Average Texting Speed vs Keyboard: What the Research Shows

Almost every number you will read about typing speed comes from somewhere unverifiable. This topic is an exception, and that is why it is worth writing about carefully.

Two studies from the same research group — Aalto University, the University of Cambridge and ETH Zürich — measured typing at a scale nothing else in this field approaches. One recorded 168,000 people typing on physical keyboards. The other recorded 37,370 people typing on phones, across more than 160 countries.

Together they answer the question people actually ask — how much slower is a phone, really — and several they did not think to ask, including one finding about predictive text that runs directly against intuition. This article works through both, and then through what the two numbers say when you put them side by side.

The headline numbers

Start with the comparison everyone wants. The mobile study measured an average of 36.2 words per minute with 2.3 percent uncorrected errors. The desktop study, using the same broad methodology on a much larger sample, put physical keyboard typing at roughly 52 words per minute.

A gap of around sixteen words per minute is smaller than most people guess. Ask someone to estimate how much slower a phone is and they will usually say half, or worse. The real answer is that a phone costs you roughly a third of your speed — and that a competent phone typist beats a poor keyboard typist comfortably.

It is worth being precise about what these figures are. Both are averages across large, self-selected but international samples, and both measure transcription of provided text rather than composition. They are not claims about what any individual can do, and the spread around them is wide — the desktop study reported a standard deviation of roughly 20 WPM, which is enormous relative to the mean.

How these studies were actually run

It is worth spending a moment on methodology, because the reason to trust these figures over the ones circulating elsewhere is entirely about how they were produced.

Both studies recruited volunteers online and had them transcribe supplied sentences while the software recorded every keystroke and its timing. That design has a clear strength: it captures what people actually did rather than what they reported doing, at a scale no laboratory study could reach. The desktop study's name comes from the volume of raw data it collected — 136 million keystrokes.

It also has limits the researchers were explicit about, and which most articles quoting these numbers omit. Volunteers were self-selected, so the samples skew toward people interested enough in typing to participate. And crucially, both measured transcription — copying text that was already written — rather than composition, which is what nearly everyone actually does with a keyboard or phone.

That second limitation matters more than it sounds. Transcription is continuous; composition is punctuated by pauses for thought that no entry-rate figure captures. So these numbers describe how fast people can move text through their hands, not how fast they produce writing. The distinction runs through everything on this page and is the reason raw speed matters less than it appears to.

What makes them worth citing anyway is that they are checkable. Sample sizes are stated, methods are published, and the papers are findable. Almost nothing else quoted about typing speed clears that bar, which is a low one.

What actually makes phone typing fast

The mobile study is more useful than the headline number suggests, because it recorded how people were typing rather than only how fast. Three findings stand out, and two of them are actionable in the next five seconds.

The two-thumb finding is the least surprising and the most useful. Roughly three-quarters of people already do it, which means a quarter are leaving measurable speed on the table for no reason other than how they habitually hold the device.

The autocorrect and prediction split

This is the finding worth the price of admission, because it separates two features people lump together and shows them pointing in opposite directions.

Autocorrect usage correlates positively with typing speed. Word prediction usage correlates negatively. Same phone, same keyboard, same person — one of these helps and the other appears to hurt.

The mechanism is worth thinking through, because it generalizes. Autocorrect is invisible until it acts: you type approximately, keep moving, and the software cleans up behind you. It removes stopping. Word prediction is the opposite — it puts options in front of you that must be read and judged, and reading a list costs attention even when accepting an item costs one tap.

So the intuition that prediction saves keystrokes is correct and beside the point. It does save keystrokes. It also inserts a decision into a process that was flowing, and for someone who already knows what word they want, that decision costs more than the keystrokes it saved. The same trade-off shows up with autocomplete on desktop, and it is discussed in what actually saves time at a keyboard.

• Correction that happens invisibly behind you removes friction and speeds you up.
• Suggestion that happens visibly in front of you adds a decision and can slow you down.
• The heavier your prediction use, the more decisions per sentence you are making.
• If you write confidently in your first language, prediction is the likelier of the two to be costing you.

None of this means prediction is useless. For a second language, for uncertain spellings, or for anyone who finds recall harder than recognition, it plainly earns its place. The point is that it is not free, which is how it is usually presented.

Swipe typing: where it wins and where it costs

Swipe entry — dragging a continuous path across the letters instead of tapping each one — is the one mobile technique where the honest answer is genuinely conditional rather than a recommendation.

It wins on common words, because the gesture shape is distinctive and the recognizer has seen it a great deal. It wins one-handed, because a thumb sweeping across a screen covers ground that the same thumb would otherwise have to reach and land on repeatedly. And it wins when your hands are unsteady, since a rough path still resolves where a mistimed tap does not.

It loses on anything the recognizer has weak expectations about: names, technical terms, deliberate misspellings, slang, and words in a second language. A wrong guess on a swipe is more expensive than a wrong tap, because you have to notice it, delete a whole word rather than a character, and re-enter it — often by tapping anyway.

• Common English words are where swipe is fastest and most reliable.
• One-handed use is the situation where it most clearly beats tapping.
• Proper nouns and technical vocabulary are where it costs more than it saves.
• The failure mode is expensive: a wrong word rather than a wrong letter.

Which is why most genuinely fast phone typists use it selectively rather than as a default — swiping the ordinary words and tapping the awkward ones. That mixed strategy is harder to describe than a rule, and it is what the speed actually comes from.

What the desktop study found that nobody quotes

The keyboard study is mostly cited for its average, which is the least interesting thing in it. Two other findings do more to explain why some people type fast.

The first is rollover. Fast typists press the next key before releasing the previous one, so keystrokes overlap rather than running in sequence, and the research found this is surprisingly prevalent and predicts speed. Typing is not a series of discrete presses; at speed it is a continuous overlapping flow.

That single fact resolves a long-running confusion about keyboard layouts. Alternative layouts genuinely reduce total finger travel, which sounds like it should produce proportional speed gains — but if movements already overlap, cutting distance does not cut time by the same amount, because the distance was being covered in parallel. The break-even arithmetic on that is in the honest math on layouts and keyboards.

The second finding is that letter pairs typed by different hands or different fingers are more predictive of speed than other factors, and that the researchers identified eight distinct typist categories varying in performance, hand and finger usage. Speed is not simply a function of using ten fingers correctly. Alternation matters more than orthodoxy, which is why some self-taught typists using six or seven fingers are genuinely fast.

That said, the ceiling argument still holds: sight typing caps out in the mid forties because visual search does not get faster with repetition. The method remains the highest-return change for anyone below that, and it is set out in the complete ten-finger method.

Why 10 to 19 year olds are the fastest

One finding in the mobile study deserves more attention than it received, because the obvious explanation for it is wrong.

The fastest age group was 10 to 19 — and they reported spending less time on their phones than the 20 to 39 group did. If speed simply followed usage hours, the older group should have been quicker. They were not, and the summary of the research on ScienceDaily flags this as one of its more striking results.

So hours are not the driver. What separates the groups is more likely when the skill was acquired rather than how much it has been exercised since. The younger cohort learned to produce text on a touchscreen as their first serious writing interface, during years when motor patterns form readily. The older cohort learned on keyboards and adapted to phones afterwards, carrying an existing pattern into a new device.

That framing has a testable implication that matches the rest of the evidence: it predicts the same effect should appear in reverse on keyboards, with people who learned to type young outperforming those who adapted later. It also predicts the two skills should not transfer, since they were acquired as separate motor patterns rather than as one general 'typing' ability. Both hold.

There is a slightly uncomfortable corollary. The generation fastest at producing text on a phone is the same generation least likely to have received formal keyboarding instruction, because schools reduced it on the assumption that screen-raised children would absorb keyboard skill on their own. What they absorbed was a different skill entirely — genuinely impressive, and not transferable.

Why the gap is closing, and which side is moving

The way this research usually gets reported is that phones are catching up with keyboards. That is true and it is only half the story, and the missing half is the more interesting one.

Phone typing genuinely improved. Autocorrect got dramatically better, two-thumb technique became near-universal, and swipe entry arrived. Those are real gains on the mobile side of the comparison.

But a population average can also close from the other direction. Formal keyboarding instruction has thinned in many school systems, on the assumption that children raised with screens absorb it — an assumption that does not survive contact with the evidence, because thumb typing transfers essentially nothing to a physical keyboard. Different movement, different layout, different number of fingers.

The detail that makes this concrete is the age finding. On phones, the fastest group is 10 to 19 — the same age band least likely to have been formally taught keyboarding. So the fastest thumb typists are, on average, the people least equipped at a keyboard. That is not a coincidence; it is what happens when a skill is acquired by immersion in one medium and never taught in the other.

For anyone teaching a child, the practical consequence is that fluency with a phone tells you nothing about whether they can type, and should not be read as evidence that instruction is unnecessary. What that instruction should look like is in how to teach a child to type.

Should you try to type faster on your phone?

For most people the honest answer is no, or at least not through practice. Phone typing is unusual among skills in that most of the available gain comes from settings and grip rather than from training.

Notice that only one of those is about typing faster. The rest are about typing less, which is consistently the larger lever once basic competence is in place — a point that holds on desktop too.

If you do want to measure yourself on a phone, the important discipline is not to compare that number against a keyboard score. They measure different motor skills, and mixing them is a reliable way to convince yourself of a change that did not happen. Why two measurements of the same person disagree so readily is covered in what makes two sites disagree about the same typing.

What these numbers do not settle

It would be easy to read a sixteen-word gap and conclude the keyboard is nearly obsolete. The entry rate is the wrong measure for that question.

Even at parity, phones remain poor for the things that surround typing: editing existing text, moving a cursor precisely, working across multiple documents, formatting, and sustaining composition over an hour. Producing characters is one part of writing and usually not the part that dominates the time.

• Entry rate says nothing about editing, which is often longer than first-draft production.
• Cursor control and text selection on a touchscreen remain substantially worse than with a keyboard.
• Sustained composition is a different activity from message-length bursts, and the studies measured transcription rather than either.
• Neither figure captures fatigue, which diverges sharply over an hour of use.

There is also the measurement caveat that applies to everything in this subject. Both studies measured transcription of supplied text, which is not what most people do with a keyboard or a phone. Composition involves pauses for thought that no transcription test captures, and those pauses are frequently the majority of the time — which is why raw speed matters less than it appears to.

That caveat cuts in an interesting direction for the phone comparison specifically. If composition pauses dominate the clock, then a sixteen-word-per-minute difference in entry rate shrinks further still in practice — two people writing the same message on different devices will finish closer together than the raw figures suggest, because both spent most of the time deciding what to say. The gap that matters in real use is smaller than the gap these studies measured, not larger.

Which makes the durable advantage of a keyboard something other than speed. It is that everything surrounding the typing — moving through a document, fixing a sentence three paragraphs up, working with two windows open — remains far easier, and none of that appears anywhere in an entry-rate measurement.

What the research does settle is narrower and genuinely useful. Phone typing is faster than people assume, two thumbs beat one finger, autocorrect helps and prediction may not, keyboard speed depends on overlapping keystrokes rather than tidy sequential ones, and being fast on a phone tells you nothing about being fast at a keyboard. In a field where most numbers have no source, five well-evidenced facts is a good haul — and the broader problem of unsourced typing claims is catalogued in the typing myths that refuse to die.