Speech vs Typing Speed: Text Entry Rates Compared

Speech recognition is the fastest everyday way to get text onto a screen. The dataset's speech figures have a median of 138 words per minute ¹. A computer keyboard has a median of 51.6 ², a phone keyboard 35.2 ³, and handwriting on paper 19.4 ⁴. Stenography, which takes years of training, reaches a median of 355 ⁵.

The speech evidence is thin, though. This page rests on 57 data points ⁶ from 25 sources ⁷, and most of them describe typing. Speech has only two rows, both from one study.

Speech beats the phone keyboard by a wide margin in the lab

The cleanest comparison in the data comes from one controlled study in which the same people entered short messages by voice and by keyboard. In English on an iPhone, speech reached 153 words per minute ⁸, while the built-in keyboard reached 52 ⁹. In Mandarin Chinese, speech reached 123 ¹⁰ and the Pinyin keyboard 43 ¹¹.

The gap is large in both languages, so it does not look like a quirk of English.

These figures count copying short phrases in a quiet room, with an older recognizer. They do not include the time spent fixing a dictated draft, long documents or noisy places. Treat them as the speed of getting words out, not of finishing a document. Our overview of dictation covers what happens after the words appear.

Phone typing is slower than desktop typing, and it varies with how you hold the phone

The phone keyboard has the most rows in the dataset. The largest is an online study of volunteers typing on their own phones, with a mean of 36.2 words per minute ¹². A field study, in which people typed in their own apps, found 32.1 ¹³. The two designs differ, yet the results sit close together.

Within phone typing, how you hold the device makes a visible difference. People who typed with two fingers averaged 37.7 ¹⁴, and two thumbs reached 38.02 ¹⁵. Age also shows. The fastest age group, the teenagers, averaged 39.6 ¹⁶, and the oldest adult group averaged 26.3 ¹⁷. Young children were slowest, at 24.3 ¹⁸.

Language experience adds a further effect. Native English speakers averaged 37.8 ¹⁹, while people who never type in English averaged 25.6 ²⁰ on English test sentences.

Screen reader users form a separate case. Blind participants averaged 7.5 words per minute ²¹ on a QWERTY keyboard with VoiceOver, and another small study measured 1.87 ²² without word prediction. These samples are small and the two results differ greatly, so they show the range of experience more than a typical speed.

Swipe typing and word prediction did not make people faster

Gesture typing is often described as a speed feature. In the data, people who reported swipe typing averaged 32.2 words per minute ²³, which was not faster than people who used no intelligent text entry feature. Word prediction alone gave 32.8 ²⁴, also no better. In the blind-user study, prediction did not help either: 1.63 ²⁵ with it, and the difference was not significant.

Do not read this as proof that these features slow people down. The use was self-reported, not assigned, so faster typists may simply need them less.

Computer keyboards span a wide range, and the average depends on who took the test

The computer keyboard has the most rows after the phone, and the widest spread. At the slow end, a school typing course control group reached 15 words per minute ²⁶ after training. At the fast end, a championship finalist typed a standard text at 163 ²⁷, an unofficial record from a single race.

Between those extremes, the large online tests agree with each other. A very large online test averaged 51.56 ²⁸. Workers surveyed by Typing.com who used all their fingers averaged 54.6 ²⁹, and a lab study put standard touch typists at 79.99 ³⁰.

Another large source reports lower. Ratatype's test takers had a median in the range Ratatype rounds to 30 to 40 ³¹. Both groups chose to take an online test, so the difference probably reflects who takes which test, not a real conflict. Our guide to typing speed covers population averages in more depth.

Touch typing matters less than its reputation

Typists who never learned touch typing did not lose much. In one transcription task, touch typists averaged 57.8 words per minute ³² and non-touch typists 58.93 ³³, a difference the authors found not significant. Another lab study found a clear edge for touch typists, 79.99 against 65.63 for nonstandard typists ³⁴. The studies disagree, and the likely reason is who was recruited: the first group of non-touch typists were already fast, with a wide range of speeds.

Health shows in the numbers too. Patients typed at 49.7 ³⁵ before carpal tunnel surgery, slowed to 45.2 ³⁶ soon after, and reached 53.5 ³⁷ about three months later.

Handwriting is the slowest method for most people

Handwriting on paper has a median of 19.4 words per minute ⁴. University students copying text averaged 22.85 ³⁸, and a school norm for the end of middle school is 16 ³⁹. Both are copying tasks, which are usually faster than composing, so writing your own thoughts by hand is probably slower still.

The dataset holds no rows for handwriting recognition, braille input or eye typing. That is a gap in the evidence, not evidence that these methods are slow.

Brain-computer interfaces now match a good typist, with caveats

The brain-computer interface median is 52.9 words per minute ⁴⁰. Individual rows run from 30.6 ⁴¹ for decoded vocalized speech to 78 ⁴² for attempted silent speech. A participant with ALS reached 62 ⁴³ and later averaged 56 ⁴⁴ at home.

Each of these is one person with an implanted device, and the faster results carried high word error rates. This is a tool for people who cannot speak or type, not a comparison for general readers. The same applies to other assistive methods: a head mouse reached 6.3 ⁴⁵ for one participant, and a joystick method reached 14.9 ⁴⁶ after practice.

Stenography is fastest, but the figures are records

Two stenography figures exist: 350 words per minute ⁴⁷ for an early shorthand record, and 360 ⁴⁸ for a realtime record. Both are single record attempts, one a short burst, so they show what is possible, not what a working reporter sustains daily.

What the data cannot tell you

The rows are not a random sample of anyone. Many come from volunteers who chose to take an online test, which tends to raise averages. Tasks differ: copying is faster than composing, and most rows are copying or unspecified. Few rows say whether time spent correcting errors is counted, so a figure may describe raw entry, not finished text. Several rows come from older devices or an older recognizer, and there is no like-for-like study that measures speech, phone and desktop typing for the same people composing real documents.

The medians also mix very different groups. The phone median blends children, older adults and blind users, and the speech median is two numbers from one study.

What this means for choosing how to write

If you are on a phone, voice is the clearest gain in the data, since the phone keyboard is where typists are slowest relative to what speech reaches. On a computer, a typical keyboard speed is already competitive with writing by hand, and learning to touch type shows an uneven benefit in these studies. If you spend a lot of time on drafts, try dictation for the first pass and keep the keyboard for editing, because the lab figures do not include editing time. Handwriting is the slowest option in the data, so choose it for reasons other than speed.

Pick a method for your task and test it on your own writing for a week. A published average will not tell you how fast you will be.