Is dictation faster than typing? What the research measured

For getting raw words down, speaking is usually faster than typing, and the best-known study found speech about three times faster than a phone keyboard. That study measured short phrases on an iPhone, not people at a desktop keyboard, and it did not measure writing something original. On a Mac, whether dictation beats typing for you depends on your typing speed and, above all, on how long you spend fixing the dictated text. The honest answer is to time yourself, and the test at the end takes ten minutes.

What did the Stanford study find?

The study most often quoted for dictation vs typing is Ruan, Wobbrock, Liou, Ng and Landay, "Comparing Speech and Keyboard Text Entry for Short Messages in Two Languages on Touchscreen Phones", by researchers from Stanford University, the University of Washington and Baidu. It was first posted in August 2016 and published in the Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies in December 2017.

What they did, according to the paper:

• Participants: 48 university students aged 19 to 32. Twenty-four were native speakers of American English and 24 of Mandarin Chinese.
• Device: an Apple iPhone 6 Plus, comparing its built-in touchscreen keyboard with Baidu's Deep Speech 2 speech recognition running on a server.
• Task: copying short phrases shown on screen, 50 test phrases per method after 10 practice phrases. The English phrases were 16 to 37 characters long and had no punctuation.
• Setting: a quiet meeting room, seated at a table.

What they found:

So speech was 2.93 times faster in English and 2.87 times faster in Mandarin, and it left slightly more errors in the finished text. Participants could fix speech errors by voice or with the keyboard, and the authors report that most preferred the keyboard for corrections.

Why does that not settle it for a Mac?

The authors are careful about the limits, and each one matters for a person at a desk.

1. It was a phone keyboard. Thumbs on glass, not ten fingers on keys. The paper compares speech with a touchscreen keyboard only.
2. It was copying, not composing. The authors chose transcription to measure what they call "upper-bound performance" and say that composing text "would reduce outcomes for both methods, perhaps differentially." When you write an email you are thinking as well as entering text, and pauses and "um" can confuse a recogniser.
3. Short phrases with no punctuation. The paper says it tested "transcription tasks on short messages without punctuation" and calls for future work on punctuation and symbols.
4. Ideal conditions. A quiet room and a fast network. The authors suggest studying noisy conditions next.
5. One recogniser, in 2016. Results for one cloud system then do not tell you how any particular app performs today.

How fast do people type on a real keyboard?

A large study gives a baseline. Dhakal, Feit, Kristensson and Oulasvirta, "Observations on Typing from 136 Million Keystrokes" (CHI 2018, Aalto University and the University of Cambridge) analysed about 168,000 volunteers who took an online typing test, almost all of them on a desktop or laptop keyboard. The paper reports an average of 51.56 words per minute with a standard deviation of 20.2, and says the fastest typists reached 120 words per minute or more. The volunteers chose to take a typing test, so they are not a random sample of computer users, and this too was a copying task.

Put the two studies side by side and you can see the problem with a simple answer. The phone typists in the Stanford study averaged 52 words per minute and speech reached 153. A desktop typist at 50 words per minute has room to gain. A typist at 100 or more has much less, before any editing is counted.

How fast do people speak?

The Stanford paper cites earlier research that English speaking rates "can be as fast as 200 WPM." That is an upper figure for speech, not for dictating finished text. In the same study, the initial speech transcription before any error correction averaged about 179 words per minute in English, and the rate after corrections was 153. The gap between those two numbers is the cost of fixing errors, even on short phrases.

Older research went the other way. The Stanford paper summarises an earlier study in which people using speech reached 13.6 words per minute for transcription and 7.8 for composition, against 32.5 with keyboard and mouse. It also notes that correcting recognition errors by voice was prone to cascading failures, where the correction itself is misheard. The authors say recognition has greatly improved since those systems, but the lesson holds: corrections decide the result.

So what decides it for you?

Total time to a finished, correct piece of text is what counts:

time speaking + time reviewing + time correcting, compared with time typing + time proofreading.

Dictation tends to win when:

• the text is prose in full sentences, such as email, notes and first drafts
• you type slowly or with discomfort
• you know what you want to say before you start
• the words are common, so few need fixing

Typing tends to win when:

• the text is short, such as a one-line reply
• it is full of names, codes, numbers, symbols or formatting
• you compose by rewriting as you go
• you cannot speak freely where you are

How much the tool matters is a fair question, and there is no neutral answer on this page. SuprFlow publishes this guide and sells a Mac dictation app. We have not run a controlled speed or accuracy benchmark of it against typing, against Apple Dictation or against any other product, so we give no speed figure for it. Use the test below with whichever tools you are considering, including the free Dictation built into macOS.

A 10-minute test you can run today

You need a timer and a text editor.

1. Pick a realistic task. For example: write a five-sentence email to a fictional colleague, Priya Raman, moving a meeting from Tuesday 14 October at 10:30 to Thursday at 3 pm, mentioning invoice 4471 and a budget of 12,500 dollars.
2. Type it. Start the timer, write the email from scratch, proofread it, stop the timer when you would be happy to send it. Note the time.
3. Dictate it. Wait a few minutes, then do the same task by voice with Apple Dictation or the app you are testing. Start the timer before you speak and stop it only when the text is corrected and ready to send.
4. Count the fixes. Note how many corrections you made and what kind: names, numbers, punctuation, wrong words.
5. Repeat with a different task. A longer piece of prose, such as a 200-word project update, usually favours dictation more than a short message does.

Compare finished times, not speaking time. If dictation is faster for long prose and slower for short replies, that is a normal result, and it tells you where to use each.

Questions people ask

Is typing or voice dictation faster for writing emails?

For a paragraph or more of ordinary sentences, dictation is often faster to get down. For a short reply, typing usually is. Time one of each.

How many words per minute is dictation?

In the 2016 Stanford study, corrected speech entry on a phone averaged 153 words per minute in English for short copied phrases. Your rate when composing real text will be lower, because you pause to think and you fix errors.

Does dictation make more mistakes than typing?

In that study speech left slightly more errors in the final text, 1.30% against 0.79%, while typing involved more errors that were fixed along the way. Dictation errors are real words in the wrong place, so they need a read-through to catch.

What is the average typing speed?

A 2018 study of about 168,000 online volunteers found an average of 51.56 words per minute, almost all on desktop or laptop keyboards.

Related guides

• How to use voice to text on a Mac
• Improve Mac dictation accuracy
• Run a fair Mac dictation comparison
• Free vs paid Mac dictation cost
• What a 20-person dictation pilot should measure
• Best dictation app for Mac

Sources

• Ruan, Wobbrock, Liou, Ng and Landay — Comparing Speech and Keyboard Text Entry for Short Messages in Two Languages on Touchscreen Phones (arXiv abstract)
• Ruan et al. — full paper PDF
• Dhakal, Feit, Kristensson and Oulasvirta — Observations on Typing from 136 Million Keystrokes (project page)
• Dhakal et al. — full paper PDF, CHI 2018