150 words per minute. That's the figure you'll find everywhere, usually credited to the National Center for Voice and Speech, and usually offered without the one piece of context that makes it useful.

150 WPM describes casual American conversation. It does not describe a conference talk, a podcast, a radio bulletin or an audiobook — all of which run at measurably different speeds, for documented reasons. If you're writing to a time limit, the wrong figure will put you minutes out.

> $ rate --silent-reading
225
WPM · reading to yourself
> $ rate --reading-aloud
183
WPM · script read verbatim
> $ rate --prepared-delivery
140
WPM · talking to an audience

There Is No Single Speaking Rate

Any source giving you one number without naming a context is oversimplifying. The defensible ranges look like this:

Context Typical WPM Source basis
Casual conversation (US English) ~150 National Center for Voice and Speech
Conversation (British English) 190 – 230 Tauroza & Allison (1990)
Conference talk (TED format) 153 – 168 Dlugan, analysis of TED talks
Radio news bulletin 170 – 190 Rodero (2012)
Reading aloud from a page ~183 Brysbaert (2019) meta-analysis
Audiobook narration 150 – 170 Industry production standard
Auctioneers, sports commentary 250 – 400 Performance, not comprehension

The slowest legitimate figure and the fastest differ by nearly a factor of three. For a ten-minute talk, picking the wrong end is the difference between writing 1,300 words and writing 1,900.

Where the 150 WPM Figure Comes From

The most-quoted number traces to the National Center for Voice and Speech at the University of Iowa, which puts average conversation among US English speakers at roughly 150 WPM. It's a reasonable figure and it's what most calculators default to.

The problem is what it measures. Conversation is not delivery. In conversation you're interrupted, you overlap, you trail off — and critically, your listener can stop you and ask. None of that is available to someone giving a presentation, recording a voiceover, or reading to camera.

> $ note --no-correction-channel

Conversational speech is listener-corrected. If someone loses the thread, they say so. Prepared delivery has no correction channel, so the speaker has to build the comprehension margin in ahead of time — and that margin is mostly slower pace and longer pauses.

The Study Worth Citing Instead

The most useful research on this is older and far less quoted. Tauroza and Allison published Speech Rates in British English in Applied Linguistics in 1990, measuring four distinct situations — conversations, academic lectures, interviews and radio monologue — rather than treating speech as one activity.

They found an overall rate around 170 WPM, but the variation between contexts was the actual finding. Conversations ran fastest, at 190–230 WPM. Radio monologue, scripted and delivered to an audience that cannot interrupt, ran considerably slower. Their working bands remain the clearest framework available:

Band WPM How it lands with a listener
Slower than normal under 100 Laboured; listeners disengage
Moderately slow 100 – 125 Deliberate; suits complex material
Moderately fast 160 – 185 Energetic; needs simpler sentences
Faster than normal above 185 Only with low information density

Faster Isn't Automatically Worse

The intuition that slower always means clearer turns out to be wrong. Emma Rodero's 2012 study in Text & Talk compared speech rate and pausing across 40 news bulletins from the BBC, Radio France, RAI and RNE, and found the optimal rate for radio news sits between 170 and 190 WPM — well above the conversational average. Below 170, recognition of the message got worse, because delivery became less dynamic and attention dropped. Above 190, listeners struggled with complex content.

> $ warning --the-real-variable

The trade-off isn't speed against clarity. It's speed against information density. The faster you speak, the simpler your sentences need to be. A subordinate-clause-heavy script at 180 WPM will lose people; the same ideas in shorter sentences won't.

News readers sustain 190 WPM because broadcast copy is written to be heard — short sentences, one idea each, names before facts. The pace works because the language underneath is doing the work.

Reading Aloud Sits Between the Two

There's a third rate that most guidance skips entirely, and it's the one that matters most if you're working from a script.

Marc Brysbaert's 2019 meta-analysis in the Journal of Memory and Language reviewed 190 studies covering more than 18,000 participants. Alongside the silent reading figures behind the 225 WPM average, it reports a reading aloud rate of roughly 183 WPM.

Faster than prepared speech, slower than conversation — and the reason is structural. When you read aloud you aren't composing. No hesitations, no false starts, no reaching for a word. What you lose is the natural pausing that comes with thinking, which is exactly why cold-read scripts often sound rushed. The full gap between reading and speaking is covered in speaking time vs reading time.

Why ReadCalc Uses 140 WPM

140 WPM is the default across the script formatter, the teleprompter and the speaking-time estimate on the main calculator. It's a deliberate choice, not a rounded guess.

It's the midpoint of the right band

Tauroza and Allison put average speech at 125–160 WPM. 140 is the middle. For someone planning a talk with no other information, the midpoint of the relevant band is the least wrong single number available.

It errs toward under-running

If your real pace is faster than 140, you finish early — recoverable, and usually welcome. If the default were 170 and your real pace were 140, a ten-minute slot would over-run by three minutes. Speaking slots are hard limits far more often than they're soft ones.

It leaves room for what you didn't script

Almost everyone speaks faster under pressure than in rehearsal — but almost everyone also adds unplanned pauses, corrections and asides. In practice those cancel out and then some.

> $ adjust --your-own-pace

140 is a starting point, not a prescription. Both the teleprompter and the script formatter let you set your own rate. Once you know your real pace, use it — a personal figure beats any published average.

Speaking Rate by Speech Length

The practical version. Word counts at three paces:

Duration Slow (130 WPM) Average (140 WPM) Fast (160 WPM)
1 minute ~130 ~140 ~160
2 minutes ~260 ~280 ~320
7 minutes ~910 ~980 ~1,120
20 minutes ~2,600 ~2,800 ~3,200
30 minutes ~3,900 ~4,200 ~4,800

Subtract for anything that isn't you talking. Slides, demos, questions and the pause after a point lands all eat into the count. For a talk with visuals, taking 10–15% off the word target is a sensible starting adjustment.

How to Measure Your Own Rate

Take a passage of your own writing — natural register, not a rehearsed script. Paste it into the calculator for an exact word count, record yourself reading it at the pace you'd use in front of an audience, then divide by the elapsed minutes. Use at least 300 words: shorter samples are dominated by how you start, and everyone starts fast.

Most people land between 120 and 165 WPM. Above 170 you're not necessarily too fast, but your sentences need to be doing more work — run the text through the sentence pacing analyser and look for the long ones. And if your material is dense, the answer isn't to slow down; it's to rewrite it so it can be heard, which is the same conclusion the audiobook comprehension research reaches from the listener's side.

Research Sources

Paste your script to get a speaking-time estimate — at 140 WPM or your own measured pace.

$ open script-formatter →