145 words per minute is not a communication strategy

Useful information? Maybe.

A communication strategy? No.

The same average rate can be perfectly comfortable for a familiar story and painful for a sentence containing three product names, a date, a condition and a changed decision. Words per minute describes how fast words passed through the recording. It does not describe how much work the listener had to do.

The better skill is local pace control: keep a natural baseline, then create a little more listener time where the information becomes dense, unfamiliar or consequential.

Two 2026 studies make the simple rule impossible

Recent evidence is useful here because it points in different directions.

Patel and colleagues studied 37 adult cochlear-implant users and 71 normal-hearing adults listening to spectrally degraded speech. In their experiment, slower sentences improved comprehension and recognition-memory accuracy and produced faster response times than faster sentences, with listener factors and context affecting the result. Effect of Speaking Rate and Sentence Context on Comprehension and Recognition Memory of Degraded Speech.

That sounds like “slower is clearer” until you read another 2026 study.

Jeong and colleagues tested young normal-hearing listeners under different speech rates and cognitive-load conditions. Fast speech harmed recognition of key content. Under dual-task load, however, both faster and slower speech could increase errors and listening effort. Both fast and slow speech can increase listening effort and impair speech comprehension in young listeners.

These studies use different populations, tasks and manipulations. They should not be averaged into a magic number.

They point to a better conclusion:

The cost of pace depends on what the listener is doing, what the message contains and what the channel does to the signal.

Why one WPM target is logically weak

Imagine two recordings at exactly 145 WPM.

Recording A:

I tried the new café yesterday. It was quiet, the coffee was good, and I will probably go back.

Recording B:

We move to release 4.7 on Thursday, unless the security waiver remains open, in which case the production window shifts to Monday the 28th.

Same average rate. Very different information density.

A listener can predict most of Recording A. Recording B asks them to hold a version number, a day, a condition and an alternative date.

If the coaching tool gives both recordings the same green pace badge, the badge has described the speaker while ignoring the task.

Local pace is more useful than global slow-down

“Speak slower” changes every part of the message.

Local pace control asks a more selective question:

Where would a little extra time help this listener recover the important information?

That time can come from several places:

• a boundary before a consequence;
• slightly more time around a name or number;
• a clean phrase ending before the next idea;
• less rushing through the critical item;
• simpler wording instead of slower wording.

Notice the last option. Sometimes the best pace intervention is editing.

The three-take test

Use a 20-second status update with one critical item.

Example:

The fix is ready. We still need the security review before Thursday’s production window.

Record three versions.

Take A — normal

Do nothing special.

Take B — globally slower

Deliberately slow the entire update.

Take C — local listener time

Keep roughly your normal baseline. Give a little extra space around security review and Thursday’s production window.

Now compare without looking at a WPM counter.

Ask:

• Which version makes the critical items easiest to recover?
• Which version still sounds natural?
• Did the globally slow version become heavy or fragmented?
• Did Take C change only what needed changing?

If you have a listener, ask them for the main point and next step after one hearing.

That test is far closer to the job than “Which recording is closest to 145?”

Pace and pauses are related, not interchangeable

A speaker can have a normal average rate and still rush every important boundary.

Another speaker can have a slow average rate and still group ideas badly.

That is why Ptichi treats thought groups as a separate control. A pause can organize a transition. Pace describes how speech unfolds over time. They interact, but solving one does not automatically solve the other.

Research suggests

The Patel study gives evidence that modestly slower speech can improve processing under degraded listening, especially in the tested cochlear-implant and vocoder conditions.

The Jeong study gives evidence that rate interacts with cognitive load and that “slower” is not a monotonic improvement even for young normal-hearing listeners.

Neither study directly validates coaching healthy professionals to change pace in meetings. Neither establishes a universal WPM target.

That limitation is the interesting part, not an inconvenience.

Our read

Pace is a perfect example of how measurement can accidentally become doctrine.

A tool can count words and seconds with impressive precision. Because the number is easy to calculate, product design starts treating it as easy to interpret.

Ptichi should keep two layers separate:

descriptive evidence: this recording averaged X WPM;

coaching consequence: for this message and listener task, a local timing change may help.

The first can be automated. The second needs context and stronger evidence.

A number can tell you what changed. It does not automatically tell you what was better.

A harder listener experiment

If you want a stronger self-test, give the listener a small secondary task.

Ask them to remember a two-digit number. Then play one of your 20-second updates once. Ask for:

1. the main point;
2. the next step;
3. the number they were holding.

Repeat later with another pace version and a different number.

This is not a laboratory-quality cognitive-load study. It is a deliberately imperfect way to expose the fact that listeners have limited attention.

Do not turn the result into a “cognitive load score.” Use it as a directional comparison.

Where this advice breaks

Some listeners genuinely need slower or clearer speech because of hearing, language proficiency, channel degradation or other factors. Some tasks reward speed. Some require deliberation.

A local-pace strategy is not a universal replacement target either.

The point is to stop pretending one average rate is an identity-level quality measure.

Transfer: new density, new decision

Use a fresh prompt:

Explain a trade-off involving one number, one condition and one recommendation.

Record it cold. Then identify the densest five seconds.

Make one local timing change there and leave the rest alone.

If that improves recoverability without making the whole answer slower, you are training control rather than obedience to a WPM band.

Continue with phrase-end stability if your real problem is rushing the final key words, or thought groups if the listener cannot hear the structure. The ten voice controls overview shows how these pieces fit together.