Should you actually stop saying um?

Start with the claim you've heard: filler words make you sound unprepared, so remove them.

It's stated everywhere, almost never with a citation. When you go looking for the evidence, what you find is considerably more interesting — and it argues for a different kind of practice than the one usually prescribed.

What the listening research found

In 2001, Jean Fox Tree published a study in Memory & Cognition examining how listeners actually process um and uh in running speech. The method was to take recorded speech, digitally remove the fillers, and measure how quickly listeners recognised the words that followed.

If fillers were noise, removing them should have helped. It didn't.

So the most-hated sound in public speaking is, at the level of comprehension, either neutral or mildly useful. "Uh" appears to function as a signal — a small flag that says a word is coming and I'm still retrieving it — and listeners use that flag rather than tripping over it.

This should genuinely lower the stakes. If you said "um" four times in a presentation, nobody's understanding suffered. The catastrophe you replayed that evening did not occur.

So why do we count them?

Here is where we have to be honest about our own metric, because the research above is an argument against it, and it deserves an answer.

Three reasons, in order of how much they matter.

1. Rate is a different variable from presence

Fox Tree studied fillers appearing at natural rates in ordinary speech. That is not the same phenomenon as someone saying "um, so, like, you know" nine times in twenty-five seconds. A filler every so often is a processing signal. A filler every six words is a texture, and it becomes the most noticeable feature of the speech — not because it impairs comprehension, but because it dominates attention.

The research tells you the floor isn't dangerous. It doesn't tell you the ceiling is fine.

2. Fillers are a symptom worth reading

This is the part that actually matters, and it reframes the whole metric.

A high filler rate is rarely a filler problem. It's almost always a sign that you started a sentence before deciding how it would end, and you're generating sound while you work it out. The "um" isn't the failure — it's the audible trace of the failure, which happened about a second earlier when you opened your mouth without a destination.

Which means the fix isn't to suppress the sound. Suppressing the sound while keeping the habit just gets you a person who starts sentences without a destination and then goes silent in the middle of them. The fix is to decide the shape of the sentence first — and when you do that, the fillers fall away on their own, as a side effect.

We count them because they're the most visible, fastest-moving indicator of a habit that's otherwise invisible to you. Not because the sound itself is a crime.

3. It is the number that moves

Practically: of everything measurable in a short spoken answer, filler rate responds fastest to deliberate attention. Pace takes weeks to shift. Sentence structure takes longer. Filler rate can halve in a few days, purely from counting it. That makes it the right first target — not the most important variable, but the one that gives you evidence you can change how you speak, which is the thing most people don't believe yet.

What to practise instead

Given all that, "stop saying um" is the wrong instruction. Here's the better one.

• Close your mouth instead. When you feel the filler coming, don't replace it with a different word — replace it with nothing. Silence reads as thinking. "Um" reads as stalling. Same pause, different interpretation.
• Decide the ending before you start. The real intervention, one layer up. Before a sentence longer than a few words, know roughly where it lands. Most fillers are generated in the gap between starting and knowing.
• Name the shape out loud. "Two things." "Short version first." This buys you a legitimate second of thinking time and gives the listener structure, which is strictly better than buying the same second with a noise.
• Count before you fix. Record two minutes of yourself describing your day and count the fillers. Most people are wrong about their own rate by a factor of two or more, in both directions. You cannot fix a habit you can't hear.

The thing worth keeping from the research

If you take one thing from Fox Tree's finding, make it this: the occasional filler is not damaging you, and the anxiety you carry about it is disproportionate to any effect it has on the person listening.

That matters because filler anxiety is self-feeding. Monitoring yourself for "um" while simultaneously trying to construct a sentence is a genuine cognitive load, and it produces exactly the halting, disfluent speech you were trying to avoid. People who are relaxed about the occasional filler produce fewer of them than people policing every syllable.

Aim for the rate, not for zero. Zero isn't a goal that any fluent speaker actually hits, and chasing it makes you worse.