How to Stop Saying Um and Cut Filler Words
Filler words are not a confidence problem. They are what your mouth does while your brain looks for the next sentence. Fix the gap and the fillers go with it. Here are seven drills, ordered from the one that works in a single afternoon to the one that takes a month.
Key takeaways
- Fillers appear when your mouth outruns your brain. Slowing down removes most of them without any drill.
- Replace um with silence, not with another word. Removing it without a replacement leaves you stranded.
- Awareness alone cuts the rate by roughly half. Record two minutes and count.
- Aim for under about one filler per 100 words in a prepared talk. Zero sounds robotic.
Why do I say um in the first place?
Fillers are not a character defect and they are not a confidence problem. They are what linguists call disfluencies, and they show up at exactly the moments when you are still choosing your next word but do not want to give up the floor. The sound is a signal to the listener: I am not finished, please wait.
That matters because it tells you what the fix is. If you delete the um without dealing with the gap that produced it, the gap is still there and something else will fill it. Speakers who train the sound away without training the pause in usually end up with a substitute instead: so, right, you know, or a long drawn-out and.
The real target is the pause. A speaker who is comfortable with two seconds of silence has nowhere for a filler to live.
How many filler words are actually a problem?
Work on spontaneous speech by researchers including Herbert Clark and Jean Fox Tree puts filler rates in natural conversation at roughly 2 to 5 per 100 words for most adults, and argues that uh and um function as signals rather than noise. In everyday talk listeners barely register them, and they can even aid comprehension by flagging that something complicated is coming.
A prepared talk is different. The audience has nothing else to attend to, expectations are higher, and clusters stand out. As a working target:
| Rate | How it reads to an audience |
|---|---|
| Under 1 per 100 words | Polished. Nobody notices anything. |
| 1 to 3 per 100 words | Normal and conversational. Fine for most business settings. |
| 4 to 6 per 100 words | Noticeable. Starts to read as under-rehearsed. |
| Over 6 per 100 words | Distracting. People begin counting instead of listening. |
Do not chase zero. Speech with no disfluencies at all sounds recited, and audiences trust it less, not more.
Drill 1: Record two minutes and count
This is the highest-return thing on the page and it takes about ten minutes. Record two minutes of yourself explaining something you know well, with no script. Play it back with a pen and put a mark on the paper for every filler.
Two things happen. You get an honest baseline, which is almost always higher than people guess. And you build the awareness that does most of the work: simply knowing your own rate tends to halve it within two weeks, because you start catching the fillers as they leave your mouth.
Repeat weekly. Watch the number, not your feelings about the number.
Drill 2: Practice the two-second silence
Silence feels far longer from the inside than from the outside. Speakers routinely rate their own two-second pause as a catastrophe and listeners rate the same pause as confident emphasis. Until you have felt that discrepancy, you will keep filling the gap.
The drill: talk for 60 seconds on any topic, and at the end of every sentence, stop completely, count one and two in your head, then start the next sentence. It will feel absurd. Record it and play it back, and it will sound measured.
Once two seconds is comfortable, you no longer need a filler, because you have something else to do with the gap.
Drill 3: Slow the whole thing down by 10 percent
Most fillers come from speaking faster than you are thinking. Cutting your pace slightly gives your language system time to find the next phrase before your mouth arrives at it.
A comfortable rate for a talk is roughly 130 to 160 words per minute. Time yourself against a script you have counted. If you are over 175, pace is likely the whole problem, and no amount of filler-specific work will help until you fix it.
Drill 4: Kill the sentence-opening so
So at the front of a sentence is the most common substitute filler among professionals, and it is stickier than um because it sounds like real language. Same with basically, obviously, essentially and I mean.
The fix is to plan your first three words. Fillers cluster at the start of sentences, especially the start of answers, because that is where you are doing the most thinking. If you know how a sentence begins, the front of it is protected. In a Q and A, buy the time openly instead: "That is a fair question" is a complete sentence, and it is not a filler.
Drill 5: Get a counter in the room
Ask a colleague to raise a finger quietly every time you say a filler while you run through your talk. It is uncomfortable and it works fast, because the feedback arrives in real time instead of after the fact.
This is close to what a Toastmasters club does with the Ah-Counter role, where one member tracks and reports every filler used in the meeting. If you want that feedback loop on a regular schedule, see where to get real speaking practice.
Drill 6: Fix the structure, not the mouth
If you know exactly what comes next, you rarely need to fill anything. Speakers who ramble produce far more fillers than speakers working from a clear three-point structure, because rambling means constantly deciding where to go.
Write your talk down to six to ten keywords in a fixed order and rehearse the transitions specifically. Transitions are where filler clusters live: the moment you finish one point and reach for the next. Scripting just the joining sentences ("that covers the cost side, now the timeline") removes most of them.
The five-session practice routine has a session dedicated to transitions for this reason.
Drill 7: Breathe at the full stops
Filler rates climb when you are short of air, because a shallow breather reaches the end of their capacity mid-sentence and produces a sound to hold the floor while topping up. It is one reason nervous speakers fill more.
Mark the natural breathing points in your notes and take a real breath at each one, from the belly rather than the chest. The breath itself creates the pause you were trying to install in drill 2, so the two reinforce each other. Slowing the breath is also the fastest way to bring nerves down before you start.
A note on video calls
Fillers get worse on video, for a mechanical reason. Latency means you cannot tell whether someone is about to interrupt, so speakers hold the floor with sound to avoid being cut off. Add slight audio compression and every um is preserved perfectly.
Two adjustments help. Speak about 10 percent slower than you would in a room, and say the shape of your answer up front: "Two things on that." Announcing the structure means nobody jumps into your pause, which means you can afford to have one. More on this in presentation skills.
Frequently asked questions
How many filler words are normal in a speech?
Research on spontaneous speech by Clark and Fox Tree puts fillers at roughly 2 to 5 per 100 words for most adults. Under about 1 per 100 words in a prepared talk reads as polished. Zero sounds robotic, so do not chase it.
Is saying um actually bad?
In conversation, no. Listeners barely register it. In a prepared talk it matters more, because the audience has nothing else to focus on and a cluster of ums reads as unprepared.
How long does it take to stop saying um?
In my coaching, most speakers roughly halve their filler rate within two to three weeks of recorded practice, because awareness alone does most of the work.