Filler words: what they are and how to cut them

What are filler words?

A filler is a sound or word you say while you plan the next one. It holds the floor, fills the gap where a pause would go, and adds no information to the sentence. Toastmasters draws the line in a useful place: sounds are "ah, um or er", and words are "and, well, but, so and you know" (Toastmasters Ah-Counter). Linguists put fillers in a wider family called disfluency, together with repeats and repaired words, because all of them come from the same place: speech is built in real time, not copied from a page (Cambridge English).

Here are the fifteen you will hear most often in English, what each one is usually doing, and what to do instead.

Two entries need a warning. "Like" and "so" have ordinary jobs, so cutting every one of them makes you sound stiff. The test is removal: if the sentence means exactly the same without it, it was a filler. Our filler word counter counts "like" separately for that reason.

Why do we use filler words?

Because you are thinking out loud, and the filler buys a fraction of a second. Steven Cohen at Harvard Extension School traces the habit to childhood: we were taught to answer questions immediately, so silence feels like a failure, and "when you use a filler word such as 'um,' you are thinking verbally". He also names the two hot spots — the beginning of a statement and the gap between two ideas — which is where nearly every um in your own recordings will be.

Nerves and distraction do the rest: HBR's guide to cutting fillers says they arrive when you are "nervous, distracted, or at a loss for what comes next", which is why they cluster in the hardest part of a talk rather than spread evenly through it. That is the useful diagnosis. A filler is not a vocabulary problem but a timing problem: the gap arrived before the word did, and the mouth filled it.

Are filler words always bad?

No. Frequency and position decide whether a filler costs you anything. A single "so" before your conclusion can signal the turn better than a pause would, and HBR makes the case for three deliberate uses: softening delicate feedback, holding the floor when people interrupt, and breaking into a conversation without cutting someone off.

Fillers are also ordinary. In a Cambridge English Corpus study of informal British conversation, eight filler forms appeared 26,712 times per million words — about 2.67% of everything said (Cambridge English). That is not a target for a speech. It is a reminder that the goal cannot be zero, because zero is not how speech works.

Context still decides. Harvard Extension's view is blunt: fillers may be natural in everyday talk, but "they do not belong in formal presentations or speeches". In an interview or a wedding toast, every filler is a small withdrawal from the attention your meaning needs. In a conversation, nobody is counting.

How many filler words per minute is too many?

There is no agreed industry benchmark, but there is one controlled study worth quoting. Laske and DiGennaro Reed (Journal of Applied Behavior Analysis, 2024) recorded speeches with filler sounds and words at 0, 2, 5 and 12 disfluencies per minute and had crowdsourced listeners rate the speakers. A low but nonzero rate — 5 per minute — did not reduce perceived effectiveness. At the highest rate tested, 12 per minute, perceptions suffered significantly, and filler sounds did more of the damage than filler words. The authors' conclusion: "a rate of five or fewer disfluencies per minute may be acceptable".

That is one study, with recorded speeches and survey listeners judging effectiveness rather than comprehension: a tested band, not a universal threshold. It also counts per minute, while most tools count per 100 words, so convert before you compare.

How to measure your own rate

1. Record sixty seconds of speech — one story section is enough.
2. Count the fillers in the transcript, not by ear.
3. Fillers per minute = count × 60 ÷ seconds recorded.
4. Fillers per 100 words = count ÷ words × 100.

At 130 words per minute, the study's 5-per-minute line is 5 ÷ 130 × 100 = 3.8 fillers per 100 words. If your own number is at or under 3.8 and the fillers are spread out, you are inside the range that study found acceptable.

Here is what the arithmetic looks like on a real retelling. The source is the opening of the fortune cookie story, which begins:

Here is a surprise for your next Chinese dinner. The fortune cookie is not Chinese. In China, people see it as something American.

A rushed retelling of the whole story, fillers marked:

"So um the fortune cookie uh is not from China you know it's from Japan like in the eighteen hundreds um they made a cookie in Kyoto and uh a paper fortune was inside you know then um a Japanese man in San Francisco like he served them with tea uh and then the war you know Japanese bakeries stopped um and Chinese bakeries made them uh and that's why like people think it's Chinese."

That take is 75 words with 11 firm fillers (um ×4, uh ×4, you know ×3) and 3 "like"s the tool would flag as possible. So 11 ÷ 75 × 100 = 14.7 firm fillers per 100 words, which at 130 words per minute is 14.7 ÷ 100 × 130 = 19 per minute — well past the worst condition in the study. Spoken at that pace it takes about 35 seconds, and it sounds exactly like that.

The same retelling, with the pauses marked as | and the detail cut:

"The fortune cookie | is not Chinese | it started in Japan | a folded cookie with a paper fortune | then a Japanese baker in San Francisco | served them with tea | then the war came | Japanese bakeries stopped | Chinese bakeries made them | and now everyone thinks it's Chinese."

That is 45 words, 0 firm fillers, 9 pauses. Nothing is missing from the story's spine, and at 130 wpm it takes about 21 seconds — so a full minute leaves room to slow the pace, put Makoto Hagiwara's name back in, and land the ending: the 1983 court case, and the fortune cookie whose message told the judge he was "not a very smart cookie". The first take was not a vocabulary failure. It was a timing failure, and the pause fixed it.

What is a Toastmasters Ah-Counter?

The Ah-Counter is a Toastmasters meeting role: one person listens for "any overused words or filler sounds used as a crutch" and reports them during the evaluation part of the meeting. The role description asks for "overlong pauses, overused words and filler sounds", with and, but, so, you know, ah, um given as examples.

Two habits are worth stealing. Count pauses as well as fillers — replacing every um with a two-second silence moves the timing problem rather than solving it. And have the count read back to you; a number someone else delivers lands harder than one you half-noticed. If you have no club, one friend with a notebook can do it for a three-minute take: fillers, and pauses over two seconds.

How do you stop saying um and like?

You stop by replacing the filler with silence, then practising until the swap is automatic. Noticing the um while you speak does not work — your attention is on the content — so the change has to be mechanical and rehearsed.

1. Close your mouth and breathe. An open mouth invites a sound; closed lips with one breath reads as thinking. This is Cohen's "pause, think, answer" in physical form.
2. Chunk the sentence. Speak in groups of three to six words, and put a full stop between groups. "The ship sank. Two men survived. One had a raft." Chunking gives the planning time somewhere to live, so the gap never opens inside a phrase.
3. Slow down before the hard word. Most ums queue behind a name, a number or a technical term. Slow the two words before it and the difficulty arrives inside a deliberate slowdown instead of inside a sound.
4. Retell a true story in 60-second takes. Read a story, put it away, retell one section aloud, then count the fillers in the transcript. Short takes keep the failure small enough to repeat, and the count tells you whether the fix worked. The full routine is in how to practise a speech alone.
5. Keep the fillers that do work. Deliberate "so", "well" and "look" can hold a turn or soften a correction. Cut the ones that stand where a pause would, and leave the rest.

If um is your main offender, how to stop saying um goes deeper on the closed-mouth pause and includes a marked transcript with the arithmetic shown. If you suspect you are also speaking too fast — which makes every filler more likely — test your speaking rate first. A rate above roughly 170 words per minute leaves no room to buy time with silence.

Try it: Set a timer for sixty seconds. Retell the fortune cookie story out loud, record it, then paste the transcript into the counter below. Write down two numbers: fillers per 100 words, and the longest run of fillers in one sentence. Do the same retelling tomorrow. Compare the second number, not the first.

Count your filler words now

The counter looks for thirteen terms, highlights each one, and reports firm fillers per 100 words, with "like" reported separately because it has ordinary uses. It is text only: paste a transcript from your phone's voice notes or any transcription app. Nothing is uploaded, and the arithmetic is the same as above — fillers ÷ words × 100.