What counts as a filler word?
Filler words are sounds and words that fill a pause without adding meaning, such as "um," "uh," "like," "you know" and "so" at the start of a sentence. Researchers group them with other disfluencies, such as restarts and repeated words. They appear in nearly all unscripted speech, including speech from confident, experienced people.
Psychologists Herbert Clark and Jean Fox Tree put it bluntly in their 2002 paper: in conversation, "fluent speech is rare." A 2016 study of job interviews defines disfluencies as "phenomena that interrupt the flow of speech and do not add propositional content to an utterance." The same paper cites an earlier estimate of about 6 disfluencies per 100 words.
So the goal for an interview is a rate that doesn't distract from your answer. Zero is neither realistic nor needed.
Are "um" and "uh" always a bad sign?
No. Clark and Fox Tree argued that "uh" and "um" are words speakers choose to announce a delay, with "uh" for a minor one and "um" for a major one. On that view, a filler tells the listener you are still working on what to say and want to keep your turn.
Their paper in Cognition also found large differences between people. Among 65 speakers in a corpus of British English who each said more than 1,000 words, the rate ran from 1.2 to 88.5 fillers per 1,000 words, with a median of 17.3. Speakers who paused more also used more fillers.
The same paper found fillers were most common at the start of a phrase, where speakers have to plan what comes next. In the corpus, the combined rate of "uh" and "um" was 43 per 1,000 opportunities at a phrase boundary, 27 after the first word and 13 later in the phrase. That pattern tells you where to look in your own answers.
How many filler words is too many in an interview?
The research checked for this page found no limit set specifically for job interviews. The closest evidence is a 2024 experiment on recorded speeches, where 5 fillers per minute did not hurt how listeners rated the speaker and 12 per minute did. Treat that as a rough guide, and count your own rate from a recording.
The study, by Matthew Laske and Florence DiGennaro Reed in the Journal of Applied Behavior Analysis, built speeches with filler sounds or filler words at 0, 2, 5 and 12 per minute. Crowdsourced survey respondents rated each speech. According to the study's abstract, higher rates, especially filler sounds at 12 per minute, lowered ratings across most categories, and the authors concluded that "a rate of five or fewer disfluencies per minute may be acceptable."
Two limits apply. The listeners rated public speeches, which are planned and one-way, while interview answers are improvised replies to a person. The raters were also survey respondents, so their judgments may differ from a hiring manager's.
| Your count in a recorded answer | How it compares with the 2024 speech study | What to work on |
|---|---|---|
| 5 or fewer per minute | At or below the rate the authors called possibly acceptable | Leave fillers alone and work on content |
| 6 to 11 per minute | Between the rates tested, so the study gives no verdict | Plan the first sentence of each part of your answer |
| 12 or more per minute | The rate where filler sounds lowered ratings most | Slow your opening and pause silently at transitions |
| Several seconds of silence before you start | Not measured in the speech study | Prepare stories ahead and use a short holding line |
Do long pauses hurt more than filler words?
In the one interview study checked for this page, delays mattered. Julie Brosy, Adrian Bangerter and Eric Mayor recorded 62 candidates answering four past-behavior questions and found that longer delays before answering predicted lower hireability ratings. Their main measure was the pause before the answer, so read it as evidence about hesitation in general.
The delays in that study ran from a fraction of a second to 35 seconds. The authors note earlier research showing that most turn changes in everyday conversation have a gap of under a second. They also found that longer delays predicted a higher chance of an inappropriate first response, which is a weak or off-target start.
A short holding line covers the gap while you choose an example. Here are some examples to copy:
"Let me think of the best example for that." "Good question. The clearest case is from my last role." "There are two I could use. I'll take the one with the bigger stakes."
Say one, then start the story. A holding line sounds deliberate, while a string of "um"s while you search sounds like you have no example.
How do you cut filler words before your next interview?
Record a practice answer, count the fillers per minute, and fix the moments where they cluster, instead of trying to fix the habit everywhere at once. Fillers tend to bunch up at the opening and at the joins between parts of a story, since Clark and Fox Tree found them most often where a new phrase begins.
- Record two minutes of one story answer on your phone.
- Play it back, tally each filler, and divide by the minutes.
- Mark where they cluster: the opening, the move from situation to action, and the ending.
- Write the first sentence of each part and say it as written.
- Practise a silent one-beat pause where you used to say "um."
- Record again and recount.
Planning the first sentence matters most, because fillers mark a moment of planning. If you already know the words that start the action section, there is less to search for. The STAR method guide gives each part of a story a clear opening line.
Endings are the other trouble spot. Many answers trail off with "so, yeah" because there is no planned last sentence. End on the result, and stop talking.
Should you think about filler words during the interview itself?
Mostly no. Trying to suppress every "um" in real time takes attention away from listening and from the content of your answer. Do the counting and fixing in practice, then in the interview focus on the question, a specific example and a clear result. Pausing silently before you start is the one habit worth keeping in mind.
If you catch yourself rambling, the fix is structural. Stop, give the result in one sentence and check in, as the page on shortening an answer that drifted explains. A rambling answer usually has more fillers because you are planning as you go.
When should you stop working on filler words?
Stop when your count is at or below about 5 per minute and each answer has a clear point. Past that point, your time is better spent on content: a specific example, your own actions and a result with a number. Fluent delivery with no evidence still makes a weak answer.
Amazon's interview guide asks candidates for STAR stories "typically lasting 2-3 minutes each," with "precise metrics and data" that show impact. Those are the things a structured interviewer scores. The page on speaking pace covers the speed of your delivery, and the guide to how long interview answers should be covers length.
To track your counts over several practice rounds, use the method in how to review a recorded mock interview and log each round in the free debrief template. The Mock room gives you practice out loud on a clock, with each answer scored out of 10 and checked against your resume. It doesn't count filler words, so keep your own tally from a recording.