People search this constantly: do "ums" lower your TOEFL Speaking score? It is one of the first things test takers worry about, and most advice tells you to stamp out every filler. So we went to the data and actually counted.
We looked at the fillers and the silence in more than 1,400 scored TOEFL Speaking answers on My Speaking Score. The result is not what most prep advice would predict. The um, on its own, carries no real penalty. The thing that quietly drags scores down is the opposite of an um. It is dead air. This post shows the numbers, explains why they come out this way, and tells you what to fix instead.
Quick answer
Filler words like "um" and "uh" do not meaningfully lower your TOEFL Speaking score. In our data, answers with many fillers looked slightly weaker at first, but that gap disappeared completely once we accounted for how much the speaker paused and how much they said. Silence is the real cost. The least-silent answers averaged around 4.4, while the most-silent answers averaged around 3.5, a gap of almost a full point. An um means you are still talking. Silence means you stalled. Stop policing your fillers and go after the dead air.
What we measured
We took scored answers from the interview task, where people speak freely and fillers and pauses show up naturally. For each answer we had the score on the 0 to 5 item scale, a measure of how many filler words the speaker used, and a measure of how much of the answer was silence. Then we compared them.
Two quick notes so the numbers are clear. The scores here come from the My Speaking Score AI scorer, which we use to find patterns across thousands of answers. And "filler rate" and "silence" are measured per answer, so a long answer and a short answer are compared fairly.
First look: do fillers track with lower scores?
At a glance, it looks like fillers might hurt a little. Here is the average score by how many fillers an answer contained.
So the heaviest-filler answers scored about a quarter of a point below the filler-free ones. If you stopped here, you would tell everyone to kill their ums. But a raw gap like this is exactly the kind of number that fools people, because filler-heavy answers tend to have something else going on at the same time. They tend to be shorter, and they tend to have more silence. So the real question is whether the um itself is the problem, or whether it is just riding along with the real problem.
The test that settles it: same length, more fillers
To isolate the um, we grouped answers by how much the person said, then compared few-filler and many-filler answers inside each group. If fillers really hurt, the many-filler answers should score lower at every length. They do not.
Look at how small those differences are, and how they even flip: among long answers, the many-filler group scored a touch higher. When we put fillers, silence, and length into one model together, the independent effect of fillers on the score was essentially zero. The apparent penalty in the first table came from something else. Filler-heavy answers were also more silent and shorter, and those are the things that actually move the score.
This lines up with the research outside our data too. Studies of second language speech consistently find that filled pauses have a negligible relationship with proficiency ratings, while silent pauses, especially in the middle of a phrase, are far more diagnostic of a lower level. Our numbers say the same thing in a TOEFL setting.
The real cost: silence
Now here is the feature that does move the score. When we sorted answers by how much of the response was silence, the pattern was strong and steady.
The answers with the most silence scored almost a full point below the answers with the least, a difference of about 0.9 on the 0 to 5 scale. That is a large gap, and unlike the filler gap, it holds up when you control for everything else. The more of your answer is dead air, the lower it tends to score.
Why an "um" is fine but silence is not
The reason comes down to what each sound tells the listener.
An um means you are still going. It holds your turn, it keeps the flow moving, and it signals that more is coming. It is a small bump in otherwise continuous speech. The listener barely registers it, and the scoring barely registers it either.
Silence is different. A long silent gap in the middle of an answer reads as a stall. It looks like you lost your place, ran out of ideas, or froze. It breaks the flow, it shortens your answer, and it leaves the impression of a speaker who could not keep going. That is what pulls the score down, and it is why chasing your ums is aiming at the wrong target.
There is a useful way to think about it. An um is the sound of a speaker who is still moving. Silence is the sound of a speaker who stopped. The test rewards the one who keeps moving.
What to fix instead
Stop policing your fillers. The effort you spend trying to sound perfectly smooth is better spent making sure you never go quiet. Here is where to put that effort.
Walk in with a plan so you are never searching mid-answer. Answer the question in your first sentence, give one reason, then move to a real example. When you always know your next move, the long silent gaps do not open up.
Prepare a few real stories in advance. When you feel a stall coming, go straight to a prepared example, because a story keeps talking on its own. You fill the space with content instead of dead air.
And if a filler slips out while you think, let it. A quick um that keeps you moving is far better than three seconds of silence while you try to be flawless.
How to practice this
Do not practice by recording yourself and counting ums. That trains the wrong habit and makes you more self-conscious, which usually creates more pausing, not less.
Instead, record an answer and mark every silence longer than about two seconds. Those gaps are your real problem spots. For each one, ask what you needed in that moment: a plan, a reason, or a ready example. Then practice that answer again with the goal of keeping the sound going, filler and all, until the long gaps are gone.
Do this across several questions and you will feel the shift. You stop fearing the um, and you start protecting the flow.
See your own silence
The fastest way to find your dead air is to look at your own answer. Take a free practice test on My Speaking Score, answer an interview question, and check how much of your response was silence and where the gaps fell. If a large share of your time is silent, you have found something very fixable, and it matters far more than any filler you were worried about.
FAQ
Do filler words like "um" lower your TOEFL Speaking score?Not meaningfully. In our data of more than 1,400 scored answers, fillers had essentially no independent effect on the score once we accounted for silence and answer length. The um itself is not what costs you.
What actually lowers a TOEFL Speaking score, then?Silence. The answers with the most dead air scored almost a full point lower than the answers with the least. Long silent gaps read as stalling and break the flow the scoring rewards.
Should I try to remove all pauses from my answer?No. Short, natural pauses are normal. The problem is long silent gaps in the middle of your answer, the ones that signal you lost your place. Aim to eliminate those, not every pause.
Is it better to say "um" or to stay silent while I think?A brief um is better than a long silence. The um keeps your turn and signals you are still going, while silence reads as a freeze. Keeping the sound moving protects your fluency.
Why do people think fillers hurt their score?Because filler-heavy answers often also have more silence and less content, so the fillers get the blame for a penalty the silence caused. When you separate the two, the fillers come out clean.
How can I reduce silence in my TOEFL Speaking answers?Have a plan before you speak, answer, reason, example, and prepare a few real stories you can pull from. When you always know your next move, the long gaps stop opening up.
Does research support this?Yes. Studies of second language speech generally find that filled pauses have a weak relationship with proficiency ratings, while silent pauses are more diagnostic of a lower level. Our TOEFL data shows the same pattern.
The takeaway
The question people ask is whether ums lower your TOEFL Speaking score, and the honest, measured answer is no. Across more than 1,400 answers, the um carried no real penalty once we looked closely. The cost was silence, which pulled scores down by almost a full point between the quietest and the most fluent-sounding answers. An um means you are still going. Silence means you stopped. Stop policing your fillers, and go kill the dead air.
Sources for the external research referenced above:
- Yan et al. (2025), "Diving Deep Into the Relationship Between Speech Fluency and Second Language Proficiency: A Meta-Analysis," Language Learning: https://onlinelibrary.wiley.com/doi/full/10.1111/lang.12701
- "Effects of speaking task and proficiency on the midclause pausing characteristics of L1 and L2 speech," Studies in Second Language Acquisition, Cambridge: https://www.cambridge.org/core/journals/studies-in-second-language-acquisition/article/effects-of-speaking-task-and-proficiency-on-the-midclause-pausing-characteristics-of-l1-and-l2-speech-from-the-same-speakers/FFE4407BEAF2A0E55D4D797C9AFE80DD
- "A longitudinal study of filled pauses and silent pauses in second language speech": https://www.researchgate.net/publication/332554262_A_longitudinal_study_of_filled_pauses_and_silent_pauses_in_second_language_speech