Your estimate of your own rate is wrong
Speakers underestimate how often they say "um" by a factor of three to five, which is documented in how to stop saying um. The reason is structural: you are busy producing the next sentence at the exact moment the filler comes out, so it never reaches the part of your attention that could register it.
That makes measurement the first move rather than a later one. Paste a transcript or record yourself into the filler word counter and you get a count, a rate, and a breakdown by type. People are usually calm about the total and surprised by the distribution. A speaker convinced they have an "um" problem often turns out to say "so" at the start of two thirds of their sentences.
"Um" and "uh" signal different things
Herbert Clark's work at Stanford separated them. "Uh" marks a short retrieval delay, half a second while your brain finds a word it already has. "Um" marks a longer planning delay, which usually means you are choosing between two directions rather than searching for one word. That distinction is practical. A speaker heavy on "um" has a structure problem upstream, and drilling pronunciation will do nothing for it.
Our review of filled pauses versus silent pauses carries the citations for that, along with how the two pause types are read differently by listeners and respond differently to training. The related review of filler frequency in professional speech covers how much context shifts the baseline. The same person produces very different rates in a hallway conversation and a performance review.
Where they cluster matters more than the total
Fillers are not evenly spread. They bunch immediately before the load-bearing part of the sentence: the number, the recommendation, the disagreement. That is why a three-filler answer can land worse than a nine-filler one, because the three arrived in front of your ask. Why you say "like" so much at work covers this pattern for the specific filler most people are self-conscious about, including why it shows up when you are reaching for precision rather than when you are unsure.
Look at your timeline before you look at your count. If the fillers stack in the first four seconds of every answer, you need a default opening sentence more than you need filler suppression.
The swap, and what it costs for the first two weeks
The replacement is silence. Every time a filler is about to arrive, you pause instead, which serves the same turn-holding function without the sound. The technique is straightforward and the practice is not, because a pause feels roughly twice as long from the inside as it does to the person listening. How to stop saying um has the drill and the timeline for it.
Filler reduction and rambling are the same skill at two scales. One is a half-second gap in a sentence, the other is ninety seconds of talking past your point, and both come from planning while speaking rather than before it. How to stop rambling in meetings handles the larger version.
Where to stop cutting
Zero is the wrong target and it is worth knowing why before you overtrain. Our review of speaker credibility and verbal disfluency synthesizes six decades of perception studies, and the findings are more conditional than the popular version of this advice suggests.
Two annotated transcripts show the range. Susan Cain opens her TED talk with "this might sound antisocial to you," a hedge placed on purpose to name the audience's objection before she answers it. Greta Thunberg scores 145 out of 150 on filler rate at the UN by removing hedging entirely, which suits an indictment and would be strange in a status update. Register decides which one you want.
Related habits get covered in vocal fry at work, and if you want a tool comparison, see the best apps for filler words.
