UmAh guides

What counts as a filler word?

A filler is any sound or word that holds your place in the conversation without adding meaning: um and uh, plus words like "like," "you know," "I mean," "basically," "actually," and "literally" when they are not doing their dictionary job. Hedges such as "I think," "probably," "maybe," and "kind of" are a related group that softens a claim. The test is whether the sentence loses anything if you delete the word.

The deletion test

Say the sentence without the word. If it means the same thing, the word was a filler. "I was, like, really tired" and "I was really tired" say the same thing, so that "like" is a filler. "It looked like rain" is not, because deleting "like" breaks the sentence. The same word can be a filler in one place and a real word in the next, which is why any honest count has to look at context rather than just matching strings.

Filled pauses: um, uh, er, ah

These are the core group. They carry no content at all; they are sounds you make while deciding what comes next. Clark and Fox Tree, working from large recordings of spontaneous speech, argue that uh and um are conventional English words that speakers use to announce a delay, with uh signaling a short one and um a longer one¹. Everyone produces them, and rates go up when the planning load goes up: describing something hard, or being the one who has to lead the conversation². They are also the ones Toastmasters' Ah-Counter role was created to track, along with overlong pauses and crutch words³.

Discourse markers used as filler: like, you know, I mean, so, right

These are real words with real jobs. "Like" compares. "You know" checks that the listener is with you. "I mean" introduces a clarification. "So" draws a conclusion. "Right" asks for agreement. They become fillers when they stop doing those jobs and start holding your place. The giveaway is position and frequency: "so" at the start of every sentence, "right?" after every clause, "like" before every adjective. A sentence-initial "So, the budget" is usually a filler; "So the budget is gone" is usually not.

Intensifiers that have gone flat: basically, actually, literally

These three were meant to add emphasis or precision and now mostly add length. "Basically" rarely simplifies anything that follows it. "Actually" is a filler unless it is correcting something. "Literally" is a filler unless the thing was, in fact, literal. They are worth counting separately from um, because they are habits of word choice rather than of planning, and they respond to a different fix: notice the word, and cut it in the sentence after.

Hedges: I think, probably, maybe, kind of, sort of, or so

Hedges are not strictly fillers, because they do change the meaning: they make a claim softer. That is sometimes exactly right. "Probably Thursday" is honest if you are not sure. The problem is reflexive hedging, where "I think" and "kind of" wrap claims you are actually sure about. Listeners pick up on it. Experiments on the feeling of another's knowing found that filled pauses and hesitant delivery lead listeners to judge the speaker as less certain of their answer⁴, and a hedge does the same thing in words. Count them, then decide one by one which are earning their place.

What is not a filler

A silent pause is not a filler; it is the thing you replace fillers with. Repeating a word while you restart a sentence is a disfluency but not a filler. Back-channel sounds from the listener, like "mm-hm" and "yeah," are not fillers either; they are how the listener tells you to keep going. And a deliberate "let me think about that" is a sentence, not a crutch.

Do fillers actually hurt?

Less than the folk wisdom says. Listeners paying attention to content often do not register ums at all, though fillers still nudge overall impressions⁵. Hearing uh can help listeners recognize the next word faster⁶, and fillers placed before new material improved recall of a story in one experiment, while a cough of the same length did not⁷. A review of the literature concludes that listeners are highly sensitive to hesitations without there being much evidence that speakers produce them intentionally⁸. The practical reading is that a low rate is invisible and a high rate is a pattern people notice. UmAh's default target of three fillers per minute is the app's own starting point rather than a figure from research; measure your own rate first and set the goal relative to that. The guide on how to stop saying um covers what to do once you have a count.