The Science of "Um": Why We Use Filler Words and What They Reveal

The Sound of Thinking

Press record on almost any conversation, then play it back. Somewhere in there — probably a lot of somewheres — you'll hear it. "Um." "Uh." "Like." "You know." Little bursts of sound that carry no dictionary meaning, sprinkled through your speech like seasoning you didn't know you were adding.

Most of us treat these sounds as flaws. Verbal litter. The thing a public-speaking coach circles in red ink and tells you to eliminate. And there's a whole cottage industry built on making you feel bad about them — count your "ums," they say, then feel appropriately ashamed.

But here's the twist that decades of psycholinguistics research keeps landing on: filler words aren't noise. They're signal. When you say "um," your brain isn't malfunctioning — it's doing something remarkably sophisticated in real time. And listeners, it turns out, are quietly using those sounds to understand you better, whether they realize it or not.

So before you wage war on every "uh" that leaves your mouth, it's worth understanding what these sounds are, why your brain makes them, and what they reveal about how fluently you speak. The goal isn't to feel bad. It's to understand the machinery — because once you do, you can work with it instead of fighting it.

What Filler Words Actually Are

Linguists have a wonderfully unglamorous term for the bumps in everyday speech: disfluencies. A disfluency is any interruption in the smooth flow of talk — and it turns out spontaneous speech is absolutely riddled with them. Estimates from conversation research suggest that a meaningful slice of everything we say out loud is technically "disfluent" in some way. It's the water we swim in; we just don't notice it.

Filler words are one specific type of disfluency. They break down into a few recognizable characters:

• Filled pauses — the classic "um" and "uh." These are the pure ones, sounds that exist for no reason except to occupy the space where silence would otherwise be.
• Discourse markers — "like," "you know," "I mean," "so," "well." These are real words doing a fake-word job. "Like" means something in "I like coffee," but in "it was, like, huge," it's functioning as a filler.
• Repetitions and restarts — "I went to the — I went to the store." Your speech stalls, backs up, and takes another run at the sentence.
• Prolongations — stretching a small word out: "annnnd then we left." The drawn-out sound buys the same time an "um" would.

What unites all of these is that they don't add information about the world. If you strip every filler out of a transcript, the literal meaning survives intact. That's exactly why they've been dismissed for so long as errors — as bugs in an otherwise clean system. But calling them bugs assumes speech is supposed to work like reading a finished script aloud. It isn't. Live speech is composed on the fly, and fillers are the visible seams of that composition.

Why Your Brain Produces Fillers

To understand why "um" exists, you have to appreciate the sheer difficulty of what you're doing every time you open your mouth to speak spontaneously.

The psycholinguist Willem Levelt spent much of his career mapping how humans turn thoughts into speech, and his influential model breaks the process into stages that all have to happen in a fraction of a second. First you conceptualize — figure out what you actually want to say. Then you formulate — select the right words from your mental dictionary and arrange them grammatically. Then you articulate — send the motor commands to your lips, tongue, and vocal cords. And you're doing all of this while already talking about the previous thought. It's like laying railway track from a moving train.

Fillers show up at the seams where one of those stages hits a snag. Usually it's the middle one — formulation. You know the idea you want to express, but the exact word or the way to phrase it hasn't arrived yet. Your brain is still searching. And here's the thing: you have two options in that moment.

1. Go silent while you search — and risk the other person thinking you've finished, or lost the thread, and jumping in.
2. Make a sound that says "hold on, I'm still assembling this" — and keep your claim on the conversational floor.

That's what an "um" really is: a placeholder that buys planning and retrieval time without surrendering your turn to speak. It's a tiny act of traffic control. Your production system needs a beat to finish loading the next word, and rather than leave dead air, it emits a sound that means "still working — don't go anywhere." Far from being a failure of speech, it's an elegant solution to a genuine timing problem. The same mechanism that produces the occasional stumble is the one that lets you speak in real time at all.

Um vs. Uh: A Real Difference

Here's where it gets genuinely surprising. "Um" and "uh" feel interchangeable — random noises you make without choosing between them. But research on natural conversation suggests they're not interchangeable at all. They appear to signal different things.

The pattern that keeps turning up in analyses of recorded speech is this: "uh" tends to precede a shorter delay, and "um" tends to precede a longer one. When a speaker is about to pause only briefly, they lean toward "uh." When they're anticipating a bigger hold-up — a harder word, a more complicated thought to assemble — "um" is more likely, often followed by a longer silence.

In other words, your brain seems to be forecasting the size of its own delay and picking a filler that fits. That's a small marvel. Somewhere below conscious awareness, your speech system is estimating "this is going to take a second" versus "this is going to take a while," and encoding that estimate into which vowel-and-consonant combination comes out.

And listeners appear to read the signal. Studies of comprehension have found that people process what follows an "uh" or "um" slightly differently depending on which one they heard — treating the filler almost like a heads-up about what's coming. It's a genuinely two-way system:

• The speaker unconsciously broadcasts how much trouble the next stretch of speech is giving them.
• The listener unconsciously uses that broadcast to calibrate their attention and their patience.

None of this is planned. Nobody decides "I'll deploy an um here to signal a two-second delay." It's all automatic, which is precisely what makes it such good evidence that fillers are part of the design of human conversation rather than a defect in it.

Fillers Aren't Always Bad

Given all of that, it shouldn't shock you that a moderate amount of filling is completely normal — and sometimes even helpful. Let's be balanced about this, because the anti-um crusade tends to overstate its case.

Listeners use fillers as cues. A well-placed "um" tells your audience that something considered or complex is coming, which can actually make them lean in. Research on memory and attention has found that words following a filler are sometimes remembered better, possibly because the filler flags "pay attention, effortful content ahead." Your hesitation can function as an underline.

Fillers hold your turn. In a fast-moving group conversation, going completely silent to think is an invitation for someone to steamroll you. A quiet "so..." or "I mean..." keeps the floor yours while your brain catches up. Strip those out entirely and you may find yourself getting interrupted more, not less.

Perfectly smooth speech can sound less trustworthy. There's a reason a flawless, filler-free delivery can feel rehearsed or slippery — like a scripted sales pitch. A few natural hesitations signal that you're thinking in the moment, responding to this conversation rather than reciting. Authenticity has a texture, and mild disfluency is part of it.

So the honest takeaway is not "fillers are bad." It's that fillers are a normal feature of live human speech that becomes a problem only past a certain threshold. The question was never whether you say "um" — everyone does. The question is whether you're saying it so often that it starts to work against you.

What Excessive Fillers Reveal About Fluency

So where's the line? When does a natural, useful filler tip over into something worth addressing?

The tell isn't the presence of fillers — it's the ratio. When a large share of your speech is filled pauses, restarts, and "you knows," it usually means one thing: your word-retrieval system is running behind the rest of your speech, and the fillers are the sound of it trying to catch up. Each "um" is your mouth buying time your brain hasn't earned yet.

This connects directly to that maddening experience of reaching for a word and finding nothing there — the tip-of-the-tongue state, where you can practically feel the word without producing it. If you want the full picture of what's happening in your brain during those blanks, we dug into it in why you forget words mid-sentence. The short version: excessive fillers and word-finding blanks are two symptoms of the same underlying thing — a retrieval process that isn't keeping pace with your intended speech rate.

A few patterns tend to travel with heavy filler use:

• They cluster before "content" words. Watch where your fillers land. They rarely appear before "the" or "and." They pile up right before the noun, the name, the specific term — exactly the words that require an active search of your vocabulary.
• They spike under pressure. High-stakes moments — an interview, a presentation, a disagreement — load your working memory and slow retrieval, so the fillers multiply precisely when you least want them to.
• They increase with unfamiliar topics. The less established your vocabulary in a subject, the more your brain has to hunt, and the more it fills the gaps with sound.

The encouraging part is that this is a fluency issue, and fluency is trainable. Excessive fillers aren't a fixed personality trait or a sign that you're "just not a words person." They're a readout of retrieval speed — and retrieval speed responds to practice the same way any timed skill does. Speed up the search, and the fillers shrink on their own, because you stop needing to buy so much time.

How to Use Fewer Fillers Without Sounding Robotic

Now the practical part. Notice the goal here is fewer, not zero. Chasing zero fillers is how people end up sounding stiff, over-rehearsed, and weirdly robotic — you replace natural hesitation with unnatural rigidity, which is worse. The aim is to get your retrieval fast enough that you simply don't need as many. Here's how.

Embrace the silent pause. This is the single highest-leverage change, and it's counterintuitive. The instinct is that silence sounds worse than "um." It doesn't. A confident, deliberate pause reads as thoughtful and in control; a string of "ums" reads as flustered. Next time you feel a filler coming, try just... closing your mouth for a beat. The silence feels enormous to you and barely registers to your listener. Trade the filled pause for a clean one, and you sound more composed instantly — no retraining required.

Slow down your overall pace. Many fillers come from outrunning your own brain — talking faster than your retrieval system can supply words. Ease off the accelerator and you give formulation time to keep up, which means fewer gaps to fill. Slower speech isn't just clearer for your listener; it's structurally less prone to disfluency.

Prepare your openings. Fillers cluster at the start of turns, when you're launching a thought from a cold start. If you know you'll have to speak — a meeting update, an answer to a predictable question — rehearse just the first sentence. Getting a clean launch off the line carries momentum into everything after it and heads off that opening cluster of "um, so, yeah, basically."

Record yourself. You cannot fix what you can't hear, and most people genuinely have no idea what their filler pattern sounds like. Record a two-minute voice memo describing your day or explaining something you know well, then listen back. It's uncomfortable the first time. But you'll learn where your fillers cluster — and awareness alone reduces them, because you start catching the impulse in the moment.

Train the underlying retrieval. All the tactics above manage fillers in the moment. To actually reduce the need for them, you have to speed up the word search that's causing them — and that takes reps. This is exactly what timed word-generation practice is built for: forcing your brain to pull words quickly, under mild time pressure, until fast retrieval becomes the default. It's the same principle that a few focused minutes a day with Flowency is designed around — short, science-backed exercises that train your brain to find words faster, so fewer gaps open up in the first place. Manage the symptom with pauses; cure the cause with practice.

Put those together and something quietly satisfying happens. You stop policing every "um" and start speaking from a system that's simply quicker on its feet. The fillers don't vanish — you'll always be human, and a few are part of sounding real. They just stop running the show.