GLOSSARY · THE DISTINCTION THAT MATTERS
Filler words vs. discourse markers.
In one paragraph
A filler word buys time and carries no meaning — “um,” “uh.” A discourse marker organizes what you are saying — “so,” “first,” “however,” “entonces.” The same word can be either: “so” is a filler when it stalls and a signpost when it introduces a conclusion. The test is whether it does a job in the sentence.
Filler word
Holds your place.
Carries no meaning and does no grammatical job. Delete it and the sentence loses nothing but the stall. “Um,” “uh,” “er,” and the emptied-out “like” and “basically.”
Discourse marker
Organizes your ideas.
Signals structure — sequence, contrast, conclusion. Delete it and listeners lose the thread. “First / second / finally,” “however,” “so,” “in short,” “entonces.”
The same word, two jobs
Watch a word flip from stall to signpost.
This is why the word alone can’t tell you. The context does — and in speech, the tell is the pause: a stall is flanked by hesitation or repeated, a signpost is not.
| Word | As a filler (stall) | As a discourse marker (job) |
|---|---|---|
| “so” | “So… yeah… so I think it works.” — trailing, holding the floor while you think. | “So, in conclusion, the data supports it.” — introduces the takeaway. |
| “entonces” | “Entonces… este… no sé.” — a stall flanked by hesitation. | “Entonces, la solución tiene dos partes.” — signposts the next point. |
| “well” | “Well… well… I mean…” — buying time before an answer. | “Well, there are two sides to this.” — frames a considered reply. |
| “o sea” | “Es, o sea, o sea, complicado.” — repeated, empty, gapped. | “Cuesta $0, o sea, es gratis.” — reformulates to clarify. |
Why it matters for feedback
A flat word-list scores one word two opposite ways.
Most filler counters match a fixed list of words. Put “entonces” or “so” on that list and every occurrence is charged as a disfluency — including the ones that are doing the work of organizing your talk.
That is a real contradiction, not a rounding error. The very same “entonces” gets penalized as a filler and, on any measure of structure, is exactly the signposting a clear speaker is supposed to use. One word, scored in two opposite directions. A well-organized speaker is punished for being organized.
The honest fix is to judge the word by what surrounds it. When timing shows a hesitation gap around it, it read as a stall; when it connects two ideas cleanly, it is a signpost and should be left alone — or credited. That is the approach behind why English speech tools misscore Spanish speakers.
Keep these
Signposts worth keeping — in both languages.
These do a job. When they are not surrounded by hesitation, they are structure, not filler — leave them in.
English
Español
FAQ
Common questions
What is the difference between a filler word and a discourse marker?
A filler word buys time and carries no meaning — “um,” “uh.” A discourse marker organizes what you are saying — “so,” “first,” “however,” “entonces.” They can be the same word: “so” is a filler when it stalls and a signpost when it introduces a conclusion. The test is whether the word does a job in the sentence or just holds your place.
Can the same word be both a filler and a discourse marker?
Yes — that is exactly what makes them hard to count. “So,” “well,” “entonces,” and “o sea” are stalls in one sentence and structural signposts in the next. What decides it is not the word but the context: a marker connects two ideas and usually has no hesitation around it; a filler is surrounded by a pause or repeated.
How can you tell a filler from a discourse marker?
Ask what the word does. If you can delete it and the logic between your sentences still holds, it was probably a filler. If removing it makes the connection between two points disappear — the “first / then / so” scaffolding — it was a discourse marker doing real work. In speech, a stall is also usually flanked by a pause; a marker is not.
Are discourse markers bad in a speech?
No — they are how listeners follow you. “First… second… finally,” “however,” “in short” tell an audience where they are in your argument, which is the opposite of a verbal tic. Cutting them to sound cleaner usually makes a talk harder to follow. The goal is to lose the empty stalls, not the signposts.
Why do filler-word counters get this wrong?
Most counters match a flat list of words, so they charge every “so” or “entonces” as a filler regardless of the job it is doing. That double-penalizes a well-structured speaker: the signpost is counted against them as a disfluency while the structure it builds goes unrewarded. Judging the word by its context — the pause around it — is the fix.
See which of yours are stalls.
Record a 30-second speech and get your filler count — free, in any browser, in English or Spanish. You can start without an account.
Try SpeakUp Coach — Free