You have spent hours eliminating every “um” and “uh” from your presentation. You have rehearsed until the delivery is smooth. Then, a week later, nobody remembers the point.
A controlled experiment produced an inconvenient result for the pursuit of perfect fluency: stories containing six filled pauses were recalled better than fluent versions.
That does not mean every “um” helps, that frequent fillers improve a presentation, or that the stories were remembered “57% better.” But it does mean a filled pause is not automatically damage.
What the researchers tested
Scott Fraundorf and Duane Watson adapted three passages from Alice's Adventures in Wonderland. Each passage was about 300 words long and contained 14 plot points.
In the first experiment, 72 participants heard three versions:
- A fluent recording.
- A recording with three “uh” and three “um” fillers.
- A recording with six coughs matched to the fillers' duration.
Participants then retold each story. The researchers scored how many plot points they recalled. A second experiment with 42 new participants tested fillers in typical and atypical positions.
In Experiment 1, the odds of recalling a plot point were 1.57 times greater in the filler condition relative to the grand mean. Coughs reduced the odds to 0.64 times the grand mean. In Experiment 2, the odds of recall were 1.38 times greater in the two filler conditions than in the fluent condition.
Why “57% better memory” is wrong
An odds ratio of 1.57 is not the same as a 57% increase in the percentage of material remembered.
Odds and probabilities are different quantities. The paper modeled the odds that each plot point would be recalled. It did not report that participants remembered 57% more of each story.
The cough result has the same problem. An odds ratio of 0.64 does not mean recall was 36% worse in ordinary percentage terms.
What may explain the effect
The equal-length coughs let the researchers reject a simple extra-time explanation. Fillers and coughs paused the story for the same duration, but they had opposite effects.
The second experiment tested whether fillers helped because they predicted a new plot point. Recall improved even when fillers appeared in less typical locations. The authors concluded that fillers probably oriented attention back to the speech stream without always predicting exactly what would follow.
The paper also notes a trade-off. Fillers may make material more memorable while making the speaker seem less knowledgeable. Better recall is not the only outcome a speaker cares about.
What this means for speakers
Do not add fillers as a technique
The study tested six fillers in short recorded stories. It did not compare professional presentations with different filler rates or identify an ideal number.
Do not chase perfect fluency
An occasional filled pause is not proof that a message has failed. Natural speech can contain disfluencies and still be understood and remembered.
Replace distracting patterns with silence
If “um” appears between nearly every phrase, the repetition can compete with the message. A deliberate silent pause gives you planning time without creating the same audible pattern.
Use a count as evidence, not a score
Count fillers to find clusters and habits. Then listen to the relevant moments or ask an evaluator whether the pattern affected clarity.
The bottom line
The quest for perfectly fluent speech can make every hesitation feel like failure. This experiment gives us a useful correction: filled pauses are not always meaningless damage. In two controlled tasks, stories with a modest number of fillers were recalled better than fluent versions.
It does not show that more fillers are better or that a public speaker should insert them deliberately. It shows that natural speech is more complicated than a score of zero “ums.”
The free speaking.app Filler Counter helps a listener record patterns without using a microphone or turning the result into a speaking score.