How the Voice Persuades: Why the Same Words Win When They Sound Confident

By Noam Shemla · The Studies

TL;DR Van Zant and Berger asked people to read the same words twice, once normally and once trying to persuade, then played one of the two recordings to thousands of listeners. The persuasive readings won, even though the words were identical and listeners could often tell they were being persuaded. The voice worked by sounding confident: mostly louder, and with more variation in loudness.

• Paper: Van Zant, A. B., & Berger, J. (2020). How the voice persuades. Journal of Personality and Social Psychology, 118(4), 661–682.
• Design: Four experiments (one in two parts) with the same core: speakers recorded the same text twice, normally and while trying to persuade; each listener heard one recording.
• Speakers: University students, 24 to 49 per experiment, recorded in soundproof rooms; the vocal cues were measured with the Praat phonetics program.
• Listeners: 4,216 online participants across the listening experiments.
• Measures: How much listeners liked the product or task, or which task they chose; how confident, sincere, likable, competent and pushy the speaker seemed; whether they felt someone was trying to persuade them.

Most persuasion research is about what people say. Alex Van Zant, at Rutgers, and Jonah Berger, at Wharton, studied how they say it: what happens to a voice when someone tries to persuade, and whether it works.1

The puzzle they started from is that persuasion often backfires. When people notice someone trying to influence them, they tend to resist. So a voice pressed into service to persuade might be expected to fail too. The authors tested two possibilities: that vocal persuasion works only because it goes unnoticed, or that it works whether noticed or not, because it makes the speaker sound confident.

The same words, said two ways

Every experiment began the same way. A speaker sat in a private, soundproof room and read a text aloud twice: once “as you normally would” and once “in a way that can persuade” a future listener, in random order. The words could not change; only the delivery could. Speakers were entered into a gift-card draw if their recording persuaded someone.

The experiments

It worked, even when listeners knew

In all three rating experiments, listeners who heard the persuasive reading liked the product or task more. The effects are small but consistent, and they held whether or not listeners had been told the speaker had a motive. What people were told did matter on its own: the paid-review notice and the speaker’s open statement of intent both made listeners like the product less. The persuasive voice kept its effect alongside them.

Experiment 3 added a direct test of whether listeners noticed. They were asked how obvious it was that the speaker was trying to persuade them, and the persuasive readings were noticed: as clearly as reviews written to persuade (d = 0.22 against 0.24). The difference was that the persuasive voice worked and the persuasive writing did not. Liking for the task barely moved when speakers wrote to persuade (4.65 against 4.73), while the persuasive voice raised it from 4.51 to 4.86. Speakers also said they had been trying harder to sound confident when they persuaded with their voice (6.36 against 4.91 on a 7-point scale), but not when they persuaded with their words.

What the voice actually did

To find out which parts of the voice mattered, the authors measured seven acoustic features of all 188 recordings from the first three experiments, and asked two questions: which features speakers changed when they tried to persuade, and which features actually moved listeners.

Which vocal features speakers changed, and which ones moved listeners

Speakers did several things when they tried to persuade: they got louder, varied their loudness more, raised their pitch and sped up a little. Only two of those moved listeners, and they are the two volume measures. The replication with clip-on microphones, which rules out speakers simply leaning towards the computer, confirmed that trying to persuade made voices louder and more varied in loudness. The average rise was about one decibel, 71.89 against 70.89: not shouting, just more.

Put together, the pieces form a chain, shown here as the authors estimated it.

Why confidence persuades, and why it did not backfire

The final experiment used just two speakers, one man and one woman, picked because their persuasive readings changed their volume by about the average amount and changed little else. 1,313 listeners heard one recording and then chose which of two tasks to do. It is the one test of behaviour rather than opinion.

Two results stand out. Listeners again noticed the persuasion, and yet rated the speaker as more sincere, not less. And the reason confidence persuaded was not that the speaker seemed more expert or more likable. It was that a confident recommendation sounded like a more strongly held one: listeners heard how much the speaker meant it.

The authors are candid that the two speakers in this experiment were chosen, not sampled, and could be unusually good at it.

Has it held up?

The central idea fits a separate line of work. In three experiments published the year before, Guyer and colleagues altered speech rate, intonation and pitch directly and found that each changed how confident the speaker seemed, and through that, how persuasive they were. When listeners were thinking hard, the confident voice tilted their thoughts in its favour; when they were not, it worked as a simple cue.2 Different vocal features, the same conclusion: listeners read confidence in a voice, and confidence persuades.

The two papers disagree on which features carry it. Van Zant and Berger found only loudness mattered among natural recordings; Guyer and colleagues, manipulating one feature at a time, found rate, falling intonation and pitch each could. A likely reading is that many features can signal confidence, and loudness is the one ordinary speakers reach for when they try.

What this study cannot tell you

• The effects are small. Around a fifth of a standard deviation on attitudes, and five percentage points on choice, spread across thousands of listeners.
• Listeners heard a recording once, online, about a product or a short task. A face-to-face conversation, where the speaker can be questioned, may work differently.
• Every message was positive. Whether a confident voice also strengthens a warning or a criticism was not tested, as the authors note.
• The acoustic analysis is partly exploratory. It was run across studies after the fact; the volume result then replicated, but the speech-rate result did not.
• The choice experiment used two hand-picked speakers, chosen to isolate volume, which the authors acknowledge could flatter the effect.

What a speaker can take from it

Two findings are worth carrying into any room. First, rewording to persuade did nothing here, while the same words delivered with conviction did. Second, being seen to persuade did not hurt, as long as the conviction sounded real.

The practical version is plain: say what you believe as though you believe it. In these recordings that meant a little more volume and more life in the loudness, not a performance. Listeners took it as a sign of how strongly the speaker held the view, and that is what moved them.

This article summarises a published paper, read from the authors’ posted copy of the accepted version. Every mean, effect size, interval, coefficient and sample size is the authors’ own, from the results of each experiment, Tables 1 and 2 and Figure 2. The two formulas restate measures the paper defines in words.

Measured

• Listeners’ liking for a product or task, and their choice between two tasks
• Ratings of the speaker’s confidence, sincerity, likability, competence and pushiness
• Seven acoustic features of each recording, measured with Praat

Inferred, not measured

• That confidence is the route: shown by mediation analysis, which is consistent with a cause but cannot prove one
• That volume carries the effect in everyday speech: exploratory across studies, then replicated for how speakers talk

References

1. Van Zant, A. B., & Berger, J. (2020). How the voice persuades. Journal of Personality and Social Psychology, 118(4), 661–682. https://doi.org/10.1037/pspi0000193
2. Guyer, J. J., Fabrigar, L. R., & Vaughan-Johnston, T. I. (2019). Speech rate, intonation, and pitch: investigating the bias and cue effects of vocal confidence on persuasion. Personality and Social Psychology Bulletin, 45(3), 389–405. https://doi.org/10.1177/0146167218787805