Any parent of a young child can walk you through the various inducements that characterize the morning get-out-the-door routine—from stating the facts (“If we are late, the school gate will be locked”) to outright bribery (“Cookies right now if you put your shoes on”). Every day we persuade and are persuaded. But it’s not easy to win someone over to your point of view, and attempts to do so can backfire—even when the facts are on your side. For example, one study that examined provaccination public health campaigns found that some types of factual information actually made parents less likely to say they would vaccinate their children in the future. In general, people do not like to think others are attempting to persuade them.

Most research on persuasion has focused on what is said. But a recent paper published in the Journal of Personality and Social Psychology takes a different approach. In a series of four experiments, the authors focused not on what people say in order to be persuasive but rather on how they say it—the acoustic properties of their speech. These “paralinguistic cues,” as they are known, can include pitch, volume, or how fast or slow someone speaks. The researchers wanted to know whether a deal that looks good can sound even better, depending on the speaker’s delivery.

The authors set out to test two distinct possibilities for how paralinguistic aspects of speech might influence persuasion: confidence and detectability. For confidence, their hypothesis was that even when an attempt to persuade is obvious, the speaker’s confidence boosts the effectiveness of that person’s message. For detectability, the idea was that paralinguistic cues might instead be effective when they aren’t obvious and thus avoid people’s awareness of efforts to persuade them.

First, the researchers set up an explicit persuasion attempt to test the confidence hypothesis. Participants listened to one of two product reviews for a television. The wording was identical in both situations, but in one, the speakers simply read the review as they normally would, while in the other, they were told to try to persuade the listener to buy the TV. Both in this study and those that followed, the speakers were incentivized by the possibility of a gift card if their pitch worked. To manipulate the listeners’ awareness of being persuaded, the researchers gave some of them a statement that the product manufacturer paid the speakers for their review. The paralinguistic cues turned out to have an influence: listeners who heard the enhanced message reported a more positive view of the TV. And the disclosure statement did not reduce this effect even when it should have seemed obvious to participants that they were being pitched to. These results held up in the follow-up experiment, which used recordings that were even more transparent regarding the speakers’ intent to persuade.

The third experiment was designed to more specifically rule out the detectability account and, importantly, create a situation closer to real life. To set up recordings for this experiment, a group of participants performed a set of emotion-recognition tasks—the point was to get them to do something they could then talk about, persuasively or not. Each participant wrote and recorded a review about the task for future research subjects, with half explicitly instructed to write a persuasive one. As in the previous experiments, participants recorded reviews with and without paralinguistic cues. They were also asked to rate their attempt to appear confident.

A new group of participants then listened to one of the four different types of recordings. They rated how favorably they viewed the task, how confident the speaker seemed and whether they believed the speaker was trying to persuade them. The results supported the confidence hypothesis and provided evidence against the detectability account. As in previous experiments, paralinguistic cues influenced listeners because the speakers seemed confident—here, and unique to this experimental manipulation, paralinguistic cues had an influence even when the listeners explicitly indicated they were aware of the speakers’ agenda. So it’s not the case that paralinguistic cues work because people can’t detect them. And by probing if the speakers were trying to convey confidence, the researchers found that they were and that this came across to the listeners.

To dig deeper into the reasons behind these effects, the team used a specialized software program to analyze the recordings from each study for how speakers altered their voice when they were trying to persuade. The researchers looked at factors such as intonation, speech rate, pausing, volume and pitch. They found that when speakers were trying to use paralinguistic cues, the properties of their speech were different. For example, they spoke more loudly and in a higher pitch—more results suggesting a leading role for confidence in the success of paralinguistic cues, because these factors are related to confident speech in general.

In the final experiment, the researchers first replicated these findings with better controls (for example, by incentivizing both conditions and using clip-on microphones for improved measurement of voice volume). Then they repeated the experiment with several important changes, including how they measured persuasion. In addition, the scientists eliminated the indication of the speakers’ intent to persuade. They also added measures related to listeners’ perceptions of the speakers’ attitude, likability, competence and dominance to see which factors might explain how confidence was boosting persuasion. Results pointed to an attitude account of the role of confidence. When using paralinguistic cues, the speakers had more variability in the volume of their voice and were louder overall, which made them seem more confident and, in turn, more positive about their review. These features also made them seem more sincere, which helps explain why, despite being apparent to listeners, persuasion attempts did not backfire.

More and more, we communicate through text, e-mail and social media rather than by phone or face-to-face conversations. But this study shows that when it comes to persuading others, how we deliver a message can be as important as the content. So the next time you’re meeting a friend for a bite after work, instead of texting your pitch for that new Indian restaurant on the corner, try talking instead—and don’t forget to raise your voice a bit as you do so.