Tone of voice—not facial expressions—may be the best way to figure out what someone is feeling.

Speedy Internet connections and cheap video calling have made face-to-face interaction easier than ever. But, a new study suggests that audio-only conversations may offer the clearest communication.

Body language and facial expressions have been extensively studied for the emotions they convey in conversations. But that’s precisely why they can be more deceptive, says Michael Kraus, assistant professor of organisational behaviour at the Yale University School of Management.

Previous studies showed that people are better at reading emotions when presented both audio and facial expressions than when they’re asked to observe facial expressions alone. But how voice-only communication ranked was unclear.

For the study in American Psychologist, researchers recruited participants online and presented them with short videos of a group of friends talking and teasing each other over a nickname. Participants were given one of three versions: one group watched and listened to the video, a second only heard the interaction, and a third group only saw the video but did not hear the voices.

They were then asked to estimate what emotions they thought the friends were experiencing, by rating feelings such as amusement, embarrassment, or happiness on a scale of 0 to 8. People who only heard the interaction—but did not watch the video — made more accurate estimates of what the friends were feeling.

In a subsequent experiment, researchers recruited undergraduate students to come to the lab and chat with each other about their preferences for movies or TV shows, and what food and drinks they liked. The students also had similarly themed conversations in a darkened room. Then, they were asked to rate their own and their partners’ emotions during both exchanges. Participants who couldn’t see each other in the darkened room fared better at reading their partners’ emotions.

Finally, the researchers presented online participants with a digital voice reciting the friends’ teasing interaction from the prior study.  But the artificial voice was the worst.

“The difference between emotional information in voice-only communication by a computer versus a human voice was the largest across all studies,” Kraus says. “It’s really how you speak — not just what you say — that matters for conveying emotion.”

One reason the voice is so effective at conveying emotion may be that speakers are less likely to be able to alter their tone to disguise their feelings.

Another possible explanation stems from our cognitive capabilities. When communicating across multiple modes, a listener must focus on many kinds of information at once: facial expressions, words, body language, and the speaker’s tone.