Nonverbal communication includes gaze, facial expression, body movement, vocal delivery, and interpersonal distance. The statistics below come from different studies and should be read as study-specific descriptive results, not as one population estimate for all communication.
Contents
- Eye gaze and eye contact
- Facial expressions and emotion recognition
- Bodily expressions of emotion
- Gestures, posture, and movement
- Voice, pauses, and vocal cues
- Interpersonal distance
Eye gaze and eye contact
Eye behavior is not one simple measure. Researchers distinguish looking toward a partner’s face from mutual eye contact, in which both people look at each other’s eyes. A 2018 study of 51 getting-acquainted conversations found that mutual face gaze occupied an average of 63% of conversation time. The range was wide: from 27% to 88% across conversations. Episodes of mutual face gaze lasted an average of 2.2 seconds. (Using dual eye tracking to uncover personal gaze patterns during social interaction)
Mutual eye contact was much less continuous. It averaged 12% of conversation time, with individual episodes averaging 0.36 seconds. Across conversations, mutual eye contact ranged from 0% to 46%. Participants also gazed away from their partner’s face during 29% of speaking time and 10% of listening time. In a separate student sample from the same study, perceived mutual eye contact averaged 72% of conversation time; that result uses a different sample and measure, so it should not be combined with the eye-tracking estimate. (Using dual eye tracking to uncover personal gaze patterns during social interaction)
The same 2018 study reported that average fixation duration on most facial regions was less than one second. For speakers and listeners, shifting gaze can therefore be a normal part of interaction rather than evidence that attention has disappeared.
Conversation role also changes gaze. In three-person native-language conversations studied in 2017, speakers gazed at listeners during 27.9% of utterance time. Listeners gazed at the current speaker during 45.4% of utterance time, at the next speaker during 48.9%, and at the other participant during 42.5%. Speakers looked toward the participant who spoke next during 38.4% of utterance time and toward the other participant during 19.3%. (Difference in Eye Gaze for Floor Apportionment in Native- and Second-Language Conversations)
| Gaze measure | Study result | Context |
|---|---|---|
| Mutual face gaze | 63% of conversation time | 51 getting-acquainted conversations, 2018 |
| Mutual eye contact | 12% of conversation time | Eye-tracking conversations, 2018 |
| Speaker gaze at listeners | 27.9% of utterance time | Three-person native-language conversations, 2017 |
| Listener gaze at current speaker | 45.4% of utterance time | Three-person native-language conversations, 2017 |
| Listener gaze at next speaker | 48.9% of utterance time | Three-person native-language conversations, 2017 |
A gaze-transition model predicted floor allocation with 43.6% accuracy in native-language conversations and 52.9% in second-language conversations. These are model-performance results, not claims that gaze alone determines who speaks next. (Difference in Eye Gaze for Floor Apportionment in Native- and Second-Language Conversations)
Facial expressions and emotion recognition
Facial expressions communicate emotion, but recognition varies by emotion and by the metric used. In a 2020 large-scale study of 1,249 participants, mean agreement between displayed and selected facial emotions was 82.0%. Using the study’s unbiased hit rate, recognition was highest for happiness at 93.2% and lowest for anger at 55.9%. The rates for other expressions were 68.4% for neutral, 61.9% for sadness, 65.7% for fear, 68.9% for disgust, 75.0% for surprise, and 58.1% for contempt. (Recognizing Emotions through Facial Expressions: A Largescale Experimental Study)
The study also reported mean agreement rates of 82.8% for women and 79.3% for men. Those figures describe agreement in that study’s participant sample; they do not establish a universal difference in emotion-recognition ability.
The choice of measurement matters. With the study’s non-unbiased agreement measure, happiness reached 96.8%, while anger was 67.7%. Because these values use a different agreement calculation from the unbiased hit rates, they should not be presented as directly interchangeable rankings. (Recognizing Emotions through Facial Expressions: A Largescale Experimental Study)
For speakers, the practical point is that a smile may be easier for observers to identify than subtler or more ambiguous expressions. Neutrality, anger, contempt, sadness, and fear had lower unbiased hit rates in this study, so a presenter should not assume that an audience will interpret every restrained facial cue consistently.
Bodily expressions of emotion
Emotion recognition is not limited to the face. A 2021 study examined whole-body displays with Mayangna participants. Recognition rates were 86% for sadness, 73% for anger, 47% for fear, and 54% when participants selected the neutral whole-body display. These figures describe recognition in the study’s task and population, not a general accuracy rate for all observers or body movements. (Beyond Face Value: Evidence for the Universality of Bodily Expressions of Emotion)
The same source reported a highly isolated 30-person subsample separately. In that subsample, recognition was 77% for sadness, 65% for anger, 43% for fear, and 50% for the neutral display. The smaller subsample is especially important to label because its results should not be treated as equivalent to the broader Mayangna findings. (Beyond Face Value: Evidence for the Universality of Bodily Expressions of Emotion)
| Whole-body display | Mayangna participants | Highly isolated 30-person subsample |
|---|---|---|
| Sadness | 86% | 77% |
| Anger | 73% | 65% |
| Fear | 47% | 43% |
| Neutral | 54% | 50% |
Across both sets of results, fear was recognized less often than sadness or anger. That pattern is descriptive of the reported experiments. It does not mean that a particular audience will read a speaker’s posture or movement in the same way, especially when context, culture, and display intensity differ.
Gestures, posture, and movement
Hand movement was prominent in a 2013 analysis of 1,903 coded observations from 11 informed-consent sessions. Hand gestures comprised 77% of observations. Within the coding scheme, general hand waving accounted for 29%, highlighting gestures for 20%, specific conversational hand gestures for 13%, resting gestures for 10%, and repositioning gestures for 5%. Head gestures comprised 12%, body-posture shifts 5%, and eye-gaze observations 5%. These categories are study-specific coded observations, so the percentages describe the recorded coding scheme rather than a share of speaking time. (The Role of Nonverbal and Verbal Communication in a Multimedia Informed Consent Process)
In an online medical communication skills course observed in 2023, all 10 lecturers leaned forward toward the camera and 90% sat upright during online feedback. The sample was only 10 observed lecturers, and the percentages represent the share of lecturers observed exhibiting a behavior, not the share of session time. (Beyond words: analyzing non-verbal communication techniques in a medical communication skills course via synchronous online platform)
For public speaking, these findings support separating gesture frequency from gesture meaning. A coded hand movement may be a wave, a highlight, a conversational gesture, a resting gesture, or a repositioning movement. Likewise, leaning toward a camera and sitting upright were observations in a particular online teaching setting, not universal rules about how every presenter should sit.
Voice, pauses, and vocal cues
Vocal delivery carries nonverbal information through intonation, pace, pauses, emphasis, and fillers. In the 2023 online-feedback study, all 10 observed lecturers used a dynamic and cheerful voice intonation. Word fillers such as “urmm” or “okay” were observed for 100% of the lecturers, and all 10 used pauses while giving online feedback. Word emphasis was observed for 10%, throat clearing for 10%, and fast speaking for 0% of lecturers. (Beyond words: analyzing non-verbal communication techniques in a medical communication skills course via synchronous online platform)
These percentages are the share of 10 lecturers observed exhibiting each behavior. They do not indicate how long the behavior lasted, how often it occurred within each session, or what percentage of all online lecturers use the same cues. That distinction is essential when applying a small observational study to presentation practice.
Recognition of vocal emotion can also differ between groups. A 2019 study reported mean vocal-emotion recognition accuracy of 68.33% in an autism spectrum disorder group versus 83.95% in a typical-development comparison group. The result is specific to the study’s groups, task, and operational definition of vocal-emotion recognition. (The Relation Between Vocal Pitch and Vocal Emotion Recognition Abilities in People with Autism Spectrum Disorder and Typical Development)
Interpersonal distance
Physical distance is another nonverbal signal, and preferred distance changes with the social task. In a 2022 UK/Germany scenario study, preferred distance averaged 1.36 meters for joint action and 1.55 meters for parallel action. Seventy-seven percent of participants preferred a smaller distance for joint than parallel action. Across five scenarios, the joint-versus-parallel preferred-distance differences ranged from 0 to 0.35 meters; the largest difference in Experiment 1 was 0.347 meters in the Storm Damage scenario. (Smaller preferred interpersonal distance for joint versus parallel action)
Experiment 2 produced somewhat different averages: 1.49 meters for joint action and 1.60 meters for parallel action. In that experiment, 68% of participants preferred a smaller distance for joint than parallel action. The difference between experiments is a reminder that scenario design and sample context matter when interpreting interpersonal-distance statistics. (Smaller preferred interpersonal distance for joint versus parallel action)
For public speakers, distance should therefore be treated as context-dependent rather than as one fixed ideal. A collaborative activity may be associated with a different preferred distance than two people working alongside each other, and the reported UK/Germany results should not be generalized to every audience or venue.