(Photo by fizkes on Shutterstock)
In A Nutshell
- When work gets mentally demanding, people tend to touch the lower face more, especially the cheek, chin, and nose. This pattern rises with stress and often happens without awareness.
- The team tracked tiny heat and sweat changes near the nose to index the body’s stress response. These changes increased during tougher tasks and lined up with more lower-face touches.
- Facial expressions were not very helpful here. People often looked neutral or “serious,” yet stress still climbed. Self-touch did a better job of signaling strain in this setting.
- Why this matters: a regular camera could help flag tough moments during solo, desk-based work, without wearables. It is a support signal, not a diagnosis. If you notice frequent lower-face touching, a short reset or break may help.
HOUSTON — Right now, as you’re reading this, you’ve probably touched your face at least once without realizing it. Humans touch their faces up to 800 times each day, and most of it happens completely outside conscious awareness. Scientists have now discovered that these unconscious gestures, particularly when the hand reaches for the lower face, serve as a reliable indicator of mental stress during intense cognitive work.
A research team from the University of Houston and Virginia Tech analyzed nearly 170 hours of video recordings from 10 academic researchers working in their offices over four days. Using artificial intelligence to track every face-touch combined with thermal imaging that detected stress-related perspiration, the study found something striking: people who frequently touched their chin, cheeks, and nose at the same time showed significantly higher levels of physiological stress responses.
The behavior appears to have evolutionary roots. Scientists have observed similar self-soothing gestures in other primates. The chin, nose, and forehead form what researchers call the “T-zone” of most frequent contact, areas packed with fine body hair and dense nerve endings. When mental demands increase, the hand instinctively moves toward these regions, potentially because touching these highly sensitive areas provides comfort during stressful moments.
“Lower-face self-touch is a solid indicator of sympathetic overactivity, which is a proxy of mental stress,” the researchers write in their paper presented at the 2025 International Conference on Affective Computing and Intelligent Interaction.
Traditional stress measurement relies on questionnaires, heart rate monitors, or sensors attached to the palms. Face-touching requires only observation, making it potentially useful for workplace monitoring systems. The behavior offers something unique: it tracks stress levels while simultaneously quantifying one way people cope with that stress, given the known calming effect of self-touch.
Most of us unconsciously touch our chins or cheeks throughout the day, especially if we’re stressed. (fizkes/Shutterstock)
How Researchers Tracked Face-Touching in Real Time
The research team recruited six men and four women, all academic researchers ranging from doctoral students to senior faculty members. Each participant agreed to be monitored in their university office for four consecutive days while they worked on regular tasks, primarily reading and writing for research projects.
Three types of cameras captured different information. A thermal camera measured perinasal perspiration, the subtle sweating around the nose that occurs during stress. This facial electrodermal activity serves as a remote measurement of the body’s stress response, which previously required sensors attached to hands. A standard visual camera recorded facial expressions and face-touching behavior. A ceiling camera tracked activities like smartphone use or breaks away from the desk.
Every morning before starting work, participants spent four minutes relaxing in their chairs, imagining natural landscapes. These baseline recordings allowed researchers to compare normal states against stress responses throughout the workday.
The study accumulated 602,737 rows of data, one measurement for each second of observation, providing a second-by-second account of stress levels, activities, facial expressions, and touch behaviors.
Processing this massive amount of video required automation. The research team developed a machine learning system using MobileNet, a type of artificial intelligence designed for analyzing images. The system identified anatomical landmarks on each person’s face and hands in every video frame, then calculated whether the hand made contact with specific facial regions including the cheeks, eyes, forehead, nose, and chin.
When the hand’s position overlapped with a facial region in the image, the system recorded a face-touch event and labeled it according to which areas were touched. Team members manually reviewed 2,000 randomly selected frames (200 per participant) and found the AI correctly classified touches 94% of the time.
One pattern validated the methodology: participants predominantly touched the left side of their faces, aligning with prior research showing people typically use the non-dominant hand for spontaneous face-touches. This pattern provided additional confidence in the classification system’s reliability.
Which Face-Touching Patterns Signal Mental Stress
After categorizing thousands of face-touches, the researchers examined how different touching patterns related to stress measurements from the thermal camera. They built a statistical model that accounted for various factors including the type of work being done, time spent on smartphones, break frequency, workload perceptions, and personality traits.
Five factors emerged as significant predictors of stress responses: the amount of time spent reading and writing, smartphone use, frequency of physical breaks away from the desk, perceived physical demands of the work, and the frequency of touching the chin, cheeks, and nose at the same time.
Touching multiple areas of the lower face at once showed the strongest association with stress. Not all face-touches carried equal weight. Chin-cheek-nose touches specifically correlated with stress, while other combinations showed weaker or no associations. Removing face-touch data from their stress prediction model reduced its explanatory power by approximately 20%.
The specificity matters. Touching different facial regions appears to serve different purposes. While someone might touch their forehead while concentrating or rub their eyes when tired, the simultaneous reaching for chin, cheeks, and nose seems particularly linked to mental strain.
Beyond face-touching, the model revealed other behavioral markers of stress during knowledge work. Participants spent an average of 68% of their time reading and writing, confirming the cognitive nature of their work. This core research activity was associated with elevated stress responses.
Smartphone use also correlated with stress responses. Participants checked their phones about 3% of the time, and these episodes coincided with higher stress measurements. Whether the stress causes phone-checking or vice versa remains unclear, but the association aligns with recent research linking smartphone use to stress levels.
Taking physical breaks away from the desk also associated with higher stress, occurring about once every two hours. Rather than indicating that breaks cause stress, the researchers interpret this as a coping mechanism. When people feel mounting stress, they walk away from their desks to reset, making break frequency an indirect marker of earlier stress accumulation.
Facial expressions, despite being commonly associated with emotions, provided little predictive value for stress in this context. Participants displayed predominantly negative emotions (50% of the time), neutral expressions (20%), and what appeared as sadness or the sober look people assume when concentrating (20%). Happiness appeared rarely (4%). Yet none of these expressions significantly predicted stress levels in the final statistical model.
Researchers surveyed participants using standard psychological instruments including the State-Trait Anxiety Inventory and NASA Task Load Index, which measures mental demand, physical demand, time pressure, perceived performance, effort, and frustration. Mental demand and effort ratings were high, averaging 12 and 11.6 out of 20, respectively. Physical demand averaged only 5.3, consistent with sedentary desk work. Frustration levels remained low at 6.2, suggesting an absence of strong negative feelings despite the mentally taxing nature of the work.
Having a mentally demanding job tends to lead to more face-touching. (© Prostock-studio – stock.adobe.com)
Individual Differences: Why Some People Touch Their Faces More
Individual differences emerged throughout the four-day observation period. Two participants touched their faces much more frequently than others, falling into the “high-touch” category identified in earlier psychological studies. In contrast, one participant rarely touched her face at all, suggesting individual differences in self-soothing strategies during cognitive work.
A 2022 study found that preventing people who normally touch their faces frequently from doing so impaired their memory performance and altered their brain activity patterns. The neural changes occurred before physical contact, suggesting the act of moving the hand toward the face (not just the touch itself) plays a role in emotional regulation.
Developmental research indicates these self-touch patterns fully form only after certain stages of brain development. One study comparing pre-adolescent and post-adolescent girls found distinct self-touch patterns only in the older group, with two forms serving self-regulatory purposes and one serving conversational functions.
Face-touching behavior, when measured continuously in natural work settings, tracks mental stress more reliably than facial expressions. This matters because facial expressions can be consciously controlled, while spontaneous self-touches typically occur outside awareness.
Practical Applications for Workplace Stress Monitoring
The research team emphasizes that their findings apply specifically to solitary cognitive work, not to conversations or other social contexts where face-touching may serve different functions. The naturalistic study design, while involving only 10 participants, provided depth through 170 hours of second-by-second monitoring and breadth by including researchers at different career stages with varying work patterns.
For workplace stress monitoring systems, face-touching may provide a valuable signal that could work with standard cameras, unlike physiological sensors that need physical contact with the body.
Those 800 daily face-touches aren’t just unconscious habits. They’re the body’s attempt to self-regulate during mental strain, revealing what facial expressions cannot.
Disclaimer: This article is for general informational purposes only and is not intended as professional medical or psychological advice. If you’re experiencing persistent stress or mental health concerns, please consult a qualified healthcare provider.
Paper Summary
Methodology
Researchers monitored 10 academic researchers (6 men and 4 women, ranging from doctoral students to senior faculty) working in their university offices for four consecutive days, generating approximately 170 hours of continuous recordings. Each morning, participants completed a four-minute baseline relaxation session. The research team used three types of cameras: a thermal camera (FLIR Tau 640) to measure perinasal perspiration as a proxy for sympathetic nervous system activation, a facial visual camera (Logitech HD Pro C920) to capture emotions and face-touching behaviors, and a ceiling camera (Logitech Brio) to identify activities like smartphone use and breaks. A MobileNet-based machine learning algorithm analyzed facial videos frame-by-frame to detect face-touches by identifying anatomical landmarks and calculating hand-face overlap. The system achieved 94% classification accuracy when validated against manual review of 2,000 randomly sampled frames. Participants completed daily psychometric questionnaires measuring anxiety and workload perception. The research team built a multiple linear regression model to examine relationships between sympathetic activation and various behavioral, situational, and dispositional factors, including the newly added face-touch frequencies.
Results
The statistical analysis revealed that touching the chin, cheeks, and nose at the same time showed a strong positive association with sympathetic nervous system activation, a physiological marker of mental stress (p = 0.004). Four other factors significantly predicted stress: time spent reading and writing (p = 0.008), smartphone use (p = 0.001), frequency of physical breaks (p
Limitations
The study included only 10 participants, all academic researchers, limiting generalizability to other occupations and work contexts. The findings specifically pertain to solitary cognitive work and may not apply to collaborative work, conversations, or other social settings where face-touching might serve different functions. The research used a naturalistic observation design, which provides ecological validity but limits control over variables compared to laboratory experiments. The four-day monitoring period, while providing substantial data depth, represents a relatively short timeframe and may not capture longer-term patterns or variations. The study relied on visible cameras, which could have affected participant behavior through awareness of being monitored, though the naturalistic setting over multiple days likely minimized reactivity effects. Two participants did not consent to having their facial images used in publications, restricting the types of analyses and illustrations that could be shared publicly. The classification system, while highly accurate at 94%, misclassified 6% of frames, introducing some measurement error. The study could not determine causality between face-touching and stress due to the observational design; face-touches might cause stress reduction, result from stress, or both.
Funding and Disclosures
This work was partly funded by the National Science Foundation (NSF) under award number 1704682, titled “Managing Stress in the Workplace: Unobtrusive Monitoring and Adaptive Interventions.” The study received approval from the Institutional Review Board (IRB) of the University of Houston. The authors obtained informed consent from all participants, with specific provisions regarding use of facial imagery in publications. Eight participants permitted their facial images to be used in publications, while two allowed use only for analysis. The researchers have made their code and annotated videos for consenting participants available through GitHub and the Open Science Framework (OSF) to ensure reproducibility and transparency. No competing interests were disclosed in the paper.
Publication Information
Kiran, F., Hasan, M. T., Ganeshan, K., Bhatambarekar, G., Sarkar, A., & Pavlidis, I. (2025). “Facial self-touches are associated with mental stress in knowledge work,” presented in Proceedings of the 2025 13th International Conference on Affective Computing and Intelligent Interaction (ACII). The paper includes contributions from researchers at the University of Houston’s Department of Computer Science and Virginia Tech Transportation Institute. The study analyzed data from the DKW (desk-bound knowledge work) dataset, an open-access resource associated with previous research published at the 2023 CHI Conference on Human Factors in Computing Systems. IEEE holds the copyright for the conference proceedings.