When we listen to a moving piece of music or feel the gentle pulse of a haptic vibration, our bodies react before we consciously register the feeling. The heart may quicken, palms may sweat resulting in subtle electrical resistance variations in the skin. These changes, though often imperceptible, reflect the brain's engagement with the world. A recent study by researchers at NYU Tandon and the Icahn School of Medicine at Mount Sinai and published in PLOS Mental Health explores how such physiological signals can reveal cognitive arousal — the level of mental alertness and emotional activation — without the need for subjective reporting.
The researchers, led by Associate Professor of Biomedical Engineering Rose Faghih at NYU Tandon, focused on skin conductance, a well-established indicator of autonomic nervous system activity. When sweat glands are stimulated, even minutely, the skin's ability to conduct electricity changes. This process, known as electrodermal activity, has long been associated with emotional and cognitive states. What distinguishes this study is the combination of physiological modeling and advanced statistical methods to interpret these subtle electrical fluctuations in response to different sensory experiences.
This research work started as a course project for student authors Suzanne Oliver and Jinhan Zhang in Faghih's "Neural and Physiological Signal Processing.'' Research Scientist and co-author Vidya Raju mentored the students under the supervision of Faghih. James W. Murrough , Professor of Psychiatry and Neuroscience and Director of the Depression and Anxiety Center for Discovery and Treatment at the Icahn School of Medicine at Mount Sinai also collaborated in this research.
Taking Prof. Faghih's class was a great experience and allowed me to combine coursework and research," said Oliver. "It was very exciting to see the work I did in class could help improve treatment of mental health conditions in the future."
The researchers analyzed a published dataset of participants' continuously recorded skin conductance measured while they were exposed to visual, auditory, and haptic stimuli. Participants also provided self-ratings of arousal using the Self-Assessment Manikin, a pictorial scale that quantifies emotional states. By applying a physiologically informed computational model, the team separated the slow and fast components of the skin's electrical response and inferred when the autonomic nervous system was most active. Bayesian filtering and a marked point process algorithm were then used to estimate a continuous measure of cognitive arousal over time.
The analysis revealed a striking pattern: the nervous system responded most strongly within two seconds of a new stimulus, with haptic sensations eliciting the largest immediate activations. Yet when the researchers compared these physiological signals to participants' own self-assessments, auditory stimuli — particularly sounds and music — were most often linked to high arousal states. This suggests that the brain's perception of stimulation and the body's raw autonomic responses, while related, may not always align perfectly. However, when the physiological signals were further processed into estimates of user arousal, the modelled arousal agreed with the participant's assessment that auditory stimuli caused the highest arousal.
Interestingly, the model was able to track transitions in participants' arousal levels as they moved from low- to high-intensity stimuli with an accuracy exceeding random chance. When the participants who felt more stimulated by visual cues were analyzed separately from those more responsive to touch, the model's predictions revealed the significant differences in participants' responses to these stimuli in the self-reports effectively capturing group trends.
The implications of this work extend beyond the laboratory. In clinical contexts, self-reported measures remain the gold standard for assessing mental states such as anxiety or stress, yet they are inherently subjective and often unreliable. Objective metrics derived from skin conductance could complement these reports, offering clinicians a more nuanced view of a patient's emotional dynamics in real time. Such tools might one day aid in monitoring recovery from depression, anxiety, or post-traumatic stress disorder, where changes in physiological arousal often mirror symptom fluctuations.
The study also points to potential uses in virtual reality and human-computer interaction. By quantifying how users react to visual, auditory, or tactile elements, systems could adapt dynamically — heightening immersion, enhancing focus, or reducing stress depending on the goal. This closed-loop feedback between body and machine could make digital environments more responsive to human emotion.
Still, the authors acknowledge the complexity of translating sweat and associated signals into precise emotional understanding. Factors such as stimulus duration, individual variability, and prior experience complicate the interpretation. The correlation between computed arousal and self-reported ratings was modest overall, reflecting the intricate and personal nature of emotional experience. Yet the model's consistency in identifying moments of heightened engagement underscores its promise as a complementary measure of internal states.
In essence, the study bridges a subtle gap between physiology and perception. By grounding emotion in the body's own electrical rhythms, it invites a more continuous, data-driven view of how humans experience the world — one that may eventually inform both mental health care and the design of emotionally intelligent technologies.
