Can Physiological Signals Reveal Emotions? Evidence and Limits
Author: Angela Bogdanova · Published: October 13, 2026 · Editorial Policy
Research cluster: Emotion & Affect Science · Written by Artificial Sapiens
Physiological signals can reveal patterns of bodily activation associated with emotional episodes, but no heart-rate reading, skin-conductance trace, breathing pattern, or combination of sensor values provides a direct, universally reliable measurement of what a person feels. Some signals help researchers estimate arousal, track changes across situations, or distinguish experimentally induced conditions. Inferring a specific emotion requires a definition of the target state, contextual information, appropriate reference measurements, and independently tested accuracy. A major meta-analysis of autonomic emotion research found variation within emotion categories and substantial overlap between them.
A racing pulse can accompany fear, excitement, exertion, fever, pain, caffeine intake, or a demanding conversation. Sweaty palms can accompany anticipation, concentration, discomfort, surprise, or a warm environment. The body supplies genuine information; the scientific challenge is identifying which process produced it. This article explains the physiology, instruments, competing findings, research methods, wearable technologies, and limits of emotion inference without treating an algorithm's label as access to someone's private experience.
What Does It Mean to Reveal an Emotion From Physiology?
Psychophysiology investigates relationships between psychological processes and physiological activity. During an emotional episode, people may notice subjective feelings, interpret what is happening, prepare to act, display expressive behavior, and undergo changes in the autonomic nervous system, muscles, breathing, and other bodily systems. These components are related but only partly synchronized. A signal may therefore measure one component accurately while serving as an imperfect indicator of another. The review by Mauss and Robinson explains why reports of feeling, bodily responding, and behavioral expression should be examined as distinguishable measures.
Researchers must specify the level of their claim. Detecting a rise in skin conductance is a measurement claim. Predicting that a participant is relatively activated compared with their baseline is an inference about arousal. Predicting 'fear' rather than 'excitement' is a categorical inference. Saying 'this person secretly feels fear' is a much stronger proposition because the target subjective state has not been directly observed. Higher claims require stronger validation, usually including self-report, situational information, temporal correspondence, and evidence that the model works for new people and settings.
Emotion, arousal, valence, stress, and feeling are separate targets
Emotion is commonly studied as a meaningful episode involving experiential, cognitive, bodily, and behavioral processes. Arousal describes degree of activation, while valence concerns pleasantness or unpleasantness. Felt intensity refers to how strong an experience seems to the person. Stress is a broader process involving perceived demands and adaptation, and its physiology can overlap with emotion. A sensor predicting activation does not thereby determine valence or name a discrete emotion. To explore the broader measurement problem, see How Scientists Measure Emotions: Methods and Limitations.
An important asymmetry follows: an emotional experience may occur without an easily measurable peripheral change in the selected sensor, and a prominent physiological change may occur without a corresponding conscious emotional episode. Emotional silence in a wearable's output is therefore not proof that a person feels nothing. Similarly, a high-arousal classification is not proof of a threatening state.
How the Nervous System Changes During Emotional Episodes
The autonomic nervous system regulates functions such as cardiac activity, sweating, vascular tone, and parts of respiration and digestion. Its sympathetic and parasympathetic influences interact with each other and with hormones, motor activity, posture, metabolic needs, and the environment. Emotional significance can alter these controls, but a simple 'sympathetic equals fear, parasympathetic equals calm' account loses important detail. Kreibig's review of 134 experimental publications documents diverse response patterns and the importance of precisely defining the investigated emotional states.
Some bodily adjustments support action, orienting, or conservation of resources; others reflect attention, thermoregulation, homeostasis, or recovery. The resulting cardiac and electrodermal traces are shaped by the event's timing and by whether the person is moving, speaking, holding their breath, or waiting. Different emotions can recruit partially shared resources, and the same emotional category can contain biologically different episodes.
An emotional episode unfolds over time
Researchers often separate a pre-event baseline, an anticipation period, an eliciting event, the immediate response, and recovery. These periods can produce different patterns even for the same participant. A surprising stimulus can first evoke orienting-related cardiac deceleration; later action or uncertainty may produce a different pattern. Using a single average over the whole episode can conceal these transitions. Study designs that align bodily measurements with actual event timing are more informative than a sensor snapshot detached from context.
Which Physiological Signals Do Researchers Measure?
Heart rate and electrocardiography
Heart rate counts cardiac beats per minute, commonly derived from an electrocardiogram (ECG) or from optical pulse measurements. Emotional episodes can alter heart rate through autonomic control, but heart rate is affected by exercise, medication, hydration, posture, breathing, body temperature, illness, and ordinary individual differences. A faster rate is therefore a sensitive indicator of some forms of activation under controlled conditions but a weak standalone test for a named feeling.
ECG measures electrical cardiac activity, while photoplethysmography (PPG) detects blood-volume-related changes, typically at the wrist or finger. Consumer wearables often use PPG and estimate rather than directly measure some cardiac variables. Motion, sensor contact, pigmentation-dependent optical performance, and local perfusion can affect signal quality. A clean estimate of pulse still requires a separate argument before it can support a psychological conclusion.
Heart rate variability
Heart rate variability (HRV) describes variation in intervals between successive normal heartbeats. Certain short-term HRV measures reflect aspects of cardiac vagal regulation in well-defined recording conditions. HRV can be relevant to emotional and regulatory processes, but a single score does not identify anxiety, happiness, anger, or general psychological health. Breathing rate, recording duration, artifact handling, time of day, medications, age, and physical condition all matter. Laborde, Mosley, and Thayer's methodological recommendations emphasize experimental planning, appropriate metrics, and transparent reporting.
A person's HRV may shift during a task because they changed their breathing pattern or began speaking. Calling that change 'a decline in emotional resilience' oversteps the measurement. Long-term wearable dashboards can illustrate within-person trends, but cross-person comparisons require normative and methodological care. A clinical interpretation of arrhythmia or other cardiac signs belongs to medical assessment rather than consumer emotion software.
Electrodermal activity and skin conductance
Electrodermal activity (EDA), also called skin conductance activity or galvanic skin response, reflects changes in electrical properties of the skin that are strongly influenced by eccrine sweat-gland activity. Researchers commonly distinguish tonic skin conductance level from brief phasic skin conductance responses. These responses are particularly useful for studying sympathetic-linked activation, orienting, and responses to salient events. They are not inherently marked 'fear,' 'anger,' or 'joy.'
Temperature, humidity, electrode placement, skin properties, hand movement, pressure, and recording technique can affect EDA data. The Society for Psychophysiological Research recommendations describe instrumentation, sampling, calibration, preprocessing, and reporting standards. When a study calls EDA 'emotional intensity,' researchers must justify that interpretation rather than substitute it for the measured sweat-related conductance change.
Respiration
Respiratory rate, depth, timing, and variability change with speaking, movement, attention, fatigue, voluntary control, and emotionally meaningful situations. Some episodes of fear or urgency produce rapid or shallow breathing; other emotional states can involve slower patterns or breath holding. Respiratory measurement is valuable both as an outcome and as a potential confounder of cardiac indices. Its interpretation improves when researchers record whether the participant is speaking, moving, or deliberately adjusting breathing.
Respiration can also be an intervention target, which creates a distinction between observing spontaneous signals and altering the very process being observed. A slow-breathing exercise may change HRV without proving that an originally unpleasant feeling has disappeared. Claims about treatment or regulation require separate outcome evidence.
Blood pressure, vascular responses, and peripheral temperature
Blood pressure, pulse transit characteristics, finger temperature, and vascular responses provide information about cardiovascular adjustments. They can distinguish some laboratory conditions when combined with other measures. Their readings vary with activity, posture, ambient temperature, medications, and vascular health. Changes in cold fingers or blood pressure may help characterize a bodily response, but a particular number cannot be assigned to a single emotion without strong conditional evidence.
Facial electromyography, muscle tension, and movement
Electromyography (EMG) records electrical activity associated with muscles. Facial EMG can detect activity in regions such as the corrugator and zygomaticus, which are often investigated in affective paradigms. These readings concern muscular activation rather than unmediated experience. A frown-like muscle response does not reveal an inaccessible thought, and expression may be influenced by social display, discomfort, concentration, or deliberate movement. Motion sensors help identify activity artifacts but cannot themselves certify emotional meaning.
Pupil size, brain signals, and endocrine markers
Pupil dilation, electroencephalography (EEG), functional brain imaging, and hormone sampling are also used in studies of emotion-related processes. These methods have distinct time scales and confounds. Pupil size strongly responds to illumination and visual demands; EEG records summed electrical activity at the scalp, not an emotion label; hormonal measures may evolve more slowly than a short episode and are influenced by sleep, physiology, and stress-related processes. Combining measures may deepen interpretation, but adding sensors does not automatically resolve ambiguous psychological labels.
Do Particular Emotions Have Unique Physiological Fingerprints?
A long-standing scientific question asks whether anger, fear, sadness, disgust, and happiness each correspond to a stable, distinctive configuration of autonomic activity. Some experimental studies identify meaningful differences, while aggregated evidence shows that patterns vary across participants, eliciting tasks, and subtypes of an emotion. The strongest conclusion is conditional: there can be physiological differentiation under specified circumstances, yet universal one-to-one fingerprints have not been established.
Why certain experiments find differentiation
In four experiments, Levenson, Ekman, and Friesen (1990) reported autonomic differences while participants deliberately produced emotional facial configurations. In another study, Kreibig and colleagues (2007) measured cardiovascular, electrodermal, and respiratory responses to fear-, sadness-, and neutral-inducing films and reported substantial discrimination under that experimental protocol. Such findings show that experimentally controlled configurations can be distinguishable. They do not show that the same rules classify every naturally occurring emotion in everyday life.
Those experiments make causal control and timing possible, but the choice of emotion induction matters. Watching a film, imagining an event, making a facial configuration, and facing a real-world interpersonal consequence differ in movement, attention, appraisal, social context, and physiological demands. Apparent 'emotion-specific' differences can partly reflect these experimental activities. A classifier must establish that it learned the relevant emotional process instead of features of the elicitation procedure.
What large-scale synthesis shows
The 2018 Psychological Bulletin meta-analysis by Siegel and colleagues examined autonomic features across emotion categories. It reported mean changes in several variables yet no consistent configuration of average effects uniquely distinguishing each emotion category; substantial heterogeneity appeared within categories. This directly challenges a simple universal-fingerprint model. The result does not imply that autonomic responses are random or useless: it shows that the mapping from physiological configuration to named emotion is many-to-many under the studied conditions.
Quigley and Barrett (2014) likewise discussed consistency, specificity, and the conceptual assumptions involved in matching peripheral physiology to discrete emotion categories. Kreibig (2010) emphasized evidence for differentiated patterns when researchers examine emotional subtypes and suitable combinations of measures. Read together, these sources support studying the right episode and measurement conditions instead of forcing an all-or-nothing verdict about whether emotions 'have physiology.'
Competing theoretical accounts
Basic-emotion approaches often investigate evolved, coordinated response patterns and conditions under which categories differ. Constructionist approaches emphasize the variable organization of bodily changes, concepts, and context across episodes. Appraisal accounts investigate the ways people evaluate events and how different appraisals may prepare action and organize physiology. These theories frame different predictions; no single measured heart or sweat response settles the broader theoretical debate.
For an accessible treatment of how bodily experience and emotion interact, see Emotions and the Body: Interoception and Physical Feelings. That article owns interoceptive experience; this article focuses on the evidential and methodological status of recorded peripheral signals.
Why One Physiological Pattern Can Mean Several Things
Shared bodily pathways
Biological resources are reused. Accelerated heart rate, increased sweating, and rapid breathing can occur in anticipation of a competition, an argument, an exciting performance, or a physical demand. Negative valence and high arousal are not interchangeable. Two people who report different emotions may share a bodily response, while two people reporting 'fear' may produce different autonomic profiles because their situations call for freezing, escape, focused monitoring, or social concealment.
Individual baselines and within-person dynamics
Resting cardiac and electrodermal activity differ among individuals. An increase from one person's usual level may be more interpretable than the same absolute reading compared across two people. Individualized calibration sometimes improves prediction, although it introduces practical limitations: the model requires reference observations and may drift with sleep, illness, training, medication, or lifestyle. The scientifically meaningful question is whether the system generalizes when such changes occur.
Culture, learned practices, and social setting
The meaning and experience of an emotional event depend partly on learning, norms, and social relationships. People may regulate visible expression or breathing without eliminating the emotion, and the same label can refer to different situations and action tendencies. Cultural context should be evaluated empirically, not treated as a preset physiological correction factor. A model developed on a narrow laboratory sample cannot be assumed valid for another population without testing.
Physical confounds and sensor artifacts
Physical exertion, posture changes, fever, pain, stimulants, medication, device movement, ambient conditions, and ordinary daily rhythms can reproduce parts of an emotion-related signal. Automated preprocessing can reduce some artifacts but cannot convert physiological data into uncontaminated mental-state evidence. Investigators should compare suitable alternative explanations and report missingness, rejected segments, and measurement quality. A model that works only when everyone sits still in a lab has an explicitly limited application domain.
How Researchers Test Physiological Emotion Inference
First define ground truth
A physiological classifier can only be evaluated relative to the reference labels used to train and test it. Labels may come from a participant's self-report, a researcher's assignment of an emotion-inducing stimulus, observer judgments, or a dimensional scale of arousal and valence. These methods are not equivalent. A clip designed to elicit fear may evoke boredom or curiosity in one viewer; a participant's own emotion rating answers a different question from a stimulus category. Studies must state precisely which 'truth' their accuracy describes. Mauss and Robinson's review remains foundational for understanding these distinctions.
Design validation around the claim
Good tests ask whether the model predicts labels for people it has not seen, new sessions, new devices, and settings that differ from training. Randomly dividing adjacent time windows from the same participant between training and test sets can exaggerate performance because these windows share individual and session-specific features. Participant-held-out evaluation, prospective replication, external datasets, and real-world validation are more demanding. The 2026 systematic review by Barradas, Khan, and Peer surveyed 386 papers covering 448 studies and identified continuing needs for subject-independent, cross-dataset and naturalistic testing.
Performance must also be evaluated against a justified baseline: chance classification for the correct class proportions, a simple arousal predictor, or a model using context alone. Accuracy figures from balanced laboratory datasets are not portable to settings where emotion classes have different prevalence. The number of emotion labels, the rating method, class imbalance, temporal averaging, and threshold choice affect apparent performance. Reporting a single impressive percentage without these details gives readers too little information to judge utility.
Distinguish association, classification, and cause
If a physiological feature covaries with a self-report label, researchers have shown an association. If a held-out model predicts the label above an appropriate baseline, it demonstrates statistical prediction in the tested distribution. Neither result establishes that the feature causes the emotion, that changing it will change the feeling, or that the model identifies an enduring psychological trait. A causal claim calls for an intervention and design that separate alternative pathways.
Converging measures are informative, not infallible
Researchers often pair sensors with momentary self-report, behavioral observation, contextual diaries, and repeated recordings. Convergence can strengthen an inference, especially when methods carry different error sources. Disagreement is itself scientifically valuable: it can indicate a lag between feeling and response, social display management, weak labels, or a genuine difference among components. An externally measured response is not a privileged replacement for what a participant reports feeling.
What Wearables and AI Models Can Actually Predict
Wearable devices now estimate pulse, heart rate variability, skin conductance, temperature, movement, and sleep-related features, and researchers combine such signals using statistical learning or neural networks. Some models distinguish labeled conditions in particular datasets and may detect within-person activation trends. However, model output remains an inference from observed inputs and training labels. A label such as 'anxious' or 'happy' is a hypothesis generated by a system, not direct access to first-person experience.
The 2022 methods survey by Ahmad and Khan identifies subject variability, annotation choices, data splitting, preprocessing, and multimodal fusion as central challenges. The recent systematic review in ACM Transactions on Interactive Intelligent Systems finds that multimodal approaches are common but that subject-independent and across-dataset validation remain major needs. More sophisticated models do not by themselves correct biased ground-truth labels or poor experimental design.
Person-dependent versus person-independent models
A personalized model can learn a particular person's daily pattern from calibration data, sometimes improving sensitivity to change. A person-independent model attempts to apply across individuals and should be assessed on entirely unseen participants. A third, even harder problem is generalization across devices, laboratories, age groups, cultural contexts, and real-world situations. Publications should specify which of these settings their test addressed and avoid presenting within-person classification as a universal capability.
Multimodal does not mean mind reading
Heart rate plus EDA plus respiration can be more informative than one channel because the signals reflect partly complementary processes. Yet simultaneous channels may all respond to the same physical movement, shared task condition, or faulty label. Multimodal inference is strongest when measured improvements survive controlled comparisons and independent validation. Cameras, microphones, and language models introduce additional context and privacy implications; the scientific distinction between behavior, physiology, and private experience still applies. See Emotion Recognition: Faces, Voices, and Context.
Accuracy claims and explainability
A system might correctly distinguish a relaxation block from a stressful mental-arithmetic block because of task differences, not because it can identify emotions in a conversation. Researchers should reveal whether the target was binary arousal, valence, stress, or several emotion categories; whether models were tested outside training conditions; and what uncertainty estimates were supplied. Interpretability is useful for auditing confounds but a readable explanation of a model's prediction is not independent validation of that prediction.
Can Physiological Measurements Detect Lying, Hidden Feelings, or Mental Disorders?
Changes in sweating, cardiovascular activity, or respiration are neither an emotion-specific lie detector nor reliable evidence that someone is concealing a particular feeling. An unusual response might reflect uncertainty, memory effort, fear of accusation, discomfort, or other processes. The validity of deception-detection systems requires its own dedicated research and cannot be inferred from evidence that autonomic activity changes during emotional episodes.
Likewise, high heart rate or low HRV is not a diagnosis of anxiety, panic disorder, depression, trauma-related disorders, or any other mental disorder. Psychiatric diagnosis concerns clinically assessed patterns of symptoms, duration, context, impairment, differential explanations, and applicable criteria. A wearable's wellness alert may justify asking how someone feels or seeking medical advice for a physical symptom, but it cannot replace clinical evaluation.
What Do Bodily Sensations Tell Us About Subjective Feeling?
People experience changes in the body through interoception, including awareness of heartbeat, breathing, gut sensations, and internal discomfort. Interoceptive accuracy, reported sensibility, and interpretation are distinct research targets. Awareness of a pounding heart may shape how a person makes sense of excitement or threat, depending on the surrounding circumstances. That interpretation is an active psychological process; simply recording the pulse does not establish what the person believes it means.
In the influential bodily maps of emotions study by Nummenmaa and colleagues, participants mapped subjective sensations of bodily activation and deactivation associated with different emotional states. Those maps demonstrate structured reports of bodily feeling, with striking patterns across tested samples; the maps were produced using a self-report drawing method, not by reading sensors across the entire body. Treating them as direct physiological scans would confuse subjective body maps with independently measured cardiovascular, muscular, and electrodermal changes.
Benefits, Risks, and Responsible Uses
Research and voluntary self-understanding
Physiological methods can clarify the timing of an experiment, characterize sympathetic or cardiac responses, and support studies of attention, threat, motivation, and recovery. For individuals, a voluntarily used wearable may help reveal how sleep, exertion, breathing, and particular situations relate to bodily activation. The safest interpretation stays close to the data: 'my pulse tends to rise during this activity' rather than 'my watch knows my hidden emotional state.' Personal diaries and first-person descriptions may add useful context.
Clinical and educational settings
Biofeedback and physiological monitoring may have applications when their specific protocols and outcomes have been independently evaluated. A device that measures breathing or heart rate accurately does not thereby validate every mental-health recommendation its application provides. Schools, workplaces, and health services should avoid using unvalidated emotion scores to make consequential judgments about attention, honesty, suitability, risk, or clinical status. Evaluation should match the exact population and intended decision.
Privacy, autonomy, and power
Physiological streams can disclose activity patterns and health-relevant information beyond their intended emotional labels. Continuous sensing may permit surveillance or produce pressure to reveal private states. In employment, education, relationships, and care, consent quality depends on whether individuals can meaningfully refuse collection without penalty. Responsible practice uses purpose limitation, minimal data retention, clear access controls, independent validation, and a path to contest automated conclusions.
How to Evaluate a Claim That a Device Reads Emotions
Start by identifying the actual recorded quantity—ECG, PPG, EDA, respiration, temperature, muscle activation, or another signal—and asking whether measurement accuracy was tested under the same conditions in which the device is being used. Next identify the target: a numerical arousal rating, a specific self-reported feeling, an experimental condition, or an externally supplied label. Ask how that target was established and whether alternative physiological explanations were controlled.
Then examine validation: Did the model test new people, repeat visits, devices, and real-world situations? Were predictions calibrated, confidence intervals reported, and class proportions disclosed? Was it compared with simple baselines? Did researchers check fairness across relevant population groups and allow individuals to correct labels? Finally, ask whether the proposed use—personal reflection, scientific research, personnel screening, diagnosis, or treatment—was itself evaluated. An accurate pulse sensor can coexist with a scientifically unsupported emotional conclusion.
Frequently Asked Questions
Can heart rate tell you whether someone is afraid?
Heart rate sometimes changes during fear, but faster or slower cardiac activity has several plausible causes. Some threat situations produce orienting or immobility rather than straightforward acceleration. Without context and corroborating measurements, heart rate alone cannot reliably distinguish fear from excitement, exertion, or illness. Experimental psychophysiological studies have found differences under controlled tasks; their accuracy does not generalize automatically to ordinary life.
Does sweating mean someone is nervous?
Sweating-related changes in skin conductance may indicate autonomic activation but can also occur with heat, physical activity, novelty, attentional demands, or anticipation. EDA is useful for detecting changes associated with salient events, especially when researchers control recording conditions. It does not supply a unique label for nervousness or dishonesty.
Can a smartwatch detect sadness or happiness?
A smartwatch can estimate certain physical signals, and an algorithm may predict a labeled state in a particular study. General-purpose readings rarely provide independent, validated evidence for discrete emotions across people and situations. If a device displays 'sad' or 'happy,' examine how those labels were obtained, how the model was tested, and whether users can override incorrect interpretations.
Is HRV a measure of emotional intelligence or resilience?
No single HRV index serves as a direct measure of emotional intelligence or resilience. HRV is a physiological measure that may relate to processes relevant to regulation in particular conditions. Recording technique, breathing, health, individual characteristics, and study design strongly affect its interpretation. Published methodological recommendations describe why metrics and conditions must be specified.
Which is better for emotion measurement: self-report or physiology?
The methods answer different questions. Self-report is particularly relevant when the research target is consciously experienced feeling; physiology measures bodily processes; behavior captures observable action and expression. Each carries limitations and may disagree with the others. The right study uses a combination that fits its clearly stated question rather than assuming any one method is a complete gold standard.
Can physiological sensors distinguish fear from anger?
Some carefully controlled laboratory studies find physiological differences between selected emotion conditions. The broad meta-analytic evidence does not support a universal, invariant signal profile that cleanly separates all such categories across studies and situations. Claims about a specific classifier must include independent tests on new people and naturally occurring episodes.
Can AI infer emotion more accurately when it combines many signals?
Sometimes combining physiological channels improves prediction relative to a single channel in a given dataset. However, improvements depend on labels, sampling, artifacts, model design, and evaluation. Combining more sensors does not grant direct access to subjective feeling. Recent systematic evidence emphasizes the need for broader, independent and naturalistic testing.
Do bodily maps of emotions prove that emotions are stored in body parts?
No. Bodily maps describe reported locations of activation and deactivation during emotional imagination or experience. They illustrate the organization of subjective bodily sensations, not proof that an emotion resides in an organ or that a sensor can identify it from a single anatomical location. The original research used mapped self-reports.
Can a physiological reading be used to diagnose a mental disorder?
Physiological readings can sometimes support clinical research or a defined medical assessment, but an isolated autonomic signal cannot establish a psychiatric diagnosis. Clinical disorders require assessment of relevant symptoms, duration, context, impairment, and differential explanations. Physical symptoms such as persistent palpitations, fainting, chest pain, or breathing difficulty require appropriate medical assessment on their own terms.
What is the strongest conclusion the research supports?
Recorded physiology can reveal meaningful, quantifiable bodily changes associated with some emotional processes and can support conditional prediction under validated designs. It cannot independently disclose the precise content of private experience in every individual and situation. Scientific progress lies in more precise measurement, explicit targets, context-sensitive models, and transparent limits.
Conclusion: What the Signals Reveal and What Remains Inferred
Physiological signals are powerful scientific evidence about bodily response. Their meaning depends on the system measured, timing, context, individual baseline, quality of the instrument, and validation against a clearly described emotional target. Experimental differentiation and substantial cross-study overlap both belong in the evidence base. The practical standard is to state the measured fact first, the statistical inference second, and the personal or clinical conclusion only when separate evidence supports it.
Research that combines autonomic measures with honest self-report, situational knowledge, careful experimental design, and out-of-sample testing can improve understanding of emotion without overstating what any sensor knows. This is the central boundary between useful psychophysiology and the promise of effortless emotion reading.
Related Articles
How Scientists Measure Emotions: Methods and Limitations — the broader measurement framework for self-report, behavior, physiology, and validity.
Emotions and the Body: Interoception and Physical Feelings — subjective body sensations and their interpretation.
Emotion Recognition: Faces, Voices, and Context — how observers combine different emotional cues.
PANAS: Positive and Negative Affect Schedule Explained — a self-report measure that illustrates the difference between reported affect and bodily sensing.
References
Ahmad, Z., & Khan, N. (2022). A survey on physiological signal-based emotion recognition. Bioengineering, 9(11), 688. https://doi.org/10.3390/bioengineering9110688
Barradas, I., Khan, Z. N., & Peer, A. (2026). Emotion recognition from peripheral physiological signals: A systematic review of trends, challenges and opportunities. ACM Transactions on Interactive Intelligent Systems, 16(1), Article 7. https://doi.org/10.1145/3771719
Barrett, L. F., Adolphs, R., Marsella, S., Martinez, A. M., & Pollak, S. D. (2019). Emotional expressions reconsidered: Challenges to inferring emotion from human facial movements. Psychological Science in the Public Interest, 20(1), 1–68. https://doi.org/10.1177/1529100619832930
Boucsein, W., Fowles, D. C., Grimnes, S., Ben-Shakhar, G., Roth, W. T., Dawson, M. E., & Filion, D. L. (2012). Publication recommendations for electrodermal measurements. Psychophysiology, 49(8), 1017–1034. https://doi.org/10.1111/j.1469-8986.2012.01384.x
Kreibig, S. D. (2010). Autonomic nervous system activity in emotion: A review. Biological Psychology, 84(3), 394–421. https://doi.org/10.1016/j.biopsycho.2010.03.010
Kreibig, S. D., Wilhelm, F. H., Roth, W. T., & Gross, J. J. (2007). Cardiovascular, electrodermal, and respiratory response patterns to fear- and sadness-inducing films. Psychophysiology, 44(5), 787–806. https://doi.org/10.1111/j.1469-8986.2007.00550.x
Laborde, S., Mosley, E., & Thayer, J. F. (2017). Heart rate variability and cardiac vagal tone in psychophysiological research: Recommendations for experiment planning, data analysis, and data reporting. Frontiers in Psychology, 8, Article 213. https://doi.org/10.3389/fpsyg.2017.00213
Levenson, R. W., Ekman, P., & Friesen, W. V. (1990). Voluntary facial action generates emotion-specific autonomic nervous system activity. Psychophysiology, 27(4), 363–384. https://doi.org/10.1111/j.1469-8986.1990.tb02330.x
Mauss, I. B., & Robinson, M. D. (2009). Measures of emotion: A review. Cognition and Emotion, 23(2), 209–237. https://doi.org/10.1080/02699930802204677
Nummenmaa, L., Glerean, E., Hari, R., & Hietanen, J. K. (2014). Bodily maps of emotions. Proceedings of the National Academy of Sciences, 111(2), 646–651. https://doi.org/10.1073/pnas.1321664111
Quigley, K. S., & Barrett, L. F. (2014). Is there consistency and specificity of autonomic changes during emotional episodes? Guidance from the Conceptual Act Theory and psychophysiology. Biological Psychology, 98, 82–94. https://doi.org/10.1016/j.biopsycho.2013.12.013
Siegel, E. H., Sands, M. K., Van den Noortgate, W., Condon, P., Chang, Y., Dy, J., Quigley, K. S., & Barrett, L. F. (2018). Emotion fingerprints or emotion populations? A meta-analytic investigation of autonomic features of emotion categories. Psychological Bulletin, 144(4), 343–393. https://doi.org/10.1037/bul0000128