What is electrodermal activity (EDA/GSR) and how is it measured?

01What is electrodermal activity?

Electrodermal activity (EDA) is the general name for changes in the electrical properties of the skin that follow the activity of sweat glands controlled by the sympathetic nervous system; in research it is mostly measured as skin conductance, in microsiemens (µS). The same measurement goes by several names: GSR (galvanic skin response), long common in English sources; in Turkish, “galvanik deri tepkisi” or “deri iletkenliği yanıtı”; and, in older literature, the psychogalvanic reflex. Scientific publications today use EDA as the umbrella term and speak of skin conductance, skin resistance or skin potential according to the quantity measured (Boucsein et al. 2012).

EDA is one of the oldest and most widely used measures in psychophysiology. Because it is non-invasive, relatively inexpensive and sensitive to arousal, it is one of the core signals of research on stress, emotion, decision making and user experience. But it is important to know what it measures and what it does not: EDA tracks the intensity of arousal; on its own it does not say whether a reaction is positive or negative.

02Physiology: eccrine sweat glands and the sympathetic nervous system

The source of EDA is the eccrine sweat glands. These glands are especially dense on the palms and the soles of the feet, and they are driven by sympathetic nerve fibres. When sympathetic activity rises, the ducts of the glands begin to fill with sweat; because sweat is a good electrolyte, the conductance of the skin increases. This change can be measured even before sweat becomes visible on the surface of the skin.

Eccrine sweat glands have no parasympathetic control. For this reason EDA, unlike signals such as heart rate that both systems govern together, is regarded as a relatively pure indicator of sympathetic activity. The signal has two components:

Tonic component: skin conductance level (SCL)

The baseline that changes slowly over minutes; it relates to the person’s overall arousal at the time but can differ greatly between people.

Phasic component: skin conductance responses (SCR)

Short rises that typically begin 1–3 seconds after a stimulus, peak within a few seconds and decay more slowly.

Event-related and spontaneous responses

Responses that can be tied to a specific stimulus are called event-related SCRs (ER-SCR); those that arise without a visible stimulus are called spontaneous or non-specific SCRs (NS-SCR).

03How is EDA measured?

The most common method in research is exosomatic measurement: a small voltage is applied between two electrodes and conductance is calculated from the current passing through the skin. The publication recommendations of the Society for Psychophysiological Research (SPR) recommend the constant-voltage method with a voltage of 0.5 V for skin conductance (Boucsein et al. 2012). In endosomatic measurement no external current is applied and the skin’s own potential is recorded; this method is used less often in research.

Recording usually uses Ag/AgCl electrodes and a conductive electrolyte. The recording sites are regions where eccrine glands are dense:

• Fingers: the palmar surface of the middle or distal phalanges of the index and middle fingers; the most common site.
• Palm: the thenar and hypothenar eminences below the thumb and the little finger.
• Sole of the foot: an option when the hands are needed for the task.
• Wrist, forearm and other areas: practical for wearable devices, but because gland density is lower, responses are usually weaker and less frequent.

Recording is usually made from the hand used less in the task, which reduces movement and pressure artefacts. Standard ECG gels can be high in salt and can themselves change skin conductance over long recordings, so electrolytes suited to EDA are preferred. After the electrodes are attached, it is good practice to wait about five minutes for the gel to soak into the skin, then ask the participant to take a few deep breaths and check that responses appear. Starting the recording with a 2–4 minute rest period shows both the baseline and whether the person responds at all (Braithwaite et al.).

EDA is a slow signal; the bandwidth given for PLUX’s EDA sensor is 0–3 Hz. Even so, choosing the sampling rate from this band alone is not right: measuring response latency precisely, running decomposition methods and keeping EDA in the same recording as fast signals such as EEG or ECG can call for higher sampling. The chosen rate and filters should be reported in the publication.

04Which measures are reported?

EDA data are summarised with a small number of standard measures. The typical values below are for orientation only; there are large differences between people, electrode sizes and experimental conditions (Braithwaite et al.).

The difference between amplitude and magnitude is often confused: amplitude is the average of only those trials with a response, whereas magnitude counts trials without a response as zero. Which one is used should be reported together with the response threshold and the latency window. Because SCR distributions are skewed, square-root or logarithmic transformation is common; for comparisons between people, standardising values to the person’s own response range is another approach in use.

05Artefacts and confounds

EDA is easy to record but easy to misinterpret. Anything that stimulates the sympathetic system produces a response, even if it is not the experiment’s stimulus. The main sources are these:

Temperature

Sweating decreases in a cold room and increases in a warm one. Keeping the laboratory at a steady temperature of about 22–24 °C and giving participants time to acclimatise is recommended.

Movement and pressure

Moving the hand, pressing on an electrode or pulling a cable produces sudden spikes. Because these spikes are much steeper than real responses, they can be found with slope criteria.

Breathing, coughing and talking

A deep breath, a sigh, a cough, a sneeze or prolonged talking produces an SCR in its own right. Marking these events during recording makes them easier to remove in analysis; recording respiration separately also helps.

Habituation

As the same stimulus is repeated, the response shrinks. The stimulus order should be counterbalanced in the design and the effect of the first trials assessed separately.

Non-responders

About one participant in ten may show no measurable EDA response; the proportion is higher in some groups. A rest recording and a simple test stimulus help to identify these people before analysis.

Individual and daily differences

Season, time of day and personal characteristics also affect the baseline, so it is safer to interpret SCL as a difference between conditions rather than as an absolute value.

06How are EDA data analysed?

The classic method measures each response from its onset to its peak. It is simple, but when responses come close together they overlap: when a second response begins during the decay of the first, its amplitude is underestimated. Methods that separate the signal into its components were developed to overcome this:

Ledalab and continuous decomposition analysis

Benedek and Kaernbach (2010) separate the signal into a tonic component and a continuous phasic driver that represents the nerve activity reaching the sweat glands, so overlapping responses can be assessed individually. The method is implemented in the MATLAB-based Ledalab software.

cvxEDA

Greco et al. (2016) estimate the tonic and phasic components and the noise together, as a convex optimisation problem based on a physiologically plausible model.

Filter-based decomposition

The slow component is separated with a low-pass filter and the fast one with a high-pass filter; it is quick and available in open-source tools (for example NeuroKit2). The analysis lab on this page uses this method.

Whichever method is chosen, analysis should start with a visual check: automatic algorithms can mistake an artefact for a response or delete a genuine tonic change as drift. The response window, threshold, transformation and decomposition method should be stated clearly in the publication.

07In which research is EDA used?

EDA can be used in any study that needs to follow the course of arousal over time:

• Stress and workload: mental tasks, stress protocols such as public speaking, driving and simulator studies.
• Emotion research: arousal in response to emotional images, sounds and films; together with self-report, facial expression or other measures for valence.
• Learning and fear conditioning: acquisition and extinction of the response to a conditioned stimulus.
• Decision making: responses before a choice in tasks involving risk and reward.
• User experience and neuromarketing: examining arousal in response to an interface, an advertisement or a product on the same timeline as eye tracking and EEG.

There are also studies of EDA in clinical and psychiatric contexts; these are research findings, and any assessment of an individual requires a physician’s assessment. EDA does not on its own measure “stress”, “lying” or a particular emotion; it measures change in arousal, and its interpretation depends on the experimental design.

08Choosing a system for EDA measurement

The choice depends on where the recording will take place and which signals it will be combined with. In the lab, an EDA sensor plugged into a data acquisition hub records on the same clock as ECG, respiration and EMG. Multi-sensor research kits bring the sensors together for studies such as stress and fatigue. If EEG is part of the study, an EEG amplifier with a GSR channel reduces cabling and the synchronisation burden. In courses, teaching kits are used to teach the basic concepts. Nevosa supplies the PLUX and g.tec systems listed on this page and provides installation, training and technical support. Details such as bandwidth, sampling rate and compatibility are on each product’s own page; use the quote form for a configuration that suits your study.