Skin conductance measurements have been used in psychophysiology for more than a century. Measurements of galvanic skin response (GSR) (also referred to by the more general term electrodermal activity (EDA)) have been reported as early as the last part of the nineteenth century (see [1] for an overview). Through many decades, the method of using direct current (DC) measurements of skin conductance with a constant applied voltage has dominated the EDA literature [2].
However, there are potential complications with the DC method, such as polarization of the electrodes and the corresponding build-up of a counter e.m.f. Furthermore, the use of an alternating current (AC) method instead would enable a simultaneous measurement of any capacitive EDA and also endosomatic EDA (potential responses) [3]. Hence, AC methods have been investigated for use in EDA studies (see [4] for an overview) [4, 5]. While it seems that AC methods solve problems like polarization of the electrodes and electrolysis of the studied tissue, some pitfalls should be avoided to utilize all the possibilities that this method offers. In this paper, we address some obvious pitfalls to be prevented.
It is generally known that the measurement depth is dependent on the measurement frequency when macro-electrodes (i.e. much larger than the SC thickness) are used for skin admittance measurements. The higher the frequency, the more similar are the electrical properties of the SC and the viable skin [6], and hence the more are the measurements dominated by deeper, viable skin layers. Martinsen
Figure 1 shows a simple equivalent circuit for human skin [8]. The upper part of the model represents the SC and the series resistance R∞ represents the deeper, viable skin layers. These layers should normally be represented by a more complex electrical model, but at low frequencies, where the SC dominates, a simple resistance is a good approximation. The polarization admittance, YPOL, represents only the frequency dependent capacitance of the SC and its dielectric loss, while GDC represents the conductive properties of both the SC and the sweat ducts [8, 9].
To measure the dynamic part of the skin conductance, which is due to sweat activity, it is therefore important to focus the measurements on the GDC. This is done by keeping the measuring frequency low and by using some kind of synchronous detection to measure only the conductance.
There are two reasons for keeping the measuring frequency low: To avoid the influence of the deeper, viable skin layers, as explained above, and to avoid the influence of the frequency dependent part of the AC conductance. The conductance measured at low frequencies is actually the sum of the DC conductance (free ions in the tissue) and the frequency dependent conductance (dielectric loss in the capacitive properties of the SC).
Figure 2 shows the results of measurements with two Kendall Kittycat® electrodes in the left palm of the hand. A total of 20 measurements were done in different sessions on a healthy, male volunteer, using a Solartron1260/1294 system in a two electrode setup (150 mV rms).
During the sessions, the sweat activity was measured in parallel on the right palmar skin site, which is shown in figures 3 and 4. A custom-made two-electrode based instrument that uses superposition of an AC voltage (500 mV peak amplitude, 20 Hz) and a DC voltage (500 mV) to measure AC and DC conductance at the same time was used. During the first measurement session, it was tried to keep the EDA at a high level (figure 3). This was done by combing through the test persons hair. For the seconded session, the test person tried to relax as much as possible. According to figure 2 the conductance measured in the excited state is considerably higher compared with the relaxed state. The difference is mainly due to the sweat activity and therefore the DC conductance.
In the case of excited state, the frequency dependent conductance contributes about 7% at 20 Hz and about 31% at 100 Hz, when assuming the measurement at 1 Hz is purely due to DC conductance. In the relaxed state, the DC conductance is low and the frequency dependent conductance plays a more significant role even at lower frequencies. The increase from 1 Hz to 20 Hz is then about 37% and from 1 Hz to 100 Hz the increase is about 112%. Please note that these results are found for this specific test subject under the described conditions and may be different for other persons and conditions.
Since the frequency dependent part of the conductance has nothing to do with the sweat activity and electrodermal activity (EDR) [10], it is obvious that the measuring frequency for EDA measurements should be kept well below 100 Hz and preferably as low as possible. The chosen frequency will be a trade-off between the ability of the measuring system to detect quick changes (requires high frequency) and sensitivity for sweat duct activity (requires low frequency), and a frequency in the range 10 – 20 Hz may be a good choice in that respect. It should be noted, however, that Nordbotten
The three-electrode system enables monopolar measurements of the impedance of the SC beneath one single electrode, without any demands for the size of the other two electrodes [12]. Hence, this electrode system is often used for skin conductance measurements.
The left part of figure 5 shows the three-electrode system with the three electrodes placed on the surface of the skin. For a frequency range where the SC typically dominates the measurements (i.e. typically below 1 kHz), only the impedance in the SC below the measuring (M) electrode is measured. Because of the feedback through the skin, the operational amplifier will generate the necessary voltage to drive the inverting input to the same voltage as the non-inverting input, i.e.
The right part of figure 3 shows an equivalent diagram for the measuring set-up, where the impedance of each of the three electrodes are denoted ZM, ZR a nd Z C, respectively. Each of these impedances includes the electrode polarization impedance and the impedance of the epidermal stratum corneum (SC) below the electrode. Compared to the impedance of the SC, the impedance of the living part of the skin is regarded as negligible, which is valid at least at frequencies below 1 kHz for regular macro-electrodes [7]. The skin DC potentials are indicated as batteries in series with these impedances.
Because of the feedback through the skin, the operational amplifier will regulate the output voltage
The voltage over ZM will then be
It follows from eq. 2 that the voltage over the measured impedance ZM is
If, for instance, the M-electrode is placed on a palmar skin site and the R-electrode on the dorsal side of the hand, a sudden increase in the (negative) skin potential inside the hand will lead to a negative response in the measured conductance. Moreover, a decrease in the potential, as we e.g. see in the diphasic signal in figure 6 (bottom), will lead to a positive response in the measured conductance. Hence, the change in skin potential can make the corresponding conductance response appear larger, smaller or even negative. This is one of two possible mechanisms giving rise to false negative conductance responses. The other mechanism is due to using wet gel electrodes and is described in the next section.
The choice of electrode type is crucial in EDA measurements. A study on electrode gels for skin conductance measurements was done by Tronstad
Figure 8 shows corresponding measurements performed on the abdomen of a test subject with the same two electrode types. There is apparently no direct response to the sound stimulus, but the physical activity generates thermal sweating. The two electrode types behave very differently. While the solid gel electrode (A) measures an initial decrease in skin conductance due to the test subject relaxing, the wet gel of electrode B penetrates into the stratum corneum, increasing the baseline level. Apparently, it also penetrates into the sweat ducts, because the conductance response during the physical activity is negative, indicating that the well-conducting gel is pushed out of the sweat ducts and is replaced by sweat of lower conductivity.
Although Tronstad
In this paper, we have argued that AC methods for measurement of skin conductance may have advantages over DC methods in EDA studies, but that there are pitfalls that should be avoided in order to use the AC methods correctly. The correct use of the AC method include a properly chosen measurement frequency, correct use of the electrode system (e.g. a three-electrode system) and the use of electrodes that do not produce significant changes in skin conductance over time.