Physiological Psychology 1985. Vol. 13 (4). 291-295 How can skin condctance responses increase over trials while skin resistance responses decrease? MANFRED VELDEN University of Osnabrock. Osnabrock. West Germany and GERHARD VOSSEL University of Mainz. Mainz. West Germany t is shown that the opposite trends over trials for condctance and resistance response magnitdes that have been fond in recent experiments can be explained entirely by the mathematical relationship between condctance and resistance. Two conditions have to be satisfied for sch opposite trends to occr: (1) a sbstantial increase in condctance level over the corse of an experiment, and (2) a certain increase in condctance response magnitdes. The opposition in trends, being de solely to the mathematical relationship between condctance and resistance, shold not be eliminated by some "correction" procedre. nstead, condctance shold be taken as the appropriate variable. Also, no correction shold be made for the correlation between condctance level and condctance response magnitde, since it reflects psychological processes rather than a biological level dependency. n a stdy pblished recently (Bocsein, Baltissen, & Eler, 1984), as well as in experiments condcted in or own laboratory, parts of which will be presented below for illstrative prposes, a rather paradoxical effect was observed. With high stimls intensities, repeated stimlation reslted in an increase in skin condctance responses (SCRs) over trials while, at the same time, the corresponding skin-resistance responses (SRRs) decreased. Since condctance and resistance are related nonlinearly, the pattern of reslts for electrodermal activity will always differ somewhat, depending on whether it is resistance or condctance that is being taken as the appropriate manifest variable to represent some latent (psychological) variable. With reslts differing in a qalitative way, as in the experiments mentioned above (measrement of condctance sggests sensitization over trials, whereas resistance indicates habitation), it is well worth asking what the reason for this difference is. This reasoning will have a strong bearing on how to proceed with the data, for example whether or not to apply some correction procedre. n trying to explain the difference, Bocsein et al. (1984) referred to the correlations they fond in their data between level (initial vale) and response magnitde. With few exceptions, they fond positive level-response correlations for both skin condctance and skin resistance, which they interpreted as a level-dependency of the Reqests for reprints shold be sent to M. Velden. University of Osnabriick. Department of Psychology. P.O. Box 4469. 0-4500 Osnabriick. West Germany. responses. They explained the opposing trends over trials for resistance and condctance responses in the case of high stimls intensities by making the assmption that, for these stimls intensities, an increase in skin condctance level (and, accordingly, a decrease in skin resistance level) occrred over trials. n this case, the positive correlations wold mean a tendency for the condctance responses to increase over trials and a tendency for the resistance responses to decrease. The athors interpreted the level-response correlations in terms of a leveldependency, and conseqently proposed a correction procedre that wold eliminate the opposition fond in the trends of the SCRs and the SRRs. Bt how can there be positive level-response correlations for both condctance and resistance, with both condctance and resistance responses reflecting the same psychological response? Even thogh there is a nonlinear relationship between resistance and condctance, larger psychological responses shold yield both larger condctance and resistance responses and smaller psychological responses shold yield both smaller resistance and condctance responses. f there is a tendency for larger amplitde psychological responses (and ths greater condctance responses) to be obtained with higher condctance levels, for example, de to increased arosal, then we wold expect aposirive level-response correlation when measring condctance, as was fond by Bocsein et al. Assming also that greater resistance responses reflect greater psychological responses, there shold, however, be a negative level-response correlation when measring resistance. 291 Copyright 1986 Psychonomic Society, nc.
292 VELD EN AND VaSSEL Greater psychological responses with higher condctance level mean greater psychological responses with lower resistance level, simply becase resistance and condctance are being related reciprocally. The reciprocity between resistance and condctance sally does not mean a reciprocity between the resistance- and condctanceresponse amplitdes. Bt how can sch a reciprocity between amplitdes evolve and case a positive level-response correlation for condctance and resistance? The reason is a prely mathematical one, one that has nothing to do with biological level dependency of electrodermal responses. Methodological implications will be addressed below. Figre 1 depicts the well-known relationship between condctance and resistance. The.::lC segments on the abscissa represent three condctance-response amplitdes; the.::lr segments on the ordinate represent the corresponding resistance-response amplitdes, the arrows indicating the direction of change. t can be seen that the condctance amplitdes become increasingly larger from.::lc t to.::lc3 while the corresponding resistance amplitdes become increasingly smaller. For the sake of graphical simplicity, the responses were chosen to be adjacent. t can easily be imagined that the same effect (reciprocity of resistance- and condctance-response amplitdes) wold reslt if every next response started at a level above the pper level of the preceding response. The effect may also occr, to a certain limit, if the.::lcs (and the corresponding.::lrs) overlap to some extent. t cannot occr, of corse, if all responses start at the same level. n this case, an increase in condctance-response amplitdes wold always be accompanied by an increase in the corresponding resistance-response amplitdes. t shold be noted that the reciprocal relationship between Rand C mst not imply reciprocity of resistance and condctance response magnitdes, even if there are sbstantial differences in level. f condctance response amplitdes were to decrease with increasing condctance level, the corc AC2 r---t--+------"'" AC1L-_-t-_-t- +- Figre 1. Possible increase in condctance responses to three sccessive stimlations with a corresponding decrease in resistance responses. R responding resistance responses wold also decrease, althogh at a faster pace. t need hardly be pointed ot that the above reasoning implies that an increase in condctance level over trials can mean that a decrease, over trials, in resistance response amplitdes corresponds to an increase in condctance response amplitdes. Whether or not this effect occrs depends, of corse, on the amont of change in level over trials. A small or zero level change will reslt in similar, rather than opposing, trends for resistance- and condctance-response amplitdes, as is sally fond (and was also fond in the stdy mentioned above for low stimls intensities). n the stdy by Bocsein et al., no data on level were presented, so it is not possible to determine whether or not the opposing trends observed for resistance and condctance responses were actally de to a sbstantial shift in level over the whole corse ofthe experiment. Empirical researchers sally tend not to flly accept a dedctive proof nless there are empirical data that behave according to that proof. We, therefore, present some illstrative data from two experiments performed in the second athor's laboratory. Since the experiments were not designed for this illstrative prpose, only the relevant data will be presented. Skin condctance data covering a broad range of level and response vales were selected from a standard habitation experiment (cf. V ossel & Rossmann, 1985) and from a stress experiment in which sbjects had to perform a complex reaction time task in the presence of discrete noise events (cf. Brnner, 1986). METHOD Sbjects The sbjects in the habitation experiment were 36 male stdents aged 19-27 years. The sbjects in the stress experiment were 19 male stdents, aged 19-36 years. Apparats Skin condctance measrement. A Schwarzer polygraph (Varioskript V 822) monitored sbjects' skin condctance responses (SCRs), skin condctance levels (SCLs), and presentation of stimli. Beckman Ag-AgCl electrodes (srface area =.64 cm') were placed on the thenar and hypothenar srfaces of the palm of the sbject's nondominant hand sing Beckman stomaseal electrode collars and Beckman electrode paste as a condcting medim. Skin condctance was detected by a constant-voltage (.5 V) skin-condctance copler (Lykken & Venables, 1971). Habitation experiment. The stimls series of the habitation experiment consisted of2-sec, l000-hz pre sine tones with a randomized interstimls interval averaging 22.5 sec (range = 15-30 sec). Tone intensity was 70 db (re:.0002 dyne/cm') as meared by a sond-level meter (Briiel & Kjaer Type 2203) at the headrest of the sbject's chair. Twenty tone stimli were presented. They were prerecorded on tape and delivered by a tape recorder (Revox A 77) throgh a speaker placed against the wall in front of the sbject at a distance of approximately 150 cm. Stress experiment. n the stress experiment, the sbject performed a compterized reaction time task. A stimls board (ZAK DTG) containing 20 colored lights (for each of ble, yellow, red,
NCREASE N SCRs WHLE SRRs DECREASE 293 green, and white) was employed. The sbject responded to each light by pressing a key with the corresponding color sing his dominant hand. Sixty lights were presented per minte. Dring performance of the reaction time task, 20 noise brsts (composed of a mixtre of white noise and different machine noises) were presented. Noise intensity was 103 db (re:.0002 dyne/cm2), mean noise dration was 10 sec (range = 5-15 sec), and the interstimls interval varied between 30 and 90 sec (mean = 60 sec). Modes of noise delivery and intensity measrement were the same as in the habitation experiment. 12,C 'f "'- Procedre The sbjects were reqested to wash their hands prior to connection of the electrodes. Both experiments took place in a sondproof chamber. Room temperatre was between 19 and 22 C. All recording and programming eqipment was located otside the chamber. n the habitation experiment, a rest period of 5 min was followed by 20 presentations of the tone stimli withot frther warning. The sbject was seated in a semireclining chair and was reqested to sit qietly dring the whole period. n the stress experiment, a rest period of2 min was followed by the instrctions for the reaction time task. The sbject was instrcted to respond to each light as qickly and accrately as possible. He was frther informed that a lod noise wold occr from time to time. The main task started after the instrction and a -min training phase. t lasted for 24 min and 20 sec, dring which the 20 noise brsts were delivered. o.c E 2-11 V'l 10 CR1 CR2 - - LR RR3 82 84 86 88 SR{k.Q) RR2 "'-, " 90 92 Figre 3. Mean condctance levels and responses of for sccessive trial blocks with corresponding resistance levels and responses (habitation experiment). sary to measre both resistance and condctance independently. Bocsein et al. (1984), who did so, actally fond no difference in their reslts when they measred resistance and transformed to condctance and when they measred condctance and transformed to resistance. - RESULTS AND DSCUSSON Scoring SCRs greater than 0.02 Jtmhos that occrred between and 5 sec following tone or noise onset were considered as stimls-evoked responses. They were measred as change in condctance. SCLs were sampled at 20-sec intervals preceding each tone or noise stimlation. To compare the trends for SCRs and SRRs over trials, the SCRs were transformed into SRRs. t shold be noted that the precision of modern skin condctance measrement makes it nnecesi R RR:3 i 36 40 44 SR(kQ) i i" 48 52 56 Figre 2. Mean condctance levels and responses of for sccessive trial blocks with corresponding resistance levels and responses (stress experiment). Figre 2 shows the mean condctance-response magnitdes and their corresponding resistance-response magnitdes from the stress experiment, with the total nmber of trials condensed to for blocks. The responses start at the mean initial level of each trial block. There is a nmerically small, bt consistent, increase in the size of the condctance responses over trial blocks, accompanied by an obvios decrease in the size of the resistance responses. The rather small decrease in resistance responses from the third to the forth trial block is de to the rather strong increase in the sizes of the condctance responses. Had the condctance responses increased still more from the third to the forth trial block, there might have been an increase in the corresponding resistance responses. Figre 3 shows the data of the habitation experiment with the same kind of data redction as in Figre 2. n this case, a decrease can be seen in the size of the condctance responses over trial blocks, as can, correspondingly, an even somewhat more prononced decrease in the size of the resistance responses, de to the nonlinearity of the relation between C and R. Note that the shift
294 VELDEN AND VaSSEL in level nder this experimental condition is nmerically very small relative to the one for the stress experiment. The effect of the small size of the segment of the crve representing the relationship between Rand C that is depicted in Figre 3 is that it appears to be nearly linear. This means that, had there been a prononced increase in the size of the condctance responses within the range of levels shown in Figre 3, there cold well have been an increase in the size of the resistance responses. This shows that, even nder the condition of increasing levels of condctance and increasing condctance response amplitdes, the reslt need not be opposing trends for condctance and resistance responses. t all depends on both the size of the level shift and the amont of change in response amplitdes over the corse of the experiment. For an overview, the reslts of the two experiments are taken together and presented in a more common form in Figre 4, with the condctance data being depicted on the left and the resistance data on the right. The qalitative difference between the trends for the resistance and condctance responses in the stress experiment as opposed to the concordance fond in the habitation experiment can be seen very clearly. These methodological considerations have strong implications for the qestion of "correction" procedres. As mentioned above, Bocsein et al. proposed a correction procedre that eliminated the difference between the patterns of reslts fond for resistance and condctance. First it shold be kept in mind that, in a psychophysiological laboratory, skin resistance and condctance are measred in sch a way as to reflect the same biological (electrodennal activity) and psychological (e.g., attention, arosal, etc.) entity, even thogh in different forms, "0.c E 3 (j) 1.4 1.0.6.2 '... -"0- - --0 2 3 4,..- stress expo --- habitation expo q trial blocks 2,, -- 3 4 4 3.:s: Figre 4. Condctance and resistance response amplitdes over trial blocks for stress and habitation experiments. 2 (j) reflecting their nonlinear relationship to each other. They are not correlated, bt they are related mathematically. This means that any difference between resistance and condctance data, like the opposing trends over trials fond, for high stimls intensities, both by Bocsein et al. and in one of or experiments, is exclsively de to the mathematical relationship between resistance and condctance. Under certain circmstances (sfficient shift in level and some increase in condctance response amplitdes over trials), opposing trends mst occr for resistance and condctance response magnitdes. t clearly makes no sense to apply a correction procedre in order to eliminate differences in trends that mst occr for mathematical reasons. t mst be asked, instead, which of the two physical variables is the appropriate one-that is, the one that is possibly related linearly to the latent (psychological) variable. Actally, condctance has generally been accepted as the appropriate variable in the psychophysiological laboratory, and the reasons that have been pt forward in the Jiteratre for this will not be repeated here. t shold be mentioned, however, that the data presented above also favor condctance as the appropriate variable. With the high stimls intensities in the stress experiment, sbjects probably became increasingly arosed over the corse of the experiment (as reflected in both the increase in condctance level and the decrease in resistance level), which means that they became increasingly responsive to the stimli, as reflected in the condctance-response, bt not the resistanceresponse, amplitdes. Note that this implies that there is also no biological level dependency for which to make some kind of correction. f the sbjects became increasingly arosed over trials and therefore reacted with increasing responses, which is most plasible, then there mst have been some kind of psychological level dependency which shold not be eliminated by some correction procedre. As arged by Lykken and Venables (1971), care shold be taken not to se correction procedres that might eliminate valable psychological infonnation. A biological level dependency in terms of a ceiling effect (which shold be corrected for) wold reqire that the condctance responses decrease with increasing condctance level. The following is a smmary of the essential points: Opposing trends, over trials, in the magnitdes of condctance and resistance responses may occr nder certain circmstances (sbstantial increase of skin condctance level and some increase in skin condctance response magnitde over trials) for prely mathematical reasons. No "correction" shold therefore be made to eliminate the difference between the trends; instead, condctance shold simply be taken as the appropriate variable. Finally, the correlation between condctance level and condctance response magnitde for high stimls intensities is not de to a biological level dependency, bt reflects psychological processes; this also arges against the se of data "correction" procedres.
NCREASE N SCRs WHLE SRRs DECREASE 295 REFERENCES BOUCSEN, W., BALTSSEN, R., & EULER, M. (1984). Dependence of skin condctance reactions and skin resistance reactions pon previos level. Psychophysiology, 21, 212-218. BRUNNER, R. (1986). Nacheffekte von Stress: Eine UberprUjng der "information load" Hypothese von Cohen. Unpblished master's thesis, University of Mainz. LVKKEN, D. T., & VENABLES, P. H. (1971). Direct measrement of skin condctance: A proposal for standardization. Psychophysiology, 8, 656-672. VOSSEL, G., & ROSSMANN, R. (1985). Stabilitiit nd Konsistenz der EDA Habitationsgeschwindigkeit. Zeitschrift for Differentielle nd Diagnostische Psychologie, 6, 10-18. (Manscript received Agst 15, 1985; accepted for pblication December 5, 1985.)