PHYSIOLOGICAL INIC-CHANISNISP ANALYSIS AND BEHAVIORAL SIGNIFICANCE OF THE ELE, CTRODERMAL RESPONSE ----------- FINAL REPORT Section Page 10 12 Va-sclilar P-ffects UPON Skin Potential 82 13 References 89 P@IYSIOLOGICAL i%IC-CHAiNIS.\IS, Ai\tALYSIS Ai\'D BE'-IAVIORAL SIGNIFICANCI',, OF q-@IE ELECTRODERNi4NL RESROLNSE- 1. LNTRODUCTIOIN This project was directed toward exploration of physiological mechanisms underlying the electrodermal response 2in the hopes of establishing a rational basis for quantitative treatment of this measure as a behavioral indicant. A further objective was to gain a better understanding of the adaptive function of electrodermal activity. It was presumed that the role of such activity in our "psychological" life can be understood if such behavior is regarded as essentially a biological adaptation, modified to fit a social struc2ture. Thus, if electrodermal activity is associated with profuse palmar sweating, and this can be shonvn to be defensive in function, one has grounds for interpreting such activity as a sign of fear or anxiety. U, on the other hand, other forms of electrodermal activity facilitate manipulation or exploration, one would put an entirely different interpretation on its apfearance. For example, if, i2n a situation whic[-l is potentially threatening, one notes evidence of the manipulative type of electrodermal activation, it would seem appropriate to conclude that ihe subject is engaged in copl-ig behavior ratj-,ier than that he is beset with alarm. The first Fortion of this three-year program was devoted primarily to pliysiological investigations. -An assortment of evidence, covered in the interim reports, hel4ped to round out the partially elaborated physiological model of the electrodermal system. Among these experiments were microelectrode observations on sweat ducts and the areas ben,@een the ducts, confiriination of potential responses from the nail bed, study fiirthe-r examination of the effect of s-pecific 2 elcctrolytes on o'leerrodermil responses, and finally continu ing studies on the sweat L real)sorption mechanism and its reflex control. These v;@-rious findings, when integrated witli other experimental evidence, lead to the formulation of a model in which sweat reabsor2ption played a prominent role and Lft which such activity was reflected in the recovery limb of the skin conductance or resistance response. According to this model the sweat g),Ia.nd has a dual function; if it secretes profusely, the skin surface becom.--s well-hy&ated and resilient and is thus protected against abrasion. In a biological sense the animal is now able to scamper over the 2 rough ground away from danger without undue mechanical damage to his contact areas. For fine manipulation, 'however, as may be involved in e,.cploration and -assessment of cbjects in the immediate surrounds, tactile requirements are such that the optimum level of surface moisture is somewhere intermediate between dry and wet. It was supposed that reoulation of surface mofsture at light to moder2ate hydration is largely a function of the activity of the sweat reabsorption mechanism. Observation of conditions under which reflex sweat reabsorption occurs led to the conclusion that this mechanism goes into action in preparation for "manipulative" tasks. This activity is reflected in increased positive-go:ing'skin potential responses and in acceleration of recovery of conductance responses. 2The major effort of this investigation was then directed to@vard the elaboration of this recovery lin-Lb measure, in terms of its measurement, its relation to amplitude, t( conductance level, and to behavioral state, and to the comparison of its discriminating strength with -:hat of other electrodermal measures. 'niis report summarizes progress to date. It is broken into a series of separate topics related to these ob1jectives. in -qc ,inn-rnnriate to render as complete a AdeL 3 picture as possible widiiii this report. One full-length ptpcr, "'nie information con- tent of the recovery limb of dic clcctrociermal response" (in press) has been in part supported by this contract and is included as part of :his final report since it summarizeo-- 2 the approach and findings in an optimal nianncr. Ii@ addi@ion to the val:ious sections on the 'recover@ limb, there is one on the relation of vascular changes to skin potential shifts at the surface. Although some of this material was described in an interim report, it has nonv beei-i completed and com- posed as an integrated manuscript to be submitted for publication, and Is, 6therefore, included in toto in d-iis report. 4 2. SU,\I,\IARY Or@' R-LNDINGS Scvcral iiietliods of nicasuring the recovery rate of an clec:trodcrrnal respoiisp- have been assessed and compared. Tticse include several manual methods and one electronic measure, all matficinatically derived from a,treatm2ent of tllc expon(.,nti@l curve. All suffer from uncertainty witli regard to the asymptotic level approached by the exponential portion of the recovery limb. One method of coping with this difficulty is to use a logarithmic writeout in which case the slope of the recovery limb is an easily measured characteristic which is proportional to recovery rate and inde- pendent of the necessity 2for deciding where the asymptotic level is. Fven then, this holds true only ff --he asymptotic voltage is zero, a situation which may be artff ically produced by the use of capacitance coupling. For analysis of recovery rate of standard DC recordings by manual.techniques, the "D". measure is the method of choice. This is a simple measurement, along the extended baseline, of the distance intercepted by 2 the altitude of the response and its steepest recovery slope. This measure was used for much of the behavioral study reported here. An elecr-ronic method usina analog computation of maximum slopes of the ascending and descending linibs of a response was highly accurate for "clean" responses, but suffered gross inaccuracies whenever the ascending limb was composed of nvo or more slurre d components.2 Another method which should lend itself readily to automatic on-line computation is the amplitude-slope method, but this has not yet been tested electronically. -An examination of the feasibility of using capacitance-coupled recordings for calculation of recovery rate shows that such a procedure is acceptable provided one uses a (-oupling time constant of 6 seconds or loncer. This procedure 5reduces the Iii@l-ilir rn.--related. At a 5 coupling time constant of 10 seconds, the loss of sensitivity becomes negli-Lble for all but the very slowest recovery limbs. Recovery rate was found to be capable of distinguishing between many be- havioral states, even when response amplitude or resp6nse frequency2 could not. lhus, to name a fenv, it distinguished: orienting responses to a light flash from responses to the same flash when it took on signal properties, responses to arL alerting signal from those to a task-execution signal, the resting state from various task situations, mirror tracing from back-viard counting or cold pressor exposure, problem solving from perceptual or psych2o-motor behavior, a deception task from a reaction time task. In the course of these comparisons it was found that recovery rate be- came slower with habituation, even in a deception task, a--id that it was also slo@,,ed by the entry of a 'Lric,,rht stimulus into a task situation. It failed to distinguish a decep- tive response from a non-deceptive response in a given series of queries when d2iffer- ences were compared across the entire population, but individual- subjects did frequently show a difference. The design of this experiment was unfortunately aimed at group analysis and the @.iumber of deceptive responses for a given subject was insufficient to evaluate this individual effect statistically. An overview of the effect of the various task and stimulus situations upon electro-- 1 dermal recovery rate indicates that acceleration of recovery reflects mobilization 'or goal-oriented performance. That the de termining factor was not general activation per se was evidenced by the slow recovery accompanying a cold pressor exposure, shonvn by other electrodermal indicants to be as activating as were performance tests associated with rapid recovery. 6 The recovery measure for a given individual in a standard situation was slion-m to be relatively stable over a pei:iod of 5 consecutive week-$. There were large characteristic individual differences bctxveen individuals even though they changed in the same direction when changing behavioral conditions. Efforts to find a behavioral trait associated wit@-1 recovery rate were generally unsuc@essful although fast recovery in a standard task (reaction time) wag found to be associated with low anxiety (SAQ), and with a tendency to maintain electrodermal response withoiit habituation during a reaction time series. This was seen as further evidence supportiri,g the interpreta- tion o'L fast reco2very as reflecting mobilization for goal-dlrected behavior, since these same subjects 1-iabituated to a series of non-signal tones just as fast as did subjects with slow recovery rate. IE is still uncertain as to whether the difference in recovery rate bet@veen problem solving and simple perception indicat@@s a specific difference in the effects of coanitive and perceptual behaviors on recovery, or whether this simply2 means that the problem-so-lving task was associated with higher anxiety. An examination of the relationship of recovery rate to other parameters of the response sho@ved it to have a low negative correlation with amplitude, that is, responses of higher amplitlde in a given behavioral state tended to have slightly slower recovery, When measured benveen different behavioral states, ho@v2ev--r, there was,often a tendency for the reverse to be true. In vie@v of the evidence showing that mobill@ation for ,:,oal- directed behavior 4-s associated with faster recovery, this observation probably reflects the fact that mobilization for task performance frequently causes an increase in activation resulting in electrz)dermal responses of high amplitude. This same consideration probably 4 e,xplains the fact that recovery rate is related (positively) to skin conductance level in 7 During tllc course of a comparison of the discriminating strength of the recovery rate measure with otlier clectroe@ert'nal nicasures, a new frequency measure was devised. 'Il-is measure is diffcrcnt from other measures of !'GSR frequency" or ficouiit" in that it examin2es, for any given task or epoch, not the total number of re- sponses but rather the maxiniun-L frequency displayed in a "burst" of three consecutive responses. This measure, termed f max, demonstrated a surprising strength in distinguishing bet-@veen stimulus conditions,, althou,,,,,h not the same conditions as were distinguished by recovery race. Thus f max distinguis2hed the deceptive response from listening to instructions, but recovery rate was not able to do so. Contrari@vise, recovery rate d'.stiiio,,liislied perception from problem so'.,ving while f max did not. The higi-iest. f max and the fastest recovery rates were both found during the reaction time test, but the level of f ma:K reached by any subject during this task, unli@2-e recovei:y rate,. bore no relation to his trait anxiety. A companion study, one directed at the physiological basis of skin potential levels and changes, demonstrated that a one minute engorgement of cutaneous vessels produces a slow negative shift and upon release of the cuff a sudden positive shift. With arterial occlusion these potential shifts were opposite in direction and2a,-reater in magnitude. Although changes were generally not over 0.5 mv they raise the possibility that vasomotor responses may be accompanied by surface potential waves. Whether these shifts were mediated by an effect of the vascul,@r state on sweat gland potentials or whether they represent changes in vascular potentials remains to be determined. 8 'TT OF ELECTRODE-r,,,\4AL Pf-,'COVERY PATE-: 3. ON TIIE \IEASUREi%,IE,',, RATIONALE@ in several The measurement of clectrodermal recovery rate may be approached 'tliere is an intrinsic recovery ways, all of which have in common the assumption that rate characteristic which may be the same for waves of greatly varying amplitudf,.s. An example of such a condition is that for the exponeitial curve in which a characteristic time.constant or rate constant may be common to all mz!ni2bers of a family of curves of different amplitudes. Darrow (1937) concluded that the recovery limb of the skin resistance resoonse (SRR) is exponential in form, and additional evidence, to be pre- sented here, supports ttxis interpretation. Methods for evaluating the rate constant of the electrodermal recovery limb are suggested from examination of the differential equation describing the exponenti2al rela- tion. Tne process described by this relation is one in which a variable changes at a rate, dE/dt, which at any instant is a linear function of the magnitude of that variable. Thus, for the case-of voltage chanoe in a condensor discharge. which may signify, for example, that the voltage drops by 5c7 per second, in which case 1 /0 .05 is the constant characteristic of such a process. Rate Constant The Lntegral of the above expression provides the common exuression for exponential decay, namely 9 Nvhere E0 is die starting lc%,cl and E is instantancous voltage at tine, t; k is the rate constant (rc). If E is decaying to zero as its asymptote, 4@ 3) When t 1/k, 2 Fol that is, it equals 37% of its original value. This t2ime, which e is equal to the reciprocal of the rate constant, is called --he time constant (tc) and, like the rate con stant, is characteristic of the process and independent of amplitude. All curves having this same characteristic regardless of their starting point may be superimposed upon the.same large exponential curve. Thus, to the extent that the electrodermal recovery jimb2 is exponential, it may be matched to an exponential curve of the same rate constant. This is the basis of the use of an overlay method for determining the time constant by curvc.mitching as described in section 6. Half-Time From equation (2) it can be shonvr,@ that decay half-time, that is, the time taken for decay to become 500,70 completed, is equal to 0.7 tc and is a cons2tant for all waves having the same time constant, independent of th eir amplitude. This measure, the recovery half-time, represents a second means of expressing recovery r-,te and is also described in section 6. Equation (2) represents a means of determining the degree to which the -recovery lin-ib fits an exponential curve. Fxpressed in common logarithmic form, 1 4) 10 from @vllict, it fOIIO%vs :Iiat a @vritcout having lo.-al:itliiiiic vcrtical compi:cssio-,i should give for exponential curves a strai-lit line N'@IIOSC SIOI)C is a - k/2.3. Tbc upper trace in Figure I is a writeout of an exponontial curve obtained by c2apacitor discharge ind I a recording of a fe,%v skin conductance responses. 13ascline of the skin conductance trace has been adjusted so that responses are recoverin(, to approximately zero voltage. Below are the same waves recorded through a logarithmic cor-ipression circuit. Note that the portion of the recovery limb which is exponential in form starts about one sec2ond after response peak-. The slope of the linear portion of the logarithmic recovery limb is proportional to the rate constant provided the asymptotic voltage is zero. Since this is not so, such a method cannot be used dir-actly. One may, honvever, apply the second derivative to achieve this end. -The time derivative of equation (1) is C2/ which may be w3@itten,. 2 4.. Integrating, which indicates tl-tat the first derivative of an exponential curve is also exponential as is its second deri-,@ative. The slope of the log-compressed wri-,eout of the first derivative is then proportional to k and unlike the case for the primary (DC) write- out, the curve, as required, decays to%vard zero because of the cauacitativ1e couplino@ olf--c--rronic coavcrsion. Since loa 'CAP s c 1 Figure 1. Upper trace: direct writeout of a condensor discharge (CAP) and skin conductance trace (SC). Paper speed I mm/sec. Lower trace: same as upper but with logari--hmic compression1. Figure 2. Diagram of method for me@@surin(r the rate zonstant by tan A/H. O/ -LL 2, '3 one is tempted to produce a writeout of d log E'/dt to obtain an amplitude reading directly proportional to ti-ie rate constant. Unfortunately, d log E//dt often has such a low magnitude in the exponential portion of the recovery limb that its analog form cannot serve reliably as the expression for the2 rate constant. A similar problem .arises if one attempts to determine the rate constant from the loa-compressed first derivative curve (loc, d E/dt) by manual measurement of the tancent of the acute angle produced betnveen the linear portion of the recovery limb, and baseline. Second Derivative Yet another method may be derived from the second derivative form. From eq2uation 'Ehus, one may determine the recovery rate constant by calctilatin- at any point on the recovery limb the quotient of the second derivative the first derivative. In practice the method is not too feasible because ti;,-- magnitude of E& is so low in the exponential region of the recovery limb that it frequently is exceeded by the noise level of the t-race, which becom2es rather high for the analog second derivative. Amplitude-Slope Method Another solution is one which requires an amplitude measure as well as a slope.. From equaticn (1), it is seen that 12 ression. rrom it., the slope of E, or its first dcrivative. 'Iliis is a most use).Ul cxp one can obtain the value for k by taking the slope and amplitude of any point On the exponential portion of the decay (Figure 2). 'llic slope is tan A. Tn practice it must be chosen at a point at or beyond the inflection point on the recovery limb. Because a measurement of wave amplitude would seem to be more precise and because such a measurement could be used independently as an index of responsivity, a test was made of the relation of H to h (Figure 2), i.e., of peak amplitude to the amplit2ude at the inflection point, 'Die product-moment correlations for 20 responses on each of 6 subjects were: .97, .99, .99, .99, .99, and .96. 'Ibus a convenient substitute measure for &,c rate constant is 6) This method showed a correlation of 0. SI with measurements made by the template method on the same 66 responses. In mak2ing the slope measurement, a line is drawn parallel to the recovery lin-ib at its inflection point. The acute angle at the intersection of this line with the horizontal is measured and its tar,.-ent obtained from tables Anothe relation in conjunction with this method pertnits a relatively simple approach to auto- mated calculation of the rate constant. The peak slore of the ascendina- 2lin-Lb is found to be linearly related to the maximum amplitude of the primary writeout (Edelberg, 196, A validation check of this relationship in the present investigation confirmed this. Measurement of 44 responses having uncomplicated ascending limbs gave a correlation of 0.94 benvee-ri the nvo measures. Hence one may substitute for H the maximum first derivativ5e o-' the- ascendincr lin-Lb, and for tan A the maxi@-inum first deri,,,ative of the c;pnaratelv 13 4. ON THE \IC- ASURE.\IEiN'T OF ELECTRODE- r,,,\IAL RECOVI,:RY RATE BY PRE- FERRE- D ,\,IE, T@@"ODS: TESTI',\G ALND CO,@,IPARISON A) Lo,-aritlim4.c Compression In the discussion of the various of2 computing clectrodermal recovery rates it was stiown that the first derivative of the logarithmic writeout of the recovery lin-Lb is directly proportional to the rate constant. This is true when logarithmic com- pression is accomplished by electronic means, but only if the wave is rec@)vering to a zero voltage level. In such cases, the reaion of the recovery lin-Lb immediately 2 following the peak (by about 1 second) is linear, and it is this portion which should reflect the ratz! "constant. " Unfortunately the DC- recorded trace rarely recovers to zero voltage as its asymptote. Since the first derivative does have essentially a zero voltage asymptote, logarithmic compressiotl of such a writeout should offer a linear section of the recove.7y lim-b whose slooe i2s proportional to the rate constant. LJn- fortunately, &.e electronically -differentiated recovery slope is often of such low amplitude that log compression and manual measurement pose a problem in accuracy. It has been shown, however, (Section 5) that the rec,@very rates of electrodermal responses recorded with capacitance coupILno- are highly correlated with those computed 2 from a DC record provided the coupling has a time constant of 6 seconds or lono-er. Such condenscr-coupled records do approximately satisfy the requirement that the asymptotic voltage is zero. If a record of this kind is subjected to logarithmic com- pression, the recovery limbs should show a linear portion whose slope is proportional to the rate constant (Figure 3). This study examined su5ch records to determine the degree to whiz-h the rate measure, computed from t@,eir slopes were correlated @vith the ........... cd @.4 2 Qj 2 14 Ef) 2 Cd r 2 0 I pit 2 bo 0 2 ID 4 r,4 w 14 Mctliod Instrumentation Skin cotiductaiice was recorded on a Beckman Dyno&,raph with direct coupl- ing (channel 1). The pen (1) output was coupled to the 2 mcgohrii input of another DC channel throug2h a 3-microfarad condcnsor to obtain the 6-second time constant, The pen (2) output of this second channel was placed in series with a 220 K resistor and a -silicon diode (Texas Instruments G-129). The voltage developed across the diode is a logarithmic. representation of. the pen (2) output (Kahn, 1962). This voltage was fed iito a third channel of the Dynograph and, with zero input into channel2 (1), and the recording completely restored to baseline, the zero position of pen 2 was adjusted until the voltage across the diode was for-ward biassed by 3.25 volt.' Polarity was arranged so tliat an electrodermal response produced increasing forward bias on the diode. ,Measuriiia- Techniqu A straight line is drawn parallel to the linear portion of the recove-2ry slope in the region immeb-ately after the peak of the response (Figure 4). On can then measure the acute angle (A) which this line makes with the horizontal, and obtain its tangent from tables. This value. is directly proportional to recovery rate constant. An alternative method is to use the L-shaped scale shown in Figure 4. 'The vert,,,al limb is set so that it passes through the intersection1 of &-e slope with the upper edge of the paper channel. The horizontal distance from this point to the intersection of the slope with the bottom edge of the paper channel can be read dirdct ly off the metric scale on the foot of the L. Since t,@n A is proportional to the rate constant, 1/tan A iq ,i ronstant). "la Upper edge of Scale Bottom edge of Scale 1 2 3, Figure 4. Measurement of recovery rate by measurement of anale A or intercept d. Also, since tan A Ii/d where H is the scale heiglit and d i@i the liorizontal distance measured, J r Since H and b are constants, d is directly pro- +,4 vi A portional to the tirr-c constant. Results The recovery time constants of 66 r2es@---nses recorded on magnetic tape were read by the electronic method described in Section 4C. This method consists of 'tive and peak negative first derivatives of the electrodermal meas'uring the peak pos-, response and dividincr one by the other. The same populat-'-on of responses %vas sub- jected to the template measurement (Section 6), t2o the amplitude-slope method (H/tan A), and also to tffe locr-compressed AC recording as described here. Produrt- moment correlations were determined between the values obtained by the logarithmic method and those from the three other methods. Values were as follows: Log/ E' v,3. electconic computer 0.59 lx2g t E vs. amplitude-slope 0.87 Logi E vs. template 0.86 The range of d in the log method was 3 to 47 mm which corresponds to an 8:1 range in time constants. The correlation of the log measure ,vith response amplitude was like many of the other measures low and positive (r = 0.37, R -<.01). 2 Because of the simplicity of the manual measurement involv@-,d in the measure- 0 because of its independence from laiowled,,a ment of loaf E, and especially ,e of baseline level, this seems to be a measure of ctioice. It has one &aw-back, namely the neces- 6 nt-inn in n@)r,-LiTiing its AMN, 16 B) N,"icasuremcnt of the Altitudc-SIOI)c Intci:cct)t Aloi,.rr f@aseline leads to yet another The notation H/tan A, that is, the amplitude-slope measure, approac2h to time constant measurement. In Figure 5 it is seen that tan A is approxi- mately equal to H/D where D is the distance along the dxtended baseline included be- tnveen the altitude aine @1) and the extension of the recover-I limb slope. As a result, H/tan A reduces approximately to D. The measurement of D in practice is not unlike the task of measuring t/2, the time for half 2recovery, but it has the advanta-e that D is about 75c7 larger than.t/2 and, therefore, incurs a smaller relative error in measure.- /0 ment. Furthermore, it allows measurement of a considerable number of responses which fall to recover by @O@'7 prior to the onset of the subsequent response. /0 To test -.he u2sefulness of this method the values of D were measured for the same 66 responses used to compare the other methods. Values of D ranged from 1.2 to 8 secon&-. Because the paper speed was only 1 mm/second, measurement was not very precise, but the correlation with the measurements obtained by H/tan A was neve-,-@ll,@--@3less 0. 90, and with those obtained with the curve-matching 2method, 0. 82. Correlation with the electronic method was 0.63 and w,'-th log/ E 0.86. At a somewhat higher paper speed, e.g., 2 to 2,5 mm/second, this is a very satisfactory r@iethod, especiauy since the value of 1), when expressed in seconds, can be readily converted to the time constant by the use of a constant factor. The use of a standard notation i6s highly desirable in making comparisons between results at different laboratories, and the use of the time constant or rate const-,int seems a reasonable choice for such standardization. To derive the conversion factor for 0, one must keep L-i mind the fact that in precise terms, tc = h/Tan A wl-Lich is equal to d in Figure 5. From this 16a e Fi,o-ure S. Veasurement of recovery rate by altiwde-slope intercept, D. 17 o r' - -------- Measurements on 40 responses gave a mean value of 0.74 for h/H + c, whence 7) O. To convert H/tan A to tc, one must consider the re"-ation of H/tan A to h/Tan A. Since a- ti A H 2A the conversion factor h/H must be determined. Calculations of h/H from 20 responses on each of 6 subje.:ts gave mean values of: 0.84 0.86 0.76 0.84 0.89 0.85 0.84 From this H 8) -LL O, -t -Dill A 'Ihe acrreement of these two relations were tested on 65 respons es, with results as follows: ,Notation Factor 4 Comt)uted te D 2.70 0.74 2.00 seconds H/tan A 2.34 0.84 1. 97 seconds 0 This agreciiient is surprisingly close and gives confi(.Icnce in tile reliability if not the -validity of these measures. 'nicir agreement N,.'itil tc values obtained by curve- matching is not nearly as impressive, the mean of that measure being 2.529 seconds. The possible cause of this discrepancy is described bcl'ow. The Positio a of the Asymptotic Level and Its Effect upon Recover-v Rate %Icasurement All methods described, whether curve-matchin-, t/2, D, or H/tan A, depend for their validity upon the accurate choice of the asymptotic level to which the exponea- 2 tial portion of the recovery limb is decayincr. The only exceptions to this are cases in which E '/E// or d aoc, F,)/dt are used to determine tc. In the other four measures, the conducta nce or resis tance level at point of response onset has been used as the asymptote, but it is clear from inspection that in numerous cases such an assumption is erroneous. Uncertainty as to the lev2el of the asymptote,' and in fact a systematic error in estimating it may explain the fact that recovery rate is found to have a correlati-on with response amplitude. This correlation though low is consistent for different methods of measurement and is significant. Examples of the correlation for measures on the same population of 65 responses are: r2 p Amplitude vs. Electr(,,-..;Ic tc .37 <.Ol Amplitude vs. Template tc .35 <.Ol Amplitude vs. log E .37 <.Ol True asymptotic level may be determined in the following way. In 4 Figure 6, tan(-,ents to the recovery lin-Lb have been dra@,in at points I and 2. It has been shon@-n Pirlit-r that the time constant of an expone,,itial curve = h/tan A or d in Fi,-,lrc 5. If <> 19 from either point should be identical. 7lic correct asymptotic level will be that at which di = d 2' When this method was applied to a nun-Lber of skin conductance responses, the asymptotic level was found to vary, sometimes falling 41.most on the baseline, but - more often considerably above it, oc2casionally near the pea'.%- of Ulle wave. As a consequence, the values for tc calculated from curve-matching, t/.2, D or H/tan A, are usually too laege, but since all are altered in the same direction by this effect, the inter-correlations are not seriously disturbed. Moreo,,,er the error introduced after conversion of eidier D or H/tan A to tc is the same, hence the agreement of 2 the mean time constants obtained from these t@vo measures. Such is not the case for the curve-matching technique. In this case the error is considerably greater and is a likely e,-,.planation for the discrepancy shown in the previous section. Since the other methods are less subject to this error, it argues in favor of the abandonment of curve matching in favor of D, H/tan A, or t/2. Since me@Lsurement of D a2llows measurement of more responses than does t/2 and entails no more work than for t/2 and less work than for H/tan A, it is considered the method of choice. More- over, it is faster than curve matching, requires less training, and the one judl-ement 0 to be made, namely the placin- o f a straight edoe parallel to the recovery linib at its steepest point is a simpler one than is curve fitting. To reiterate, the most advantageous mediod for measurement of tc is to measure D, using the conversion: tc = 0.74 D 20 or, for rate constant: 1.35 rc It will be recalled that LNvo alternative metliods for measurement of recovery rate are inherently independent of any knowledge of baseline or asymptote. The first of these is E/Ell , i.e., the ratio of the first to second derivative at any2 point on the exponential portion of the recovery limb. The difficulties with this approach have already been discussed. It does -,iot appear to be feasible at this time. The seccnd alternative, d aog E)/dt, that is, the slope of the linear portion of the logarithmically-compressed recovery lin-Lb does appear useful, but only with a condensor-coupled recording and,,,iith loa-com2pression circuitry. C) Electronic Computation As discussed in Section 3 the notation H/tan A forms the'Dasis of an approach to automatic on-line calculation by an analog computer. The analoo- approach takes advantage of the high linear relation of the peak first derivative of the ascendino, limb of the DC elect:rodermal recording and amplitude of the wave (Pearson's 2r - close to 1). Also convenient is the fact that tan. A ;.-. the maximum first derivative of the recovery limb. Hence: H - = -L@)v) A where E@ is the maximum first derivative of,the ascendin- limb, E- is the same for the 8 recovery limb, and g is a constant. There are several complications in the use of this expression for computino- 0 21 so that a storage requirement e,%ists. The use of Ii/tan A (Figiir,.,,, 5@ @-;ould be more pre- cise and because h and tan A are taken simultaneously, would be easier to program, but it would require measurement of li in terms of its distance above the baseline or 2 preferably above the asymptotic level. This becomes a problem in DC recording, because of expected baseline shifts, and such an approach does not lend itself readily to analog computation without cumbersome programming. Secondly, because responses in close sequence interact, contingencies must appear in the program for selection of waves meeting standard criteria. These cr2iteria are: (a) 'Ihe time of onset of any measurable response must be at least 7 seconds after the onset of the previous response. This is neces- sitated by the fact that -recovery rate of a small wave superimposed upon the recovery of a preceding larger one is spuriously rapid. The effect is shown in Figure 7a. (b) R2esponses which do not last at least 4 seconds prior to the onset of a successive wave must not be accepted for measurement. This requirement prevents measurement of a wave whose peak negative- going first derivative is cut short by the orset of another wave (Figure 7b). Method A corrlbination of analog computer 2and digital logic circuitry Is used to meet the conversion, storage, and contingency gating requirements of the computation. @i o (/I (D (D 0(D (A 0 4 0 o p OQ (D 0 (D (D 2 @j 0 0 < 10 ' (D (D LO E3 p 0 Fi 2 0 o r- Cl) (t) En (D cn (n Id cn. 0 En (D 5 0 It ct 0 0 0 07 ('D H rt 22 .@n4l2E- olqi@zial Cor@ crl@@ n@_ SL2i -,C, and@Cop,@p,,,-,Ii@on The skin conductance rcsponsc is recordcd Nvith a constant voltage bridge whose output, p-zoportional to skin conductance, is fed in*lo a Beckman Dynograpti 2 amplifier. 'nie voltage at the I)cn output is fed into two cther channels, each using a 0. I-second couplin- time constant, thereby achieving e--.rivation of the trace. The output of one of these is used to fire a Schmitt trigger to start analysis at the onset of a response. The output of the other is sorted into positive-going and negative-going components by fhe use of diode c2lippers which feed their respective signals to storage capacitors. This storage is arranged as a peak memory circuit so that after a response is over, one of the nvo capacitors is charged up to the peak voltage of the positive first derivative, the otlier the negative. Upon a command signal from the associated logic circuitry, both capacitors are read simultaneously by two separate voltage-to-freq,ue2ncy converters. These feed into a preset counter programmed to divide one frequency by the other, thus accomplishino- the computation of E@/E- or its reciprocal. The quotient is fed into a dio,,ital printer and at the same is pr'--nted out on another time the value of the positive-goino- derivative (E printer 2 to furnish amplitude data. Details of the arrangement are shown in FigiLres Sa and Sb. Control Loo,,ic System This system must recogni7.e the onset of a response, must screea out responses which fail to meet the t@,vo time criteria, must time and command readout and print, and must reset the storage can,-icitors. It must also decide, on the bas--i's of a minimum 4 amplitude crit erion, which responses are large enough -o measure. @-,iti@,out inc,-lrring unacceptable error because of low si.-iial-to-noise ratio. These demaads are,met by 22a -7 5 PEN OUTPUT 7,5 k REED CONTROL RELAY LOGIC REED 0- 2 RELAY STORAGE 2 0- 0 0 0 2 STORAGE RESET PRINT COvIMAiND PF' ESET 4 @COUINTER PRINTER A/B Figure 8a. Diagram of electronic programmino@ for automatic analysis of recovery rate. 22b NEG POS OUT OUT PEN 2-Z K RQFUT colm STORAG2E RESET Figure Sb. Storare circwt. 23 A free-running pulse generator (,c\) is set to fire at a rate of one per second., Its output is fed into a binary counter (Figure 9). The onset of a response serves to zero the counter sc tl-iat tin-ie gatin(, of subsequent operations may be standardized. This is accomplished by using the first dcri2vative signal from peii 2 of the dyno,,rraph. This signal is fed to a Schmitt trigger which fires @viienever a positive-going wave occurs in the first derivative trace. Firing of the Schm'itt trigger not only resets the binary counter to sta.ct counting at the next second, but also fires a one-shot having a 4. I-second output pulse. The-ascendinc, lin-Lb of this pulse fires a2 second one- shot whose 0.5-second output is used to trip a reed relay to discharge the storage capacitors (reset). The primary purpose of the 4. 1-second one-shot, however, is to act as a gate to prevent processing of the response.if a second wave starts with- in 4. 1 se6onds after onset. This is accomplished by feeding these one-shot pulses into a quadruple and-gate: the othe2r three inputs of which are fed from the 0-0- I terminals of the b4-nary counter. When 4 seconds of counting are up, the and-gate can fire only if the one-shot has not returned to its off-state. If this condition is met, the and-gate output acts as a command to the pre-set counter to take the quotient of the two storage outputs. The preset counter, after finishin- the division, sionals the printer to prir-t. The quadruple and-r-te also fires a one-shot whi'ch advances a decade counter used to number the responses on the printout. U a second or third response occurs during the initial 4-second period, the 4. I-second one-shot is reset and its off condition output serves as an inhibit signal to the and-gate. The requirement for at least 7 seconds benveen processed waves is8 met by the use of a triple and-gate and two flip-flops interposed bet@Nveen the Schmitt, tri@Ser and the 4. I-second one-shot. The triple a,,id-gate is led by a I- I- 1 output frori Llie AMIL F r-o- -m Pen One Reed Shot Relay 0. 5 Sec 2 Sclunitt 2 Trigger In One. Shct 4. 1 Sec Reset IrNbit Reset 2 Pulse Generator Preset Counter Decade Counter 1 0 1 4 Printer BCD Decoder Fio.ure 9. Control lo,,7ic c;,.rcuitr-y. 24 biniry counter so that it fires at 7 seconds after the Schmitt tri,,,,-Icr lias signalled start of the response. Its output resets the two flip-flops and allows a subsequent firing of the Scli-nitt trigger to fire the 4. I-second one-sliot. Until this triple and- gate fires again seven seco2nds later, any pulse coi-nin(- through after the 4 seconds of processing cannot fire the one-shot and therefore canhot initiate -tnodier computing sequence The -LNvo flip-flops are connected in such a way that the 4. 1-second one- shot cannot be fired again until the triple and-gate fires again. Thus no computation can be started u--iless t2he binary counter has been allowed to count for seven seconds after the start of the last response. Any response occurring during these seven seconds starts the count over again.. Results This system proved to be very accurate, the prir-t-out data ag-reeing very closely with values obtained by hand measurement of the positive and negative peaks o2f the first deriiative trace. Values of Eland of the time constants obtained by the t%vo methods for each of twenty solitary responses of a Nvriteout are plotted against each other in Figure 10. From this standpoint the system is very reliable. The major source of error is caused by responses in which the ascending limb is in fact a slurring of tnvo responses into a single one without an intermediate p7eak. In such a case the Schmitt trigger fails to recognize the second response because the derivative has not recrossed baseline (Figure 11). The maximum slope is that of one of the two slurred components rather than beina additive and is not consistent with total H. The value computed electronically is, therefore, considerably less than that computed II-- T- Cd 4-) 2 (n 4-4 i 2 0 Cd LLI 2 Cd Cd 2 Cd Cd OTUO.110;Dl a 2 'O r4 6 @T4 AMNL 24b Figure 11. BreakdoNvn of linear relation bet@veen primary amplitude and first derivalk-ive amplitude in duplex response F. Simple responses D and E show linear relation. 25 The above clcficicncy may explaiii the relatively 10%@,rcorrelation of ti.-le constants derived electronically from those obtaiiicd by the various manual methods. The (-6.rrelations with od-ior measures for the same population of 65 responses analyzed elsewhere in this paper w2ere as follonvs: Electronic vs. Log' F, 0.59 Electronic vs. H/tan A 0.68 Electronic vs. Template 0.42 Another problem with the electronic method lies in the selection of the minimum. threshold for trigc,,,erino, the Schmitt trigger. If this is chosen too low, many clean high ampl2itude waves are lost because of the occurrence of a miniscule wave in the previous 7 seconds. If set too high many perfectly useable waves are lost. The compromise between these t@vo conditions is difficult. With the particu[ar threshold used in this test (16c7 of half scale), the system accepted 63 of 86 responses which met amplitude /0 criteria, the remainder being rejected because of temporal conting2encies. The total aumber of responses, however, was far greater than the 65 having adequate amplitude. For a threshold 8"7 of half scale (i.e., 407 of full channel width), the total number of .,O /0 responses reachir..g amplitude criterion would have been 188. While this automatic system meets minimum requirements for an operationally 9 satisfactory system, its susceptibility to spurious readings of relative wave amplitude when multiple responses fuse in the ascending limb is a rither serious disadvantage. It appears possible to construct an on-line analog system lsi.'Io, one of the other approaches and in this regard the measurem--,-it of the slop-- of ulic log-compressed 26 D) Sun-ini@in, oi' Coni-)arison,5 Bct,,@,cen ',\Tcisures For convenience a summary of the various intercorrelatiol,,s examined is shown in the matrix in Tablz- 1. Table 1. Correlations bet-,veen time constants measured in different ways 2on the same 63 responses. Template Electronic Log /E H/tan A D t/2 .94 Template .42 .86 .81 .82 Electronic .59 .68 .63 Log.*E 2 .87 .86 H/tan A .90 The overall t igh correlation bet@veen the various manual measures gives conf idencc@ that a fundamental form characteristic of th e recovery limb :-s in fact under examination. <> 28 various couplin-s into a Beck-iiia,-i Dynograpli equipped with rectilinear ink writers. Time constants of 1, 2, 3, 4, 6, S, 10 seconds and DC were used. Time constants were measured by the amplitude-slope method described elsewher2e. Forty responses were selected wizh tlic,@ qualification that they must start at least seven seconds after the start of the previous wave, must have an amplitude at least 5 mm (full scale was .40 mm) and must not be interrupted by a successive wave until at least 50@o recovery was completed. In each case the rate constant was calculated from tan A/H. Values for each coupling condition w2ere matched with those from the DC record to compute a Pearson s pro@Llct-moment correlation. Results The correlations found at each value of coupling time constant are shown in Table 2. Table 2. Correlation betnveen recovery rates (tan A/H) computed from IDC records and t.-iose from- ca2pacitance-coupled records taken with various coupling time constants. R-C Slope Of Time Constant Pearson's r Line Of (Coupling) Regression ISeconds .33 2 .32 2 .51 .44 3 .72 .27 4 .80 .59 6 .86 7 .43 8 .90 .68 10 .86 .86 Note that Iiigii correlations are found for 6, 8, and 10 second couplin- constants, Lliat is wlien the coul)ling time constant approaches the time constant of the ,;Io%vest waves in the population (see Sections 6 and 8). The correlation reaches a relative2ly conslant level at coupling coil'St'ants of 6 seconds or longer. This constant lcv-! is less than 1, and is in part indicative of the departure of the relation from linearity and in part of tile repeat measurement reliabUiLy. By comparison the repeat measure- ment reliability for the template measure found in the earlier study was 0.93 (co2efficient of concordance). The DC values of the recovery Iii-nbs are plotted against tile rate constants found with each coupling constant in Fiaure 12., As expected, rate constants are increased as compared %vith th@-, DC value, with greatest acceleration fcund when shortest coupling constants were used. '\Ioreo,.,er, at any given coupling2 constant, those recovery limbs having the sho:etest rate constants Gongest tifne constants) are affected the most. Be- cause of this, the relationship benveen DC rate constants and those found with capacitance coupling is non-linear. Short coupling constants not only cause a greater scattering of points, making the measure less sensitive, but also affect the slope of the relation as seen in Table 2. Since, as discussed elsewhere, the rate constant of the first derivative is the same as that of the primary wave, one may wonder why intermediate coupling con- stants, e.g., I second, do not give as good a relation. The answer lies in the fact tl-,at the time constant for obtaining the first derivative must be so short that all responses, long and short, are affected similarly. Witi t3he 29a RC I Sec 0 C A @D RC 3 Sec 0 00 0 0 0 0 0 0 0 0 c p 0 0 0 Figure 12. Relation of rate constants of direct coupled responses to rate constants r,omputed on same waves recorded with capacitance couplinc,,. 2 9b RC 4 Sec CAt RC 6 Sec 29c RC 8 Sec CA? RC 10 Sec 30 Lntcrmcdia te coupling constants used licre, tile slower waves N,;ill be differ ciitially affected and the relationship breaks do,.Nrn. These results imply that the use of a capacitance-coupled system having a time constant of 6 seconds or longer is suitable for the de-.erminati0on of recovery time constants. Though values obtain ed from such recordings are not directly comparable with tjlosc from DC recordi-rigs, they have similar capacity to reflect changes incidental to shifts in behavioral state. 31 6. TFIE- IiNFOr,.,%[.,\TIO,\' COiN-A@-EiNT OF TIIE- l@E-CO%'Ei\Y ELr-'CTRODEr\,\IAL rE-SI"ONlr'>E t@sliowed that during an clectrodcrmal res A previous 2study poiise (E-DR) there is frequently a sudden decrease in tllb hydration of the covered skin surface, and that this plicnoi-nonoii occurs mucli more often during co.-iiitive activity such as listeiiing to instructions than N@,itli startle responses. These liydra- tion responses, Nvhicli are attributed to absorption of water from 2the surface, commence at about one secord after the onset of the skin resistance response (SRR) and reach peak in about three to four seconds. Their occurrence is not determined by the ampli- t-ude of the SRR but is associated Nvith the presence of positive waves in the skin po-, tential response (SPR). @Vhere pure ne(rl-)tive SPRs occur, hydratioli under the electrode e2ither increases (Figure 13) or remains unchancred. These obsery@ations sur, est that .9 two different kinds of EDRs might be occurring, a supposition consistent with con- clusions drawn in an earlier study Moreover, since skin conductance is in part determined by the level of2 fluid in the sweat ducts and by the hydration of the corneumi an absor-otion res- ponse micht be expected to speed return of resistance or conductance to base level and thereby steepen the recovery limb. An indication of t@is is seen in Figure 13, where recovery rate of SCRs in the second panel averages twice that of the first2. The absorp-tion response appeared to-hold promise as a useful qu alitative index O.' behavioral set, bull- there are serious d4.j'ficult,.es in its dire--t quantitative measure- mer,it due to the conplicating effects of simultaneous sweat secre@Lion. It was hoped that the recovery limb of the SRR or SCR micht carry tl-,e iTLformLi4lion reflected i-,l the absor@,",'c@n rcs:Do.,isc ard at *the same time be more ame-iabl-- to qua,-,@icati%,o measure- <> 32 mciit. This scctioii is concci--licd @%,itli tl-c of c,,ua,itit;-,tivc methods for examining the rcco%,cry rate of the L,;DR ird witit an o.,-aniina@ion of its relation to other measures and its sc,isitivity to 2behavioral state. Method t@,,.iluatioii of the Rccg@v r@R,,te had developed a measure called the Recovery Quotient for describinc, the rate of return of skin resistance to base line after re:s- 0 ponse to stimulation. They did this by 2determining the percent recovery reached in five minutes after peak displacement. Their measure ordirarily was applied to a complex toncr-lasdrcr response and they interpreted it as an i ndication of the capacity of the central nervous system to reestablish homeostasis follo%ving a disturbance. The Recovery Quotient did not deal Nvith wave form and theoretically should be unrelated to 2the measure of concern in this paper. For the present purposes, the recovery limb was, as a first appro%imation, assumed to be exponential, a conclusion also reac!aed by One fundamental characteristic of such a curve, its rate constant, or i.,i reciprocal terms its time constant, is inde-endent of amplitude and was selected as a best first appro-.i,mation of recovery rate. It is viewed as a 2useful reflection of this rate, but not as indicatina, a truly exponential form for the recovery limb. Exami- nation of the exporential equation reveals numerous methods for evaluating this con- sta.nt, but t-,-io convenient ones were adopted for use in tl-ds study. For either one, I)C recordings are mandatory. Half-time measure. To determ. inc,@ the tiric constart (tc), one should measure the time reaji red to attain 631 recovery (i.e. , 1 - lllc), b,"t rcco%,e.-N haLf-ti@-,Ie 33 1 - 43 t/2) and is more cas ily measured. It can be cluickly dctcr- relation to cc (tc mined by ttic use of a transparent overlay containing a series of parallel horizontal lines bisccted by a single vertical line. The central parallel line is made longer tha2n all the rest and contains a metric scale on the rig'it side of the vertical, with zero at the intersection. The vertical line is made to pass throu-h the peak of the wave perpendicular to the base line. The template is moved up and down until the central line is midnvay bet@veen onset and peak of.the wave, as indicated by the short parallel lines. The 2distance from the ver-Lical line to t[ic intersection of the central horizontal line with the recovery slope is read from the scale and converted to half- time by the appropriate Qalibration. This method ca-a only be usea on responses Nvmch recover at lear-t 50/10 before a second response occurs, Curve-matching. A second method is based on 2 curve-matcEn,-. As seen in Figure 14a, if zhe members, of a class of responses of differing amplitudes all have the same recovery time constant, it is possible to superimpose them on a single exponential curve. One may use a transparent overlay bearing a family of exponential curves of k-no@va time constant (obtained, for example, by recordin,-s of a condensor2 dischar e through various resistances). Thb tc for any resp@)nse mav be quict-,Iy deter- 9 mined by placirg the overlay such that its base line (asymptote) Is horizontal and passes through the point of onsell- @@)f the response. A straightedge is then held against the loi@-er margin of the overlay, which should be parallel to the b0ase line, and the overlay is moved horizo-,itally until one of the standard curves coincides %vith the recovery limb or a best i@-iterpolatio:i is made (Figure I'llb). '@@",ch of the a@-,ove methods has its advantages and disadvanta(@,es. Tle half-time -a,uno I-epuauodxa u-e uodn p,?soduluadns Z)I-C.1 laDA00al @ures inq sopmt @o C[ .IdLur 3uaiapTp jo sasuodsax Iruliopoiloala aailla ;o lui,-A-c-i a-i 33b Figure I'zb. Illustrat-ion of curve-matching to deterrni-ne the "time constatit" of the steenest por-don of the recovery Umb. 34 bc evali,,aLed. The not recover by 50@" prior to oiisct of a subscqtient %@.,avc aiill cai,,,,iot curvc-matcl,ilig MCLI10d is SOIIICW',Iat MORC (liffiCUlt; it soon becomes clear t2hat many N%,aves depart from the cxtoiiciitial sliapc and niatcliiii- is -anibi,(,,-tious. HoNicvc r, by using the early portion of recovery just arter the steepest slope has been reached, satisfactory matcliiii- can be made for practically any response, Since tl-ds can be 2 done on a portion wliicli reaches only Z5 to ')01/'o of rccovc::y, many more waves lend themselves to tWs measure than t.) the half-time measure. As might be expected, curvilinear recordinm produces a distortion in the recovery limb which becomes appre- ciable if the writeout extends beyond the middle tl-iird of the 2 scale. The half-time measure is therefore recommended for curvilinear recordings since it is possible to make objective corrections for this effect, for example, using an overlay Nvith an appro- priately curved -vertical line. Corrections for the curvili near effect in curve-matching is considerably more difficult and more approximate, and it is recommended that the use o2f this method be restricted to rectilinear recordinas or to the middle third of the curvilinear scale. Procedure Subjects were 106 normal adults of both sexes, ao-es 17 to 45. The number used in each experiment varied and is indicated in the presentation of results. Each e%-peri- ment was run on a different population. Electrodes were silver-silver chloride pl2ates applied to masied sites, 1/4-inch diameter, Nvith a thickenina agent in 0. INl 'L\aCl as an electrode paste. A constant current brid-e (current density 8 niicroamp/cm2) was used to obtain SRRs and a constant volta-e bridge (0.7':-@ V across t-,vo active sites) for SCR' All exosomatic electroderrr,-6@l placcmen@,s @@.,erc on the volar hand. For polcn';al the reforence clo-trocie %vas over the ulziar bone, onc-fiftli t!ic distance from cll)ow to %,irist, Fl@,clr,-Lioii was nicasured N@,,itli the contact device described clsc%@,licrc the placement bcin- on the volar sur- 2 C-1 face of tllc fin-crtip. Rccor(lin(-, of the above variabl es was accomplished using a dircetly-coupled Boc'%-man Type R Dynograph. In on(-- series of runs, a reflectance plethysmo-r,ipli unit was applied to the tlicnar eminence of the non-preferrcd hand and finger pulse volt.,2me was recorded using capacitance-cou-.)Ied amplification The subject was- seated in a small room and after a 15-minute st,--biLization period was exposed to stimuli which differed for each experiment as described below. ExiDeriment 1. This was a replica of an experiment by ',--,,ihich was run for another purpose. In this the subject is given a 3-mi2nute rest period followed by a 1-minute task in which he is frustrated by the experimenter in his attempts to count back--@vards. He then participates with the experimenter in a guessin- game in w-dch the experimenter attempts to guess the number which the sub- ject selects. If he fails, the subject presses a button which supposedly shocks the experimenter. Follo%vin2- this t@vo-minu,Le "agoressive" game, he is criven a second rest -period. Ex-oerimeTit 2. A mixed series oi'L eight tones and eicht light flashes of moderate intensity is presented to the subject who has been instructed to relax @vith his eyes open. Follo%,,-ino- this he is instructed as follows: "This time, when you hear a tone it will be a warnina- signal fcr a second one which will occur at any time up to a half-minute C) later. At the sound of the second tone you are to press tl-ds footswitch as -apidly as poss-@ble so l@;c can measure your reaction time. In the same way a lirht flasli be to C..-.pcct I sccoiicl lil-IIL fllsh at tinle @-ou arc to 10',.C tilc FOSitiO.-I Of the nio%,in- pointer. A ti-dr(i ilasli %,;ill occur sliorlty after tliit, at which time you arc 0 to repor-t the letter izidiciti2n- its position. " The inter-stiniulus interval for each task ranged froi-n 10 to '00 seconds. The pointer rotated at 60 r.p.m. This series @vas continued undl eight trials of each task- had been run in randomized order. Ex2criment 3. The subject was told to rclax Nvitli his eyes open and .was ex- posed to a series of five tones (approximately 75 db at the subject, I second duration) presented throu(,,Ii a speaker at a varying intcr-trial interval (10 to 25 seconds). As soon as tl,,is habituation series ended, he was instructed for the next sequence which followed immediately. In this series, tones of the sai-ne intensity and ranoe of inter- 2 trial intervals as for the-first series co-,istituted signals for a reaction time efforl (fin-er-press) -without a foreperiod. ExiDe-Liment 4. The subject is criven two reactior- dme t-,4Lals with a foreperiod of 4 to 10 seconds. Both warrliii r and execution signals are 0. 5-second tones delivered f2rom a speaker (approximately 65 db at. the subject). Tliiese trials were intermixed in ra-,idom order with t@@vo word association trials to form a block in which the inter-trial interval varied between 20 and 60 seconds. Blocks were repeated Nvith t,.vo-minute interveril:-,io- res,6- periods. Ext)eriment 5. In tMs series, run ea2rlier for another purpose the sub ect wore a contact hydration detector as well as skin resistance and potential electrodes. Stimuli were a mixed series of sounds and lichts of moderate intensity i.-iterrupted at intervals Nvith a period of conversation. 37 Results and Discussion Characteristics of the Rccovcrv \Icastire ReliaLl!!Iy. Inter-scorer reliability of the more subjective of the t@vo methods of measuring recovery rate, namely curve-matcllin-, was dt-,termined by having each of three ind2ividuals score the same 30 isolated rcsponses chosen at random by a lottery method from a population of 100 useable responses. Time constants ranae,_l from 2.2 to 7.2 seconds. The Ke-,idall coefficient of concordance among their scores was 0. 93. The time consumed in making a single measurement was appro,-,Imately seven seconds. Despite the high inter-scorer reliability in2dicated in the measurement of Isolated responses, i.e. , responses whichoccurred at least five seconds after the peak of the Iast previous response, there is an apparent source of error in the measure- ment of recovery rates of responses which themselves fall on the steep portion of the recovery limb of a previous larger wave. For this reason, scoring of responses is best accomplished on waves which otcur at least 2ten seconds after the peak, of the preceding wave. Relation between tc and t/2. The reliability of the balf-time measure should be even greater than that of the curve-rnatcl-dng measure, in view of-the -rigorous method by which it is obtained. It does, however, frequently demand the reading of very short time Intervals with precision, a chore which can be difficult when for conven2ience in making other measurements, slonv paper speeds are used, -e.g., 1 mm- per second. To see how the two measures compared, 22 consecutive responses'recorded with rectilinear pens were analyzed by t@vo persons, each using one of the t@%ro methods. Time constants ran-ed from 1.6 to 8.7 seconds. The mean values -Lor tc and t/2 res- nectivelv we re 5. 40 and 35. 83 sccorlds. The ratio of these values, 1. ' )9, a,-r--es rather 33 i@-ell Nl,,itli the theoretical ratio of 1.43. The product-monient correlation be,@,,,Ccn tc and t/2, 0.94, can be regarded as a combined test both of the accuracy of measure- ment and of the validity of the exponential treatment. 'fli-is result, together with the 2 1.39 ratio, implies that either measure is acceptable, dnd that the portion of the curve used does not deviate appreciably from an exponential form. Resistance vs. conductance responses. In some instances responses were measured concurrently from nvo sites, one Nvith a constar..t'Voltage bridge, the other with constant current. Recovery half-times %vere det@,-n-ni2ned for corresponding SCRs a.nd SRRs and found to be highly correlated, though different in absolute magnitude. Correlations between 20 pairs of responses on each of three subjects were .33, .82 and .95. Mean values for measurements on SCR and SRR respectively were: subject A, 4.9/6.3; subject B, 3.6/7.3; subject C, 3.6/3.8. Thus although either SCR or 2 SRR may be used for studying chan-es in recovery rate, if individuals are to be com- pared it would be desirable to Standardize on one of the nvo systems. Differences be- tween the mean SCR and SRR values, however, may well --eflect differences between sites rather than between methods of recordin-, since despite high correlations, appreciable differences are found between measures tak-e2r. from simultanous palxnar and dorsal SRRs as seen in the example in Figure 18. Relation of rEERI@@@@ to amulitude. The assumption of an exponential form for the recovery limb implies that the rate constant should be independent of amplitude. This assumption was tested by an examination of the relal,-,io-.i of recovery rates to ainplitude using the half-time measure on 20 SCRs from each of six subjects. The -product-morrient corre lations were .21, -.20, .28, 62, @7 4 and .19. These results AWL ^9 0 t,,@,o may for sorne sLI)jccts be related tilrou,,Ii a co,,iinioii irLRticlicc. Tile Positive sigi,is for the- t-,vo si-idficant correlat-ioiis (.62 and .71112) @voulct inciicatc ti,at there is cct s for responses of 1-dgllcr amplitudes to I)c a tendency in some stibi associated with slower recovery rates. Relation to Positive SPR - From the relations described in the introductory secti2on, it was predicted that positive SPRs shoulcl be associated with fast recovery rates. Out of "30 subjects run in Experiment 2, 13 showed both clear positive SPRs biphasic) and "Pur3" ne-ative SPRS. These were used for -he comparison, Since the magnitude of positive-@oing activity associated with a negative SPR cannot at present be quantitatively evaluated 2 a test of associadon rather than correlation was made. In each case two clearly utiiphasic nefative res- ponses and two bi@hasic rosponses N@,-ith a pronou-,iced posit:ve deflection were selected from the skin potential record. The ti-ne constants of the corre@pondinc, sMn resis- tanc& responses were measured and for 12 of the 13 subjects the ave2rage time constatit associated with positive SPRs was shorter than that associated v.-Ith "pure" negative SPRS. Tne respe--tive group means were 3.7 and 7.4 seconds @p < .005). Amplit-@.,des in the t@vo categories were not si-gmfic-,,ntly different, although responses accompanying positive S?.Rs tenied to be somevihat2 larger, the mean log ratio being 0. 126, equivalent to an amplitude ratio of 1. 34: 1. If anything this would reduce the differences bet%,,,e'en time constants rather than account for them. The relation of recovery rate to occurrence of absorption responses was also exa,-nined l'or an associative relationship rather than a correlazion be--ause absor-@-,"ion, Li'lle the SPR, is Oifficult to measure quan@,italivelv Data e@n -'Hp lik@cl-@--itiorl 10 -3 slio%@.,ing mar@-c(J z)j)sorptio,,i and t-\,.-O siio\,.,iii- an i-,lcreasc in li,'d,-atoti were selected in a randz)niizccj niaiiiier, @,.,itliotit regar2d to stiniu(its. The time constatits of the recovery limbs of the corresponding SRRs %vere measured and an lntor-ilidi@vidual comparison made between the two categories. In 12 of,tlio 13 subjects examined, recovery was faster in association with absorptioii (p < .001). The group means were2 6.0 and 9.1, The association of faster iccovcl-y rate with the occurreiic'c of absorption responses a.nd with positive SPRs is consistent with the third combination of these variables, narnely that between occurrence of absorption and of posit-.vc SPRs previously reported Sensitivitv to 13chavioral'State Both me-,2sures, t/2 and te, proved highly sensitive to change in behavioral set. In some cases, for example in the comparison of a rest period @%,ith a task period, the conventional ampit-ude or frequency measures would discriminate just as well, but the sr.-en-gth of the recovery rate in differentiating between ccriditions did not de'Dend upon an. plitude. 'ivforeover, in many cases, it distin-,ruished between conditions when a-Tiipli- tude per se could not. In this initial examination of the resolving strength of this measure, t/2 has been sampled in some experiments, tc in others. Each clearly demor.strated its value, and the choice of which to use is lar-ely a matter of preference. The results of the variou2s experiments are presented belo%v. ExDcrirrient 1: Rest vs.-ai@,,gressive game. In examining differences between the recovery limb lialf-times of spontaneous waves in the rest period and those durincr the a,(-,rcrressive g-@me, an effort was made to minir.-iize anv intrusion of an amplit-ude effect. The fi9rs.' three rosponscs of at leas*6 5 mm an, p',-itude met the liall-tim, e 0 subjects thcsc were taken from the pro-task- resc pcriw, for the other half from the Post-tasL, period. 'flic first tlirce responscs from the game period wliich fell in the same amplitude range as those of the rest period were used for comparison. Figure 15 sh2ows representative strips taken from each of these periods. The pre-task and post-task relaxa'ion periods (upper and lower panels) both show responses which re- cover more gradually than those during the task period. As for the other subjects, t/2 in the pre- and post-task periods are rather similar. It is significant that the sharpening of the recovery limb started dur2ing the instruction period. A portion of the instruction period is shonvn in the section of the middle strip to the left of the arrow designating start of the task. The group mean t/2 for the rest period was 5.6 seconds and for the task-'period 3.3 seconds (p < .001). All of the 12 subjects showed a reduction durino- the aggressive game, the mean decrease being 411/70 (Tables 3 and 4). 2 The amplitudes of the t@vo groups of response samples were well matched and the small mean difference of 1270 Nvith the task, responses being larger fell far short of significance (t 87). Experiment 2: Sim2le stimuli vs. _@!zna.1s. The half-time measure was also used in this comparison made on 16 other subjects. Here the responses to light fla2shes during the relaxation period before instructions were con--pared with responses to the warning lights (not the execution signal) for the perceptual task. The last three useable responses during the.rest period were compared with the first three useable responses during the task period. An acceleration of recovery rate occurred 4tv 13 of the 16 sub- jects (Table 4) when the light flash took on signal prop4erties. The average decrease in t/2 was 291 (p < .05). The amplitudes of the response samples avera-ed 9-' less ,,O /0 durirg t@e task period (t = 1. 14, NS). 41a CAL REST 10,u TASI( s PTtTASr% F,' S T T/2 Figure 15. Recordinrrs of skin f-o-cL%,,',uct-,,-nce during pre-task rest (upper), a instr,,jctions and task (middle) and post-task (lower), showing acceleration of recovery rate durinor instructions and task-. Illustration o-F half-time measure is shown in lower trace. Time line is 10 seconds. ANIL '2 -i Table 3, Comparison of a,,-crage values of reco%,cry lialf-tir,-ic (t/2) for resting state and gucssin- game task- in 12 subjects. t/2 (seconds)2 Subject Resting Tas< % qhange I 6.7 5.6 -i6 2 5.0 3.6 -28 3 4.9, 3.i -33 4 2 5.7 3.1 -46 5 5.3 2.9 -45 6 7.5 2.8 -63 7 3.6 3.1 -14 8 10.0 2 5.0 -50 9 3.9 3.2 -22 10 3@9 2.9 -26 11 2.2 1.7 -23 12 8.1 3.i 2 -63 Table 4. Summary of results from three experiments shdwino, change in recovery 0 rate Nvith change in stimulus condition, S@-mbol "n" designates number of subjects who showed a decrease, Mean 2 Measure Change N 11 Used Condition A Conditior- B A to B P- 12 12 t/2 Rest Aggressive Game -2.3 <.001 5.6 sec 3.3 sec (-4170) 16 13 t/2 8 Liiht Flashes Perceptual Task -2.3 <.05 7. sec 5.6 sec - 2 9 35 32 t.c. Tones Reaction Time 5.5 <.001 10. 4 sec 4.9 sec 43 Ex-octinient 3: Simole tones v@, reaction time cffo.-rt. The cur,,c-matcldng method was used to dctcmliiie tc on the 35 subjects used in this experiment. The largest response to the five tomes presented durin,-, the habituation series was Com- pa2red ,vitli the response which was nearest in amplitude during the five reaction time trials. Of the 35 subjects, 32 showed acceleration of the@ rec overy limb durin-Y the reaction time series, as seen in the representative example if Figure 16. The mean decrease in recovery limb tc for the group was 53 percent (p < .001). Tne mean tc presentation of non-signal tones was 10. 42 seconds for this po@ulaticii (Table 4). In Experiment 2, the mean haif time for responses to non- signal li-ht flashes was 7.9 seconds. Convertincr this to tc by multiplying by 1. 43 gives 10. 3 seconds, in remarl,-ablc-- ag-reement %%,ith the results of Experiment 3, This similarity though perhaps fortuitous provides an added degree of confidence in 2the measure. Change in Arousal So far this section has been concerned with the power of the recovery rate in discriminating bet-%veen responses in t@@o qualitatively different stimulus situations. It can a parently also distinguish benveen t-,vo similar stimulus periods in which the state p of the su7oject is different. Representative examples may he2lp illustrate this point without statistical treatment. Figure 17 shows strips from two consecutive blocks in Experiment 4. In trace A, the subject is exposed to his first block of reaction time and word association stimuli. In trace B taken from a later block, the subject has apparently habituated to the situation, as indicated by reduction in response arnputude and by the fact --hat2 he has ceased respondina to the alertina- signal for the reaction 0 0 time trials. The slowincr of the recovery limb @@rith reneated trials is espe--ialiv marked R T Figure 16. Representative skin conductance recordings from Experiment 3 sho%ving acceleration of recovery rate during reaction time trials. Time line is 10 seconds. RT W- A RT RT VJA RT Figure 17. Recordinas of skin resistance from blocls I (upper) and 2 (lower) of Experiment 4 showing slonvincr of recovery rate with adaptation. Stimuli are 0 reaction rime trials (PT) an7d word association trials (WA). 'Vertical caubratio,,i line is 25 K; time line, 10 seconds. '@Resistance of unper trace is 200 K, lo%@ier trace 190 K. ,14 tendency, only 7 of tli,2sc sl,,owcd an increase of o%,,cr ')O,;'O- stiiguisli bct@,.,ccn respo,-,scs to dissimilar stimuli 'Me recovery Limb may also di in the sm-ne task pericd. An example is shown in Figure 18, taken2 from E,%periment from the dorsal and paimar surfaces of the fin-ers 2. It shows simultaneous recordings 0 of nvo individuals. 'Me letter A indicates an alertin- signbLI for a forthcoming reaction time trial and E, the execution signal. S denotes a spontanco-as wave occurrin- during the f2oreperiod. The number below each respclnse is the value of its time constant. For the first subject, the spontaneous response has a-bout the same time constant as does l@ie alerting r4@sponse, but the execution response is considerably faster. The dorsal and palrnar surfaces reflect this same pattern despite absolute.differences in their recovery rates. For the second subject, tl-ie constant for the spontaneous res- 2 ponse is over twice as Iona,, as that for alerting; the constant for the execution response is also longer. The difference between recovery rates of responses to alerting and execution responses is in general not a reliable one, but six of the 19 subjects did show an appreciable difference bemeeti the nvo. Two Nvere si --nificant at the . 005 level of 2 IT corlffdence, three at the .025 level and one at the .03 level. In five of these six cases, the shorter tc occurred Nvith the ececution signal. Individual Differences Tae sensit!N-ity of the recovery rate to differences in stimulus condition and in state of the subject sLLgaested that inter-individual differences in recovery rate, seen for example in Tabl5e 3, might be associated with differences in characteristic behavior pati,erii. 'Me behavioral index chosen to test this idea, namely rate of habituation to tones or to reaction time trials, was derived from Experi ment 3. The qi,,otient of t-he 40 Adlilk 4-lia A S 7.2 5,p 3) 49 5.0 2.8 4.1 2 5.7 3@ 3P Figure 13. Simultaneous recordin(Ts of skin resistance from dorsal surface of 0 ub;ects. Stimuli are the finoer (ijpper) and palmar surface (lower) from t"vO s ale-,ting tone (A) and executio2n tone (E) for a reaction dme trial. S denotes a spontaneous response. Numbers denote recoverv limb tim-a constants in seconds. 'Tirrie line is 10 seconds. link, 45 of habiwation rate, hi(,Ilcr values indicating faster habituation. Time constant measures 0 were talen only from tlic.rcaction tinic series and for ')5 of the subjects were the same as obtained for the earlier analysis of this experiment2. In 18 other subjects who showed no responses to the tones, the response of median amplitudc was used. The group correlations bet-,keort this time constant and the various measures of habituation are shown in Table 5. Comparison of tc N@,itli rate of habituation of the reaction time res- Table S. Summary of results from Experiment 3 sho%ving product-moment correlation bet@2veen recovery rate and habituation rate. Habituation N Measure r P- 53 SCR durino, .33 <. 05 0 Reaction Time 32 SCR during .13 NIS Tones 51 2 FPV Change .30 <. 05 durin- Tones ponses (r .33, N 53, p <05), shows that individuals @@,ith fast recovery rates tend to habituate more si-owly during the reaction time series. Of these 53 subjects there were 32 whose records permitted analysis of habitua- tion rate of the SCR during the tone series, The correl6atio-P between this measure of habituation and the -ime constant was very low and not si.-nificant. However, when. instead of SCR, fincer pulse volune chan(re was used as the response measure, the rate of habituation to tones sho%ved a si.-nifican'L correlation tc (r l')O, iN = 51, 46 gression. T'he tvo positive correlations found bet%,,,cen r-2covery limb tc ard different habituation measures indicatc that individuals with fast rccovery rates tend to habi- tuate more slowly (Table 5). Conclusions It has been demonstrated th2at the time constant or the half-time of the recovery limb of the exosomadc electroder'mal response can be used to distinguish bet@,ieen a restino, state and a task state, between responses to the same physical stimulus under qualitatively different sets, bet@veen responses to the same stimulus at different stages of habituation an;i finally between individuals2 Nvho habituate at different rates. In some cases, significant discrimination would also have been rrade by treatment of amplitude data but in other cases, amplitude could not have distinguished between the r@,;o condi- tioris. The rate of recovery accelerated markedly under three conditions, all of which involved activatioii from a resting state for performan2ce of a task. However activation per se does not imply acceleration of the recovery limb. Bursts of stroncr spontaneous activity during "relaxation" periods were composed of EDRs showing the gentle recovery Limbs characteristic of isolated responses elsewhere in -.he rest period. A reasonable interpretation is that fast recovery rates reflect a mobilization for goal-directed be- havior2. Spontar-eous activation during rest does not possess this quality. To the extent that slow habituation represents longer maintenance of a set to respond, the inter-indin,idual relation between fast recovery limbs and slow habituation rates is consistent with such an interpretation. For the electrodermal (but not the vasomotor) response the difference in the relationship ben@reen7 tc and lia:oiti-,ation rate for the reaction tl-ne se-.ics, where it %,Iras appreciable, and for the tone se@ries, where it was.nccrL,.-ble, 0 0- T',,,csc also ai impo,,:-Lart iiiii)lica,,io,-i in to the nature of cicctrodc-.rlat -activity. 'llic rccovcry Liiiib time coiistaii@@ is a reslioiise cli-,ractcris- tic by which two exosomatic clec@Lrodcrnial responses nia,,- bo qu@,.Utatively differentiated .V according to the nature of the stimulus. Another nictliod is by comparison of relative amplitudes of palmar and dorsal finger respoiiscs This stimulus specificity, to(,ctlier with the independence of tli:a recovery rate from a d2irect amplitl,de effect, its relation to the occurrence of positive components in the SPR and its relation to the occurrence of absorption waves suggests the C.'@-,istence of a second component in the EDR undar independent control. The steeper recovery slopes apparent- ly associated with goal-directed activation are viewed as representing the activity of a s-6%,,,eat reabsorption medhardsm which in some way serves as an adaptive process duripo- such beha%-ior, v 7 TI-IE- ELECI'I',CDERNli@L r,,T-CONIE-l@Y LL.%IB AS -,@N 1:\DL-X OF, COAL-DIRT--CTED BLflj%VIOR It was demonstrated in Section 6 Lliat the rate of rceovery of an cxosonlatic clec- tr2odermal response varies as t function of the state of behavioral activity of the in- din-idual, becoming faster when the subject clianaes from rest to any- of a variety of tasks. The rate of recovery, %vliicli was expressed in terms of the time constLT',t (te) of the exponential re,-,ion of. the recovery limb, is faster in the presence of positive- going skin potential responses, -and in the presence of the recently discovered sweat reabsorption response Although observations in the initial study appeared to indicate that acceleration of recovery rate indicates a mobilization for goal-directed behavior, the evidence Nvas sug,-,estive rather t@an definitive. The tasks and stimuli used generally caused an increas2e in activation level as reflected in ele--tro- d--rmal activity or fir-crer vasomotor activity, and an equally t:anable hypotl,.ests would be that acceleration cf recovery rate merelk reflects @@ncreased general activation'. The obvious experiment needed is one in which activation is a--bieved in one instance by task perfori-nance, and in the other by a none-tas2l,, conditio-.i, for example, noxious st:imulation. Such an experiment was one of the key aims of t-'ds experiment. iln addition fur,--her data to'evaluate the de-ree of the deperdence of recovery upon such things as response amplitude and base level were needed. Information was also i--eded in regard to the stability of this measure in a given individual over a period of time. 0This section deals with additional findings relev-lnt to some of these questions. od In one -hase cf this experiment a group of 16 m-,Ie medical stL:doi-,'Ls @,,-as -,:,I on 49 N@,cre monitored. Twelve of tl,,csc subjects @,,,crc tested on four successive occasio-,is one week apirt, under the same stimulus schedule, Five additional subjects were run only on selected tasks from this list. 2Electroden-nal electrodes were applied, after iil-dch the subject was seated and a stabilization period of 12 minutes allowed, The follonving schedule of conditions was followed %%,itliout randomizincy the order. Constant order was used to allow maxi- mum similarily of the 5 repeat sessions despite the danger of an uncontrolled order effect.2 Sequence Duration Situation 1 2 i-nin. Relax (eyes open) 2 1 1/2 min. Cou-it by ones at own ace p 3 1 1/2 min. Cou-it backward from2 500 by 7s. A new startino, point near 500 0 was used on each successive week. 4 1 1/2 min. Take deep breaths on command (3 at app2roamately 30-second intervals) 5 1 1/2 min. Read aloud 6 2 min. Mirror tracing 7 2 min. Relax (eyes open) 8 40 sec. Cold pressor (hand immersed to 0 ct in ice water) wri. Time constants were measured with the curve-matcmn(r overlay technique described earlier, the farst useable @vave follo,,vir.9 the first 15 seconds of each pro- cedure bei-ir taken as the sample. One req,,iiremert was that the wave used should 50 their recovery rates. 'nic an,cra-c of the initial and final rcsistance lc,,,cls for each of the ei-ht procedures was coriiputed for each subject. Results Effect of Task, and Activation on Time Constan t 2 The d-me constant (tc) was found to vary considb--ably as a function of task- situation, as found in the earlier study, The range of time constants ran from an average of 3. 1 seconds in the mirror tracing task to 7. 8 seconds for resting with eyes closed. The time constants for the 8 task conditions @vere ranked for each of 16 siib- jects and the average rard,- for each task computed. Results are seen in Table 6. The Table 6. Averaoe ranks and standard deviations for each of 8 task, conditions on 16 0 subjects.. ConditLon Order Run Rank S.D. Rest (eyes closed) 7 2.5 (2longest tc) 1. 5 Rest (eyes ooen) 1 3.2 1.5 Cold Pressor 8 3.3 2.3 Deep Breath 4 4.2 2.2 Count Fonvard 2 4.8 2 2,0 Count Bac'zi@,,ard 3 5.2 1.9 Read Aloud 5 5.9 2.0 lurror Tracing 6 6.4 1.5 average ranks have b-ten arranced in order Nvith the longest tc a7t the top. ne order in wldch these procedures were run is also shown, and a Spearman's rho test irdi- cated no sio-nira:--ant relarion benkeen the rarl',s of the time constants and the order rim (rho = -.05, N. S.). AWilcoxon signed-rar@,zs test sho@ved that many of t.Liese 0 51 test) are shown in Table 7. Slo\vest recoveries were associated with the two rest conditions and with the cold pressor exposure. 'Dic remainin- conditions appear to rank themselves essentially in order of iiicreasing task complexity. In view of the fact that recovery rate during the ice water exposure was not 2 significantly different from that at rest, although that of.the last 4 tasks was, it appears that activation per se is not an adequate condition for acceleration of electro- dermal recovery. One may challenge the contention that the cold pressor was as activatin- as mirror tracing, but the report of stron(- pain by 14 of 16 subjects would suggest that there at least occurred the kind of arousal associated with 2strong noxious stimulation. Moreover, mean skin conductance levels fcr the test population were very similar'for the col(f pressor and mirror tracing corditions, being 28.1 and 27.6 micromhos/cm2 respectively (t = 0.36, N.S.). The rate of chan,-e of skin resistance during these two tasks also did not differ (t = 0. 12, N. S.), nor did the frequency of spon2taneous responses (t = 0. 12, N. S.). Thus, at least according to a few so-called el--ctrodermal activation measures, activation levels during ice water exposure and m-'-rror tracing were not demonstrably different, although both differed from the resting condition. Twelve of the above 16 subjects were run once per week for 5 weeks a,-id an analysis of variance was performed 2on the time constants (Table 8). This also Oemon- strated a signif[cant difference between tasks (p <.001). A final test of the sensitivity of tc to differences in task situ,'Ition was by analysis of the similarity of rankings of these tasks across the 16 subjects, using Kendall's Coefficient of Concordance. This confirmed the si,-nificance of the dlfferen- 5 tiation (@V = .31, X2 = 34, p <.001). 52 Table 6. Averige ranks and starciarcl, deviations for cacti of S tas'.@ conditicrs on 16 subjects. Condition Order Run Bank S.D. Rest (eyes closed) 7 2.5 Oongest tc) 2 1.5 Rest (eyes open) I 3.2 1.5 Cold Pressor 8 3.8 2.3 Deep Breath 4 4.2 2.2 Count Fon%,ard 2 4.8 2 2.0 Count Backivard 3 5.2 1.9 Read Aloud 5 5.9 2.0 Mirror Tracing 6 6.4 1.5 Table 7. Levels of significance of Wilcoxon signed-ranks test for various con-Lbinations 2 of task situations. Back-@vard Deep Cc-Id Rest (Eyes Rest (Eyes Condition Counting Breath Pressor Open) Closed) ,Mi-rror Tracing .05 .02 .01 .002 .005 Reading Aloud .203 .03 .02 .005 Back-%%,ard Counting .005 Deep Breath .1 Stability of the Time Constant in Receated Trials Not only did the tc offer a stable index of differences bet'Nveen task s2ituations, but also of differences bet@veen subjects. The analyses of variance of the tc's of 12 subjects run in t:ie eight task situations on each of 5 consecutive weeks (Table 8) demonstrated a significant difference between subjects (p <.001). In add,,tio,-.i there was a si-0, ificant task-by-subject interaction Co <.005) as 2 well as the difference be- t,,veen t--s':'-S al4-eady noted. 0 53 Table S. Analysis of N-arian,cc of time constant data from 12 subjects, run in 12 task situations on 5 consecutive wecks Mean Source df Square F p Subjects 2 11 109.7 14.5 <.001 Ta s k.9 1 355.0 47.0 <.001 Subjects x Tasks 11 28.4 3.& <. 005 An indication of the stability of this measure on repeated trials was obtained by examining the de-ree to which the 12 subjects maintained the same rank for2 a given task, over the 5 sessions. The KendaLl Coefficient of Concordance for the 2 Backivard Counting task Nzas .53 (,X= 29, p < . 01) and for the cold pressor exposure 2 .65 (,X= 36, p< .001). These results indicate that a given subject may show variation in time constant accordino- to the 2task, but the magnitude of his time constants takes a characteristic position in relation to the rest of the population. 'The 5-wee@,, data for the backward c-@-,Lnting task are presented graphically in Figure 19. Each vertical line represents the 5-week mean and + I standard deviation of one of the 12 subjects. They have been ranked in order of their mean time constants (upper series) and rate 2 constants @ower series). The rate constant scale increa-.es downnvard. L\Tote that fast recovering subjects are consistently fast and slow recovering subjects are con- sistently slow. Note also that the rate constant shows less dependence of variance i,non mean (r = -D. 2 1) than does the time constant (r = 0. 34). Tne solid circles represe.,it the mean rate constants fo3r the cold pressor ccndition over the five weeks aud in everl case they are slower than those for backivard countino,. An e.--,amole of p I the range of time cons'Lant means over the S conditions for the 5 weeks for a single a Figure 19. i%leans and + 1 SD of recovery limb time constants (uppet group) and rate constants (lower group) for each of 12 subjects tested over 5 consecutive weeks. Vertical line data are for the backward countinc, task. So@d circles represent the correspondin(r means for the cold pressor rate9 constants. 0 5 subject chosen ac ranc!om is given in Figure 20. Eacli @y,ertical line represents the mean time constant + 1 stan(iard deviation. Thus the recovery limb measure, preferably its rate constant, constitutes a rather stable measure of differences in behavioral situations, and appears to reflect the l2evel of involvement in goal-oriented performance., Tt is capable of discrimination under conditions in which base conductance level and (from the earlier study) skin conductance response amplitude do not. Relation of te to Conductance Level Although conductance usually increases when a resting subject becomes activated for a task, there is no reason to suspect a necessary correlation between co2nductance and tc, i..e., that the t@vo are causally related. If the type of activation which produces a rise in conduc:ance does not entail mobilization for a aoal-directed task-, it is to be expected that the correlation would break down. Thi.s predictio.1-was tested upon a o-roup of 21 subj--cts. Two tasks were chosen, both activatincr, but only one of em 2 0 th involving a cognitive task. The first was a Cold Pressor exposure, the second, the BacNivard Counting task. Correlations were essentially as predicted. The Pearson's r between SC and tc during Cold Pressor wa@ -. 23 (,N. S.) whereas that for Bac-l,-@,,,ard Counting was -0.49 (p < .03). The @bove determinations are in themselves inadequate to justify the conclusion that ic is independent of skin conductance level. Another an.-ilysis does, however, point toward such a conclusion. The data for the above cor- rel,@ion-il analysis was subjected to a t-test for differences betveen tas't,-s. The time were si@,milli7cantly different (p <.01), but the SC levels were not. JF REC RI-0 CP DB CF CB RA MT Figure 20. %4eans and + I @D of recovery limb time constants for a single subject for each of 8 conditions over a series of 5 consecutive weeks. Abbreviations are: REC resting (eyes closed); REO, resting (eve1s open); CP, cold pressor; DB, deep breaths; CF, countin(r fonvard-, CB, countincr bact,-\vard; RA, reading aloud, @\IT, mirror tracino-. r-',clation of Co,-sLar.L to The earlier intcrprctitioii IliLt the re.-OverY lin,,b ciiilc co;istiit rclaEC(i to mobilization for goal-dircctect activity implied a relation between the degree to which mobilization occurred an(I the level of pcrf-,-niince. The relationships bcc%,Iccn tc and two measur-2s of performance in ttic r>ack--",ard Couiitin ' (by 7's) task were examined. 9 One measure was the nuiiiber of errors made, the other die rate of counting. 2The number of errors was determined by determining the number of times subtraction of seven from die Drevious nun-Lber was in error. 'I-he rate was expressed in terms of total span for die back-@%,ard count in the 90 seconds allowed. Results on a population of 12 showed that subjects widi sliort'er time constants tended to count faster (t = 1. 96) and to make f2ewer errors (t = 1. 82). Lev--Is of confiderce for these, using a t@vo- tailed test, were at the 0. I level. (Note: Althounli the direction was predicted, this investigator has discontinued the use of one-tailed tests). Functional Interpretation In loo.Ung for a functional interpretation, one may point to the results of experiments over the last sever2al years which show the relation of the recovery limb to rate of reabsorption of sweat. In some rec6nt experiments, a iew method has been utilized for demonstratilig this relation. If a prisni is illuminated for maximum internal reflection and placed on the skin as was done by sweat droplets adoear as black points on a white field. 7bis method may be carr'ie@ to its looical next stage, namely, the placement of a photocell at the proper point in the light path to irtegra-,e and follow the field of punctate sweat droplets. Fig,,ire 2.1 shows such a re--ordin-. The abbrc,.,iation OSR refers to N@.,Iiat I call :he oc-tical s%@.,eat resoo.,is--, the C) F--l ;A (D cn El ". 0 4C) CD 2 -0 cn 0 (A m c@, 0 6CD Z3 C-L qp 0 36 covered by a glass plate, do not simply accumulate on die surface during and follow- ing a response, but are immediately reabsorbed, so that d2spfte continuing activity, the moisture content of the surface remains relatively constant. 'Me second point of interest is that when the skin conductance trace shows slow recovery limbs a2s in. the upper pair, so does the optical sweat response as compared with the lower pair in which electrical and optical recovery rates are rapid. Such demonstrations, along with those already reported using other methods, imply that changes in the recovery limb are at least in part indicative of a change in the rate of the reabsorption process. Because differences in the state of hydration of the 2skin surface are known to influence tactile and manipulative performance, one is tempted to hypothesize that control of reabsorption, like control of sweating, may be part of the adaptive prepara- tion for these'types of behavior. This may explain the apparent relation of recovery rate to mobilization for goal-oriented performance. Conclus ions The time co2nstant measure appears to be capable of reflecting differences in '-e- havioral s,.tuations with a relatively high resolution. it h,7s been demonstrated to dis- criminate when the base level measures did'not, and in the earlier study when amplitude did not. It is rel-atively stable over tl-ime, at least insofa-r as the position of an individual in a group is concerned. Finally, since te is l0o:@gest with the subject at rest or exposed to a cold pressor task, and becomes progressively shorter as the task becomes more goal-di.rected fcalculat-ion, reading, and mirror tracing), the findings support the earlier interpretation of tc as reflecting mobili@zation for goal-directed performance. 0 S2 12. @',,Sr'Ul 4R EFFE-.CTS t'PON SL,'IN FO'rl'% - @:,% I Altllou-h' -Iproduced experi- mental evi(iance 2rjiat vascular clian,,,,cs play oiily a ne-ligible role, If any, In "CcOunting for clectrodermal activity, tllc focus of these Oxperinic-nts was upon conductancc clian,cs. The possil)ility that vasomotor effects could account for at le.ast some skin potcntiil clianri-es remained. Tne observation that2 potential re- spctt.ses could bc- recorded from the nail bed which is devoid of sweat giands su<-r,rested that blood vessels mfght indeed account for some pot- enEL'al shilrts observable at the skin surface, TWs studv examines the possible le of -vasomotor chan,,es in producing 2 ro change of surface potential. The approach was to ixiduce alterations in the state of peripheral vessels by mechanical means And to observe any corc@)mitant potential changes not ascrib'able to central reflex effect. Venous occlusion was used as a -iethod for engorgl-ig the veins while maintaining flow and arterial occlusion for 2 inter-zupting flow %vithout engorgement. A simultaneous recording of a no.'I-occluded homologous con@alateral s ivould ser,,,e as a control fo2: centrally induced changes la potential. ivethod St.@bjects v@,ere 11 males and 2 fe,-nales in t,ie age range from IS to 42. Ski0n poter.cial (SP) was recorded from the palmar and the dorsal surfaces of the middle phalan,x o'L tlie mfddle finger of each hand. Each site was recorded with reference to an inactive skin site above the ipsilateral clavicle. This -reference was chosen to be oi.itside Llic occluded re-ion so that comr-.on-mode effects %%@ould ror cancel out elicited 83 changes. A direct-couplecl reflectance pieLh),sii,.ogral)li was attached to the d,)rsurn of the middle segment of the index fin,,cr of the experinicntal arni, I.e., the arm 0 around which the blood pressure cuff was placed2. Procedure The pressure cuff around the upper part of the experimental arm was rapidly inflated to 60 mm Hg or to 180 mm Hg and kept at that p:essure for one minute. Following deflation at the end of the one minute period oil inflation a rest period of one to two minutes preceded the next inflation. Each S received at least 5 trials at 60 and at least 4 at ISO mm Hg. The experimental arm was alternated from one S to the next. Lfflation of the cuff to 60 mm Hg prevented venous return from the arm and lead the arm to become engorged distal to the cuff due to the unimpeded arterial flow. Inflation to ISO mm Hg collapses both the arterial and vecous systems, thereby completely occluding blood flow to the arm below the cuff. Plethysmographic record- ings co2nfirmed these differential hemodynamic effects associated with the t-,vo levels of pressure. Data Treatment The analysis was guided by the need to answer two experimental questions. ations of SP at -.First, does e,.ther of the two inflation pressures result in fluctu experimental sites and not at control sites? Se7cond, if shifts of SP level are demon- strated for the experimental sites, are these shifts different for venous engorgement than for arterial occlusion? Each of Lic four SP channels was scored identically. Since it appeared that the neriod of zreatest change in the t-,,,7o exper,,rre.@ital char-nels occurred about 5 seconds 4 -Z after the release of pressure, this point was chosen as an arbitrary reference. The slope of SP 6uring tile 10 seconds precedin- t[iis point was determined by measuring the difference in potential between the ref2erence point and a second point 10 seconds earlier. I,ikewise the SP slope was also found between the reference point and tjic level of SP at a point" 10 seconds later. If tJ-;e change during any 10 second period was towards increasing pos-Ltivity the slope score was also assigned a positive sign, and conversely. A change score (,6-slope) was calculated from 2 the tnvo slopes by taking the absolute difference between them. Thus, each slope score represented the magnitude of the aloebraic change in slope bet-@,veen the first and last 10 seconds of the 20-second scoring period, independent of the direction of change. Results Comparison of ExT)erimental Versus Control Sites 2 It was predicted that either engorgement, occlusion or both would lead to alterations in SP levels at the experimental sites but not at control sites. This assumption was tested by a bet,.veen-S comparison of A -slope scores between experimental and control s ites . Analys es N@ere ac'compl [shed s eparateli for the 180 mm and 60 mm conditions and, within each of these, palmar and dorsal compar2- -ately. Irt each of these four conditions, the average Lsoi-L-s were also exam,ined sepa3. -slope score of the control site was computed for each S. These average n-slope scores for experimental and control sites across subjects were subsequently sub- jected to analysis by t-tests for correlated means. The results of the four bet-,veen9-.S t-tests are summarl-ed in Table 15. 0 Table 15. Change in slope of SP after release of pressure: between-S comparisons of experimental vs. control sites. Pressure Level Site 60 mm Hg 180 mm Hg 2 Dorsal 3.30* 5.74* Palniar 3.40* 5.25* *p < .01 The significant levels of these t analyses indicate that, for the group as a Nvhole, there were la-rger shifts at experimental sites than at control sites. This was true for both 2 palmar and dorsal regions and for both 60 and 180 mm Hg. An effort was also made to examine the intra-subject reliabilittl of this effect. This was accomplished, for each of the 13 Ss, by computing four t-rests for correlated means, based on the distribution of his t@, -slope scores across trials. Table 16 in- dicates how many of these 52 coi-nparisons were significant (2i.e., magnitude of SP change greater at experimental than at ntrol sites), and at what level of significance. Table 16. Change in slope of skin potential: within S comparisons of experimental vs. control sites. Values are in terms of the frequency of each level of sigiiifi- cance obtained by t-test. Signi2ficance Level Condition .01 05 .10 Nlons ig. 60 mm H -Dorsal 7 1 2 3 9 60 mm Hg - Palniar 5 I 1 6 180 ,nm Hg - Dorsal 10 0 1 2 180 mm Hg - Palmar 9 2 0 2 S5 Table 15. Change in slope of SP after release of pressure: bct%@,,ccn-S comparisons o.f experimental vs. control sites. Pressure Level Site 60 riim l@Ig 180 mm Hg Dorsa2l 3.30* 5.74* Palmar 3.40* 5.25* *p <.Ol The significant levels of these t analyses indirate that, fDr the group as a whole, there were larger shifts at experimental sites than at control sites. This was true for both palmar and dorsal regions and for both 602 and 180 mm Hg. An effort was also made to examine the intra-subject reliability of this effect. This @%,as accomplished, for each of the 13 Ss, by compu:ing four t-tests for correlated mee.ns, based on the distribution of his A -slope scores across trials. Table 16 in- dicates how many of these 52 comparisons were significant (i.e., magnitude of SP change greater at experimental 2than at control sites), and at what level of significance. Table 16. Change in slope of skin potential: within S comparisons of experimental vs. cont::ol sites. Values are in terfns of thifrequency of each level of sig-Liifi- cance obtained by t- test. Significance Level Condition .01 2.05 .10 Nonsig. 60 rr,,m Hg - Dorsal 7 1 2 3 60 mm Hg - Palrnar 5 1 1 6 180 mm Hg - Dorsal 10 0 1 2 180 mm Hg - PaLmar 9 2 0 5 2 86 As can be scei from this table, at least 8 of t!ic 13 Ss in cacti grol.,p, except 60 mm Hg-Palmar, demonstrated t values si(,-,iificant at the .05 level or better. In addition, the physiological recordings for certain comparisons which were not significant, clearly appeared to show the experimental,effect, but due to variability of the control site significance was not obtained. iolarity of SP Shift as a Function of Venous vs. Arterial Occlusion The first two analyses confirmed the existence of a shift in SP slope at a point approximately 5 seconds after release of pressure. The two sites on the experimental hand exhibited significantly greater SP fluctuations than control sites2 for both 180 and 60 mm Hg. . However, the pattern of change in SP seemed different for these two degrees of pressure. Typically, with the initiation of venous occlusion, SP became gradually more negative during occlusion and upon release reversed this trend and became more positive beginnin- about 5 seconds after release. 7bis cont'rast@d with the pattern of SP change induced by 2arterial occlusion. In this condition SP dipped sharply towards increasina- positivity about mid@vay through the in flation p2riod and then changed to a more negative level, again beginning at a point about 5 seconds followincr deflatio.,i. These distinct response patterns for 60'and 180 mm Hg may be illustrated simply by counting the number of Ss whose SP tended to s.ii2ft in a positive direction following the termii-lation of venous occlusion and in a negative direction upon the conclusion of arterial occlusion. Table 17 contains these figures, In three groups there was nearly complete congruence; all Ss but one shifted in parallel directions. Five Ss in the 6'ul mm Hg-Dorsal cell shifted in a negative direct,-on, but for only two of these Ss was this paradox,.c4al reversal -,t all proncunced. 87 Table 17. Fre-luency and averige niagiitudc of positive and nc-&,@ative SP sllifts after release of pressure (experfniont,,il sites o.,ily). Pos. Shift Neg. Shift Pos. X ineg. 60 mm ILiz 2 Dorsal 8 5 0. 3i niv Palmar 13 0 0. 17 mv 180 mm-!!g Dors'al 0 13 0.73 mv Palmar 1 12 0.56 mv The average magnitudes of the positive shifts occurring at the offset of 60 mm 2 and of the negative shift following the end of a 180 mm trial are listed In the last two .columns of Table 17. The five subjects whose mean che-nge was negative were not included in the 60 mm Dorsal mean; likewise, the one subject foi whom a positive sliift appeared after 180 trials and not added to the 180 mm Palmar group for this analysis. As these magnitude figures indicate, chan2ges of SP levels that occurred in relation to 180 mm Hg trials were more salient than 1-he less marked shifts assocfateii with the 60 mm Ho- conditio,-i. Summary It has been demonstrated that certain changes in surface potential are consistently brought about by manipulation of vas cular state. These changes are particularly marked 6 during the sudden return from the altered state and may iridlcate a local reflex involving compensatory changes in smooth muscle of the blood vessels. This exideriment does not demonstrate that s-,tc!i clianges are also produced by central ei'^ccs but such a likeliliood is rnade pliusiblc by dicsc findings. Wltitc sccon(!.Iry alterations, of sweat glands, could explain ttiese data, the most parslinonious explanation Is that they rcflccc local vasoniotor alterations.