- Email: [email protected]
Elicitation and habituation of the electrodermal orienting response in a short interstimulus interval paradigm
International Journal of Psychophysiology, 15 (1993) 241-253 0 1993 Elsevier Science Publishers B.V. All rights reserved
INTPSY
247 0167-8760/93/$06.00
484
Elicitation and habituation of the electrodermal orienting response in a short interstimulus interval paradigm Robert
J. Barry a,b,*, Sabine Feldmann a, Evian Gordon and Chris Rennie d
‘, Kathryn
I. Cocker
’
’ School of Education Studies, University of New South Wales, Kensington, NSW (Australia), b Department of Psychology, University of Wollongong, Wollongong, NS W (Australia), ’ Department of Cognitive Neuroscience, Westmead Hospital, Westmead, NSW (Australia) and d Department of Medical Physics, Westmead Hospital, Westmead, NSW (Australia) (Accepted
Key words: Orienting
response; Habituation; Skin conductance
19 July 1993)
Response response;
recovery; Dishabituation; NlOO; ERP paradigm
Electrodermal
activity:
The present experiment was carried out to investigate elicitation and habituation of the electrodermal Orienting Response with stimulus trains utilising a short interstimulus interval (ISI) of 1.1 s. We sought evidence for within-train response decrement to repeated stimulus presentation, response recovery to a change stimulus and dishabituation following the change stimulus - the three properties necessary to unequivocally identify a decremental process as habituation. No autonomic study could be found using such a short ISI. Autonomic studies on this time scale are necessary if these measures are to be integrated with central event-related potential (ERP) measures of electrical brain function. Overcoming this paradigm gap required the development of novel measurement procedures to estimate the small electrodermal responses obtained, usually occurring on the recovery slope of the response to the previous stimulus in the train. With our novel measurement procedures, evidence was found indicating that electrodermal activity in such a paradigm exhibited the three classic criteria of habituation.
INTRODUCTION The Orienting Response (OR) is the primary reaction of the body to a novel stimulus, and its elicitation may be considered as one of the most fundamental behavioral properties of living organisms (Sokolov, 1963). If a stimulus has no major consequences for the organism, repetition leads to extinction of the OR, i.e., there is response decrement with repeated stimulus presentation. If another stimulus is presented following extinction of the OR to the original stimulus,
* Corresponding author. Correspondence to: Department of Psychology, University of Wollongong, Wollongong 2522, Australia.
there is an increased response to the change stimulus (response recovery), and responses to subsequent presentations of the original stimulus are also increased (dishabituation) (Thompson et al., 1973). Response decrement, recovery and dishabituation are well-established concepts in the context of autonomic OR research, and are generally considered as robust phenomena. This pattern of decrement, response recovery and dishabituation has been examined in autonomic measures only with relatively long interstimulus intervals (ISIS), usually of several tens of seconds. As an example, Barry and James (1981) used visual stimuli with a duration of 2 s and ISIS of 40, 50 and 60 s. In contrast, with studies of central measures of electrical brain function such as the event-related
24x
potential (ERP), response decrement over trials commonly is examined using repeated trains of stimuli of short duration (generally less than 100 ms) presented with short ISIS (generally of the order of 1 s). These differences illustrate the paradigm gap which has developed between workers studying autonomic versus central measures of essentially similar perceptual and cognitive functions. An important focus for our group is the bridging of this paradigm gap, by ultimately using simultaneous recording of both peripheral and central measures, so that benefits can be gained by examining both sources of information and their interrelationships. At this stage of the development of integration of the autonomic and central nervous system measures of brain function, there is no obvious reason to prefer one family of paradigms over another. In The Netherlands, Verbaten’s group (e.g., Kenemans et al., 198X) has provided a number of useful examples of how single-trial estimates of the ERP can be obtained in an autonomic OR paradigm, but the lack of widespread implementation of their technique suggests that it has not been as universally accepted in the literature as might have been hoped. As a tentative step towards the development of a range of choices in integrative techniques, the present study focussed on an attempt at bridging in the opposite direction, from the traditional ERP paradigm towards the autonomic. In ERP habituation studies, EEG activity is usually averaged across trains to obtain ERPs, which are then tested for habituation within train. In a study by Barry et al. (1992), we reported that the NlOO component of the ERP showed response decrement over trials, and response recovery to a change stimulus, but failed to show any evidence of dishabituation in the response to the stimulus following the change stimulus. The question arises as to whether this failure to show one of the classic signs of habituation concerns variables associated with late ERP components, the short IS1 paradigm, or both. We sought information relevant to the resolution of the short IS1 problem in the present study. Thus, in order to examine autonomic habituation in a short IS1 paradigm typical of ERP studies, we used an
A
\\, “I\, B
C
Fig. 1. Panel A illustrates the standard baseline-to-peak measurement of the SCR in the context of a stable baseline. Panel B shows a response formed by adding the curve in A to a linearly-decreasing level, and illustrates a situation in which the SCR occurs in the context of a rapidly falling baseline. With the application of the standard procedure there is no scored response. Panel C shows the procedure adopted here: the falling baseline is extended linearly below the response, and the maximum from the extended baseline to the observed response is scored as the SCR. Note that the full SCR shown in panel A has been recovered by this procedure.
analogous procedure with electrodermal activity, i.e., averaging activity across trains and examining within-train habituation. Electrodermal activity is an often used autonomic OR variable, and the changes in skin conductance response (SCR) over trials in the common types of autonomic paradigms are very reliable. In long IS1 paradigms the individual response can be seen to commence in an interval 1 to 3 s after the stimulus is presented (Barry, 1990). Traditionally, the amplitude of the response is measured by calculating the difference between a relatively stable baseline and the response peak (see Fig. 1A). As might be expected from the paucity of data in the literature, a number of methodological problems appear with measurement of the SCR in ERP-type stimulus sequences. One of these is the relatively small size of the electrodermal response to be expected from the short duration of the stimulus (here 40 ms rather than a typical autonomic-OR duration
249
of 1 or 2 s). This is compounded by the cumulative effect of reductions in response recovery associated with the very short intervals between stimulus presentations. Further, normal habituation over stimulus repetitions can be expected to reduce the response amplitudes contributing to the average - in ERP studies, something in the range of hundreds of stimuli might be used, compared with the lo-20 usual in autonomic studies. Although we hoped that our averaging procedure would help overcome this problem by increasing the signal/noise ratio, the expectation of very small responses remains a problem. In addition, because electrodermal activity produced by one stimulus may be returning towards baseline when the response to the next stimulus begins, there is no expectation of a stable baseline from which to measure the response peak. The combined effect of relatively small responses and a relativelyrapidly falling baseline is illustrated in Fig. 1B. A traditional baseline-to-peak measurement procedure would lead to the scoring of such a response as zero. In the normal type of autonomic study, the omission of such a small response might not introduce significant error, but in the present context, such an omission could be expected to seriously distort the response picture. Accordingly, in this study each reaction was measured by linearly extending the recovery slope of one response (or the shifting baseline trend) to a point below the peak of the following response, and taking the difference between this extended baseline and the peak as the response amplitude (Fig. 1C). This combined procedure, of response averaging and measuring response amplitudes from a projected baseline, has not previously been reported in the literature. An additional problem concerns the onset latency of the electrodermal response. As mentioned above, the usual latency range used to define a stimulus-evoked response is from 1 to 3 s after stimulus onset. With an IS1 of 1.1 s, it is readily apparent that the responses to each of two consecutive stimuli can fall within the traditional latency range used to define the response to either stimulus. Thus, response identification was an additional problem addressed in the study.
Hence, we sought evidence of within-train response decrement of the averaged evoked electrodermal response occurring in a typical ERP paradigm. Our paradigm consisted of 15 trains, each of 10 stimuli, and we averaged responses between trains. In Barry et al. (1992) the EEG epoch containing the first stimulus in train 1 was added to that containing the first stimulus in train 2, and so on up to train 15, in order to obtain the averaged response to the first stimulus in the train. This was repeated for each stimulus position in the train. The same procedure was carried out here using electrodermal activity rather than EEG data. We also sought evidence of response recovery to a change stimulus in the train, and dishabituation of the response to the following stimulus. Such an investigation of an autonomic variable in an ERP paradigm has not previously been reported. In this way we hoped to help bridge the gap between central and peripheral investigations of brain functioning in the OR context.
METHODS Subjects 1.5 normal subjects (nine females and six males), aged between 19 and 35 years, volunteered to participate in a study in the Department of Psychiatry at Westmead Hospital. Apparatus and procedure In each of two conditions, the subjects received 15 trains of auditory stimuli, each of 40 ms duration, 12-ms rise and fall times, and 60 dB above threshold. Each train consisted of 10 stimuli with an IS1 of 1.1 s. There were 9 standard tones of 500 Hz in places 1 to 7, 9 and 10, and one change tone in place 8 of each train. In trains 1 to 13, and 15, the change tone was at a frequency of 1000 Hz; in train 14 tone 8 was 2000 Hz, but the effects of this difference are not examined here. The intertrain interval (ITI) was 5 s. The tones were produced by a software-controlled sine-wave generator and delivered binaurally by circumaural headphones.
250
Electrodermal activity was recorded from Ag/ AgCl electrodes on the volar surfaces of the medial phalanges of the second and third finders of the right hand. Contact area was limited to 0.8 cm* by double-sided adhesive collars, and 0.05 M NaCl in an inert viscous ointment base was used as electrolyte. A Contact Precision Instruments SC4/A skin conductance device, controlled by an Apple microcomputer, impressed a constant voltage of 0.6 V across the electrodes. Skin conductance was sampled at 20 Hz for 15-s intervals commencing 0.8 s before the first stimulus of each train, and displayed in real time on the Apple monitor. The onset of sampling was triggered by a PDPll microcomputer controlling stimulus presentation. The period between sampling epochs was used to automatically control the back-off level of the SC4/A device. The data were recorded on floppy disk for off-line analysis. Subjects sat in a comfortable chair, with the video screen in front of them. In the same room, but out of the view of the subject, were the experimenter and the equipment. The room was dimly lit. The subject was instructed to remain alert and to watch the screen where a word anagram was presented. In the first condition (IGNORE) the subject was told that his/ her task was to form as many words as possible from the word anagram, and the tones were not mentioned. In the second condition (COUNT) the subject was instructed to count the deviant tones in the sequence, while maintaining his/ her gaze on the screen to reduce eye-movements. Subjects were only included in the study if their reported total was 15 i 1.
RESULTS The averaged evoked skin conductance response (SCR) was obtained across all 15 trains for each condition and subject separately. The data set was slightly smoothed using a ‘running average’ over three consecutive data points. The resulting data were plotted, as shown in Fig. 2 for a representative subject, and the evoked SCRs for the 10 auditory stimuli were measured by hand for this exploratory study.
, ,8 SKIN CONDUCTANCE *
T
i
2
3
4
4.16
(US)
u 5
6
7
a
9
IO
L,
4.12
-
-
4.11 12
3
4
6
6
7
6
9
TIME FROM TRAIN ONSET
la
11
12
13
(8)
Fig. 2. An illustrative averaged response obtained from a single subject in one condition. Note the relatively small responses to individual stimuli, and the falling baseline of the bulk of the record. The vertical lines numbered 1 to IO correspond with the response onsets predicted by taking that to the first stimulus as a model (1) and seeking subsequent responses at time-locked intervals corresponding to the ISI. The response onsets nearest these vertical lines were used to identify the SCRs to the sequential stimuli in the train.
The response to the first stimulus in the train, occurring with an onset latency between 1 and 3 s after the train onset, was taken as the model SCR for the subject in that condition, and its onset latency was marked (see vertical line labelled 1 in Fig. 2). Since we had time-locked stimulus onsets at 1.14 s intervals, we sought time-locked response onsets. Thus further vertical lines were drawn at 1.14-s intervals following the initial response to train onset (see lines labelled 2 to 10 in Fig. 2). The SCR with onset closest to each of these lines was taken as the response evoked by the corresponding sequential stimulus in the train. Having thus identified a response to each stimulus in the train, we then obtained a baseline for response amplitude estimation by extending the recovery slope or baseline drift prior to the onset of that response to a point below the response peak. The deviation from this point to the response maximum was measured in nS. SCRs for each condition were obtained in this way for each subject. These data were then square-rooted to reduce the skew associated with small responses.
251
SQRT SCR (JnS)
3.3 -I-
12
3
4
3
6
7
6
9
la
STIMULUS POSITION Fig. 3. Square-root SCRs as a function of stimulus position in the train, shown in each condition. Stimulus 8 is the change stimulus.
Mean values for each stimulus in the train, in each condition, are shown in Fig. 3. Examination of Fig. 3 suggests that, in the first condition, there was rapid response decrement over the first seven stimuli in the train, followed by response recovery to the change tone 8. The response to the subsequent standard tone 9 appears somewhat larger than that to the preceding tone 7, suggesting dishabituation. In the second condition, the response to tone 1 is reduced from that in the first condition, suggesting that reduction due to stimulus repetition was larger than the enhancement expected due to the task requirement. Similar changes across the train occurred as noted in the first condition. However, the response to the target change tone in the second condition appeared larger than that to tone 1, an apparent enhancement not observed in the first condition. These data were analysed in separate repeated-measures ANOVAs, designed to evaluate the classic signs of habituation outlined above. The first analysis, to investigate response decrement with stimulus repetion, was a two-way ANOVA over position (1 to 7) and condition (IGNORE/ COUNT). Within position, linear and quadratic trends were used to examine habituation. Both linear and quadratic trends over these stimulus positions were significant (F(1,14) =
12.69, p < 0.01 and F(1,14) = 20.63, p < 0.001, respectively). The difference between the two conditions failed to reach significance (F(1,14) = 2.23, p = 0.16). The interaction of condition by linear trend over repeated tones approached significance (F(1,14) = 3.69, p = 0.08). The second analysis was carried out over the responses from tones 7 and 8, to evaluate if response recovery occurred at the change tone. There was a significant increase in the response to the change tone (F(1,14) = 12.22, p < 0.01). Again, no significant difference between conditions was found (F < 11, and there was no interaction between position and condition (F < 1). A similar analysis was used to compare responses to tones 7 and 9 to examine evidence for dishabituation. The effect of position almost reached significance (F(1,14) = 4.55, p = 0.051), but again there was no suggestion of condition or condition X position effects (F < 1 in each case>. Because the analysis of positions 7 and 9 failed to yield unequivocal evidence of dishabituation, the data were transformed to z-scores within subject and reanalysed. This transformation equates all subjects in their contribution to the pooled data set, and hence increases the sensitivity of the data analysis to genuine within-subject effects. Means of these z-scores over positions in the stimulus trains are displayed in Fig. 4. Over the pre-change stimulus positions 1 to 7, the linear and quadratic trends were strengthened (F(1,14) = 17.33, p < 0.001 and F(1,14) = 37.36, p < 0.001, respectively). There was no effect of condition (F(1,14) = 2.43, p = 0.141, and the interaction between linear trend over stimulus positions and condition again only approached significance (F(1,14) = 3.51, p = 0.08). The analysis over the responses to stimuli 7 and 8, carried out to investigate response recovery at the change, again found no effect of condition (F < 1). There was a strengthened effect of position (F(1,14) = 45.74, p < O.OOl>, but no interaction between position and condition (F < 1). The analysis over stimulus positions 7 and 9, investigating dishabituation, found a significant position effect (F(1,14) = 7.17, p < 0.021, but no effect of condition (F < 11, or interaction between condition and position (&X1,14) = 1.59, p = 0.23). That
252
Z-SCORES
1.6 T
1
2
3
4
STIMULUS
s
6
7
3
9
10
POSITION
Fig. 4. The effects of stimulus position in the transformed data. Data were transformed to z-scores within-subject to reduce between-subject variability and increase the sensitivity of the analyses.
is, this set of analyses over z-scores provided clear evidence of response decrement with stimulus repetition, response recovery at a change stimulus, and dishabituation of the response to the following stimulus. These data thus indicate that habituation occurs in the electrodermal responses elicited in this paradigm.
DISCUSSION In this experiment we investigated an electrodermal analog of a standard ERP study, and focussed on within-train habituation of the OR. Thus, we looked at changes in the electrodermal data averaged across trains, time-locked to the onset of individual stimuli in the train. The interstimulus interval used in this experiment (1.1 s> was extremely short when considered from the perspective of workers in the field of autonomic ORs. This often resulted in more than one response occurring in the traditional latency range associated with each stimulus in the train, and the response to one stimulus falling on the recovery slope of the response to the previous stimulus. After identification of responses by timelocking response onsets (using the response to the first stimulus in the train as a model re-
sponse), a satisfactory method was found to measure these responses by extending the recovery slope/ baseline to a point under the peak of the response. This method was able to be used with each subject, even those with low levels of responding. If our two conditions (IGNORE VS. COUNT) had been counterbalanced, we would have expected enhanced response amplitudes (particularly to the target tone) in the COUNT condition. However, in our paradigm, designed for use in evaluating perceptual/cognitive processing in individual patients, these two conditions were not counterbalanced. Thus the attentional enhancement was expected to be negated by the reduced novelty associated with the COUNT condition (stimulus trains 16 to 30) being presented after the IGNORE condition (stimulus trains 1 to 15). This was indeed the case, with littlc evidence of any conditions effect. Nevertheless, this use of a two-condition paradigm is useful in providing additional data for the repeated-measures analyses without the problems of boredom, etc.. which might arise from a long single-condition paradigm. Evidence of response decrement with stimulus repetition, response recovery and dishabituation was obtained over both conditions. Response decrement was clearly shown in the linear and quadratic trends over tones 1 to 7 in both conditions. There was some difference between linear trends over these tones in the two conditions, suggestive of slower relative decrement in the COUNT condition, but this only approached significance. Response recovery at the change tone was substantial in both conditions. Subsequently, dishabituation was apparent in enhanced responding to the standard tone 9 (after the change tone) compared with the preceding standard tone 7. Again, this did not differ between conditions. Thus, we can conclude that, in this study, the electrodermal OR showed the three classic signs defining habituation: response decrement with repeated stimulus presentation, response recovery to a change stimulus, and dishabituation to subsequent presentations of the standard stimulus. This preliminary finding indicates that meaningful autonomic measures can be obtained in short IS1 paradigms.
253
In the same paradigm, we previously found that the NlOO component of the ERP failed to demonstrate dishabituation following the change stimulus, and hence failed to meet all the classic criteria for habituation (Barry et al., 1992). In the light of this failure, the response decrement process evidenced by NlOO cannot be identified as habituation. The present demonstration indicates that such a failure is not merely a correlate of the short IS1 ERP paradigm, since the electrodermal OR demonstrates dishabituation. Rather, it suggests that such a failure to demonstrate habituation is a reflection of the intrinsic nature of the NlOO. If this is the case, the NlOO as elicited in this short ISI paradigm cannot be considered to be an index of the OR, since habituation is a fundamental characteristic of that system. There remains, however, the possibility that some subtle aspect of this paradigm, perhaps reflecting an interaction with specific temporal parameters of the NlOO mechanism, may explain the present discrepancy between the SCR and NlOO. Such a possibility is currently under investigation in our laboratory. This study provides a clear demonstration that the paradigm gap, currently separating workers using autonomic and central measures to explore similar aspects of cognitive neuroscience, may be
bridged by the examination of autonomic data collected in an ERP paradigm. We hope that our novel measurement techniques, used here in an may serve as directionexploratory fashion, markers for future integrative efforts.
REFERENCES Barry, R.J. and James, A.L. (1981) Fractionation of phasic responses in a dishabituation paradigm. Physiol. Behal,., 26: 69-75. Barry, R.J. (1990) Scoring criteria for response latency and habituation in electrodermal research: A study in the context of the Orienting Response. Psychophysiology, 27: 94-100. Barry, R.J., Cocker, K.I., Anderson, J.W., Gordon, E. and Rennie, C. (1992) Does the NlOO evoked potential really habituate? Evidence from a paradigm appropriate to a clinical setting. ht. J. Psychophysiol., 13: 9-16. Kenemans, J.L., Verbaten, M.N., Sjouw, W. and Slangen, J.L. (1988) Effects of task relevance on habituation of visual single-trial ERPs and the skin conductance orienting response. ht. .I. Psychophysiol., 6: 51-63. Sokolov, E.N. (1963) Perception and the conditioned reyex, Pergamon Press, Oxford. Thompson, R.F., Groves, P.M., Teyler, T.J. and Roemer, R.A. (1973) Dual-process theory of habituation: Theory and Behavior. In H.V.S. Peeke and M.J. Herz (Eds.) Habituation: Behacioral studies and physiological substrates, Vol. 1, Academic Press, New York.
INTPSY
247 0167-8760/93/$06.00
484
Elicitation and habituation of the electrodermal orienting response in a short interstimulus interval paradigm Robert
J. Barry a,b,*, Sabine Feldmann a, Evian Gordon and Chris Rennie d
‘, Kathryn
I. Cocker
’
’ School of Education Studies, University of New South Wales, Kensington, NSW (Australia), b Department of Psychology, University of Wollongong, Wollongong, NS W (Australia), ’ Department of Cognitive Neuroscience, Westmead Hospital, Westmead, NSW (Australia) and d Department of Medical Physics, Westmead Hospital, Westmead, NSW (Australia) (Accepted
Key words: Orienting
response; Habituation; Skin conductance
19 July 1993)
Response response;
recovery; Dishabituation; NlOO; ERP paradigm
Electrodermal
activity:
The present experiment was carried out to investigate elicitation and habituation of the electrodermal Orienting Response with stimulus trains utilising a short interstimulus interval (ISI) of 1.1 s. We sought evidence for within-train response decrement to repeated stimulus presentation, response recovery to a change stimulus and dishabituation following the change stimulus - the three properties necessary to unequivocally identify a decremental process as habituation. No autonomic study could be found using such a short ISI. Autonomic studies on this time scale are necessary if these measures are to be integrated with central event-related potential (ERP) measures of electrical brain function. Overcoming this paradigm gap required the development of novel measurement procedures to estimate the small electrodermal responses obtained, usually occurring on the recovery slope of the response to the previous stimulus in the train. With our novel measurement procedures, evidence was found indicating that electrodermal activity in such a paradigm exhibited the three classic criteria of habituation.
INTRODUCTION The Orienting Response (OR) is the primary reaction of the body to a novel stimulus, and its elicitation may be considered as one of the most fundamental behavioral properties of living organisms (Sokolov, 1963). If a stimulus has no major consequences for the organism, repetition leads to extinction of the OR, i.e., there is response decrement with repeated stimulus presentation. If another stimulus is presented following extinction of the OR to the original stimulus,
* Corresponding author. Correspondence to: Department of Psychology, University of Wollongong, Wollongong 2522, Australia.
there is an increased response to the change stimulus (response recovery), and responses to subsequent presentations of the original stimulus are also increased (dishabituation) (Thompson et al., 1973). Response decrement, recovery and dishabituation are well-established concepts in the context of autonomic OR research, and are generally considered as robust phenomena. This pattern of decrement, response recovery and dishabituation has been examined in autonomic measures only with relatively long interstimulus intervals (ISIS), usually of several tens of seconds. As an example, Barry and James (1981) used visual stimuli with a duration of 2 s and ISIS of 40, 50 and 60 s. In contrast, with studies of central measures of electrical brain function such as the event-related
24x
potential (ERP), response decrement over trials commonly is examined using repeated trains of stimuli of short duration (generally less than 100 ms) presented with short ISIS (generally of the order of 1 s). These differences illustrate the paradigm gap which has developed between workers studying autonomic versus central measures of essentially similar perceptual and cognitive functions. An important focus for our group is the bridging of this paradigm gap, by ultimately using simultaneous recording of both peripheral and central measures, so that benefits can be gained by examining both sources of information and their interrelationships. At this stage of the development of integration of the autonomic and central nervous system measures of brain function, there is no obvious reason to prefer one family of paradigms over another. In The Netherlands, Verbaten’s group (e.g., Kenemans et al., 198X) has provided a number of useful examples of how single-trial estimates of the ERP can be obtained in an autonomic OR paradigm, but the lack of widespread implementation of their technique suggests that it has not been as universally accepted in the literature as might have been hoped. As a tentative step towards the development of a range of choices in integrative techniques, the present study focussed on an attempt at bridging in the opposite direction, from the traditional ERP paradigm towards the autonomic. In ERP habituation studies, EEG activity is usually averaged across trains to obtain ERPs, which are then tested for habituation within train. In a study by Barry et al. (1992), we reported that the NlOO component of the ERP showed response decrement over trials, and response recovery to a change stimulus, but failed to show any evidence of dishabituation in the response to the stimulus following the change stimulus. The question arises as to whether this failure to show one of the classic signs of habituation concerns variables associated with late ERP components, the short IS1 paradigm, or both. We sought information relevant to the resolution of the short IS1 problem in the present study. Thus, in order to examine autonomic habituation in a short IS1 paradigm typical of ERP studies, we used an
A
\\, “I\, B
C
Fig. 1. Panel A illustrates the standard baseline-to-peak measurement of the SCR in the context of a stable baseline. Panel B shows a response formed by adding the curve in A to a linearly-decreasing level, and illustrates a situation in which the SCR occurs in the context of a rapidly falling baseline. With the application of the standard procedure there is no scored response. Panel C shows the procedure adopted here: the falling baseline is extended linearly below the response, and the maximum from the extended baseline to the observed response is scored as the SCR. Note that the full SCR shown in panel A has been recovered by this procedure.
analogous procedure with electrodermal activity, i.e., averaging activity across trains and examining within-train habituation. Electrodermal activity is an often used autonomic OR variable, and the changes in skin conductance response (SCR) over trials in the common types of autonomic paradigms are very reliable. In long IS1 paradigms the individual response can be seen to commence in an interval 1 to 3 s after the stimulus is presented (Barry, 1990). Traditionally, the amplitude of the response is measured by calculating the difference between a relatively stable baseline and the response peak (see Fig. 1A). As might be expected from the paucity of data in the literature, a number of methodological problems appear with measurement of the SCR in ERP-type stimulus sequences. One of these is the relatively small size of the electrodermal response to be expected from the short duration of the stimulus (here 40 ms rather than a typical autonomic-OR duration
249
of 1 or 2 s). This is compounded by the cumulative effect of reductions in response recovery associated with the very short intervals between stimulus presentations. Further, normal habituation over stimulus repetitions can be expected to reduce the response amplitudes contributing to the average - in ERP studies, something in the range of hundreds of stimuli might be used, compared with the lo-20 usual in autonomic studies. Although we hoped that our averaging procedure would help overcome this problem by increasing the signal/noise ratio, the expectation of very small responses remains a problem. In addition, because electrodermal activity produced by one stimulus may be returning towards baseline when the response to the next stimulus begins, there is no expectation of a stable baseline from which to measure the response peak. The combined effect of relatively small responses and a relativelyrapidly falling baseline is illustrated in Fig. 1B. A traditional baseline-to-peak measurement procedure would lead to the scoring of such a response as zero. In the normal type of autonomic study, the omission of such a small response might not introduce significant error, but in the present context, such an omission could be expected to seriously distort the response picture. Accordingly, in this study each reaction was measured by linearly extending the recovery slope of one response (or the shifting baseline trend) to a point below the peak of the following response, and taking the difference between this extended baseline and the peak as the response amplitude (Fig. 1C). This combined procedure, of response averaging and measuring response amplitudes from a projected baseline, has not previously been reported in the literature. An additional problem concerns the onset latency of the electrodermal response. As mentioned above, the usual latency range used to define a stimulus-evoked response is from 1 to 3 s after stimulus onset. With an IS1 of 1.1 s, it is readily apparent that the responses to each of two consecutive stimuli can fall within the traditional latency range used to define the response to either stimulus. Thus, response identification was an additional problem addressed in the study.
Hence, we sought evidence of within-train response decrement of the averaged evoked electrodermal response occurring in a typical ERP paradigm. Our paradigm consisted of 15 trains, each of 10 stimuli, and we averaged responses between trains. In Barry et al. (1992) the EEG epoch containing the first stimulus in train 1 was added to that containing the first stimulus in train 2, and so on up to train 15, in order to obtain the averaged response to the first stimulus in the train. This was repeated for each stimulus position in the train. The same procedure was carried out here using electrodermal activity rather than EEG data. We also sought evidence of response recovery to a change stimulus in the train, and dishabituation of the response to the following stimulus. Such an investigation of an autonomic variable in an ERP paradigm has not previously been reported. In this way we hoped to help bridge the gap between central and peripheral investigations of brain functioning in the OR context.
METHODS Subjects 1.5 normal subjects (nine females and six males), aged between 19 and 35 years, volunteered to participate in a study in the Department of Psychiatry at Westmead Hospital. Apparatus and procedure In each of two conditions, the subjects received 15 trains of auditory stimuli, each of 40 ms duration, 12-ms rise and fall times, and 60 dB above threshold. Each train consisted of 10 stimuli with an IS1 of 1.1 s. There were 9 standard tones of 500 Hz in places 1 to 7, 9 and 10, and one change tone in place 8 of each train. In trains 1 to 13, and 15, the change tone was at a frequency of 1000 Hz; in train 14 tone 8 was 2000 Hz, but the effects of this difference are not examined here. The intertrain interval (ITI) was 5 s. The tones were produced by a software-controlled sine-wave generator and delivered binaurally by circumaural headphones.
250
Electrodermal activity was recorded from Ag/ AgCl electrodes on the volar surfaces of the medial phalanges of the second and third finders of the right hand. Contact area was limited to 0.8 cm* by double-sided adhesive collars, and 0.05 M NaCl in an inert viscous ointment base was used as electrolyte. A Contact Precision Instruments SC4/A skin conductance device, controlled by an Apple microcomputer, impressed a constant voltage of 0.6 V across the electrodes. Skin conductance was sampled at 20 Hz for 15-s intervals commencing 0.8 s before the first stimulus of each train, and displayed in real time on the Apple monitor. The onset of sampling was triggered by a PDPll microcomputer controlling stimulus presentation. The period between sampling epochs was used to automatically control the back-off level of the SC4/A device. The data were recorded on floppy disk for off-line analysis. Subjects sat in a comfortable chair, with the video screen in front of them. In the same room, but out of the view of the subject, were the experimenter and the equipment. The room was dimly lit. The subject was instructed to remain alert and to watch the screen where a word anagram was presented. In the first condition (IGNORE) the subject was told that his/ her task was to form as many words as possible from the word anagram, and the tones were not mentioned. In the second condition (COUNT) the subject was instructed to count the deviant tones in the sequence, while maintaining his/ her gaze on the screen to reduce eye-movements. Subjects were only included in the study if their reported total was 15 i 1.
RESULTS The averaged evoked skin conductance response (SCR) was obtained across all 15 trains for each condition and subject separately. The data set was slightly smoothed using a ‘running average’ over three consecutive data points. The resulting data were plotted, as shown in Fig. 2 for a representative subject, and the evoked SCRs for the 10 auditory stimuli were measured by hand for this exploratory study.
, ,8 SKIN CONDUCTANCE *
T
i
2
3
4
4.16
(US)
u 5
6
7
a
9
IO
L,
4.12
-
-
4.11 12
3
4
6
6
7
6
9
TIME FROM TRAIN ONSET
la
11
12
13
(8)
Fig. 2. An illustrative averaged response obtained from a single subject in one condition. Note the relatively small responses to individual stimuli, and the falling baseline of the bulk of the record. The vertical lines numbered 1 to IO correspond with the response onsets predicted by taking that to the first stimulus as a model (1) and seeking subsequent responses at time-locked intervals corresponding to the ISI. The response onsets nearest these vertical lines were used to identify the SCRs to the sequential stimuli in the train.
The response to the first stimulus in the train, occurring with an onset latency between 1 and 3 s after the train onset, was taken as the model SCR for the subject in that condition, and its onset latency was marked (see vertical line labelled 1 in Fig. 2). Since we had time-locked stimulus onsets at 1.14 s intervals, we sought time-locked response onsets. Thus further vertical lines were drawn at 1.14-s intervals following the initial response to train onset (see lines labelled 2 to 10 in Fig. 2). The SCR with onset closest to each of these lines was taken as the response evoked by the corresponding sequential stimulus in the train. Having thus identified a response to each stimulus in the train, we then obtained a baseline for response amplitude estimation by extending the recovery slope or baseline drift prior to the onset of that response to a point below the response peak. The deviation from this point to the response maximum was measured in nS. SCRs for each condition were obtained in this way for each subject. These data were then square-rooted to reduce the skew associated with small responses.
251
SQRT SCR (JnS)
3.3 -I-
12
3
4
3
6
7
6
9
la
STIMULUS POSITION Fig. 3. Square-root SCRs as a function of stimulus position in the train, shown in each condition. Stimulus 8 is the change stimulus.
Mean values for each stimulus in the train, in each condition, are shown in Fig. 3. Examination of Fig. 3 suggests that, in the first condition, there was rapid response decrement over the first seven stimuli in the train, followed by response recovery to the change tone 8. The response to the subsequent standard tone 9 appears somewhat larger than that to the preceding tone 7, suggesting dishabituation. In the second condition, the response to tone 1 is reduced from that in the first condition, suggesting that reduction due to stimulus repetition was larger than the enhancement expected due to the task requirement. Similar changes across the train occurred as noted in the first condition. However, the response to the target change tone in the second condition appeared larger than that to tone 1, an apparent enhancement not observed in the first condition. These data were analysed in separate repeated-measures ANOVAs, designed to evaluate the classic signs of habituation outlined above. The first analysis, to investigate response decrement with stimulus repetion, was a two-way ANOVA over position (1 to 7) and condition (IGNORE/ COUNT). Within position, linear and quadratic trends were used to examine habituation. Both linear and quadratic trends over these stimulus positions were significant (F(1,14) =
12.69, p < 0.01 and F(1,14) = 20.63, p < 0.001, respectively). The difference between the two conditions failed to reach significance (F(1,14) = 2.23, p = 0.16). The interaction of condition by linear trend over repeated tones approached significance (F(1,14) = 3.69, p = 0.08). The second analysis was carried out over the responses from tones 7 and 8, to evaluate if response recovery occurred at the change tone. There was a significant increase in the response to the change tone (F(1,14) = 12.22, p < 0.01). Again, no significant difference between conditions was found (F < 11, and there was no interaction between position and condition (F < 1). A similar analysis was used to compare responses to tones 7 and 9 to examine evidence for dishabituation. The effect of position almost reached significance (F(1,14) = 4.55, p = 0.051), but again there was no suggestion of condition or condition X position effects (F < 1 in each case>. Because the analysis of positions 7 and 9 failed to yield unequivocal evidence of dishabituation, the data were transformed to z-scores within subject and reanalysed. This transformation equates all subjects in their contribution to the pooled data set, and hence increases the sensitivity of the data analysis to genuine within-subject effects. Means of these z-scores over positions in the stimulus trains are displayed in Fig. 4. Over the pre-change stimulus positions 1 to 7, the linear and quadratic trends were strengthened (F(1,14) = 17.33, p < 0.001 and F(1,14) = 37.36, p < 0.001, respectively). There was no effect of condition (F(1,14) = 2.43, p = 0.141, and the interaction between linear trend over stimulus positions and condition again only approached significance (F(1,14) = 3.51, p = 0.08). The analysis over the responses to stimuli 7 and 8, carried out to investigate response recovery at the change, again found no effect of condition (F < 1). There was a strengthened effect of position (F(1,14) = 45.74, p < O.OOl>, but no interaction between position and condition (F < 1). The analysis over stimulus positions 7 and 9, investigating dishabituation, found a significant position effect (F(1,14) = 7.17, p < 0.021, but no effect of condition (F < 11, or interaction between condition and position (&X1,14) = 1.59, p = 0.23). That
252
Z-SCORES
1.6 T
1
2
3
4
STIMULUS
s
6
7
3
9
10
POSITION
Fig. 4. The effects of stimulus position in the transformed data. Data were transformed to z-scores within-subject to reduce between-subject variability and increase the sensitivity of the analyses.
is, this set of analyses over z-scores provided clear evidence of response decrement with stimulus repetition, response recovery at a change stimulus, and dishabituation of the response to the following stimulus. These data thus indicate that habituation occurs in the electrodermal responses elicited in this paradigm.
DISCUSSION In this experiment we investigated an electrodermal analog of a standard ERP study, and focussed on within-train habituation of the OR. Thus, we looked at changes in the electrodermal data averaged across trains, time-locked to the onset of individual stimuli in the train. The interstimulus interval used in this experiment (1.1 s> was extremely short when considered from the perspective of workers in the field of autonomic ORs. This often resulted in more than one response occurring in the traditional latency range associated with each stimulus in the train, and the response to one stimulus falling on the recovery slope of the response to the previous stimulus. After identification of responses by timelocking response onsets (using the response to the first stimulus in the train as a model re-
sponse), a satisfactory method was found to measure these responses by extending the recovery slope/ baseline to a point under the peak of the response. This method was able to be used with each subject, even those with low levels of responding. If our two conditions (IGNORE VS. COUNT) had been counterbalanced, we would have expected enhanced response amplitudes (particularly to the target tone) in the COUNT condition. However, in our paradigm, designed for use in evaluating perceptual/cognitive processing in individual patients, these two conditions were not counterbalanced. Thus the attentional enhancement was expected to be negated by the reduced novelty associated with the COUNT condition (stimulus trains 16 to 30) being presented after the IGNORE condition (stimulus trains 1 to 15). This was indeed the case, with littlc evidence of any conditions effect. Nevertheless, this use of a two-condition paradigm is useful in providing additional data for the repeated-measures analyses without the problems of boredom, etc.. which might arise from a long single-condition paradigm. Evidence of response decrement with stimulus repetition, response recovery and dishabituation was obtained over both conditions. Response decrement was clearly shown in the linear and quadratic trends over tones 1 to 7 in both conditions. There was some difference between linear trends over these tones in the two conditions, suggestive of slower relative decrement in the COUNT condition, but this only approached significance. Response recovery at the change tone was substantial in both conditions. Subsequently, dishabituation was apparent in enhanced responding to the standard tone 9 (after the change tone) compared with the preceding standard tone 7. Again, this did not differ between conditions. Thus, we can conclude that, in this study, the electrodermal OR showed the three classic signs defining habituation: response decrement with repeated stimulus presentation, response recovery to a change stimulus, and dishabituation to subsequent presentations of the standard stimulus. This preliminary finding indicates that meaningful autonomic measures can be obtained in short IS1 paradigms.
253
In the same paradigm, we previously found that the NlOO component of the ERP failed to demonstrate dishabituation following the change stimulus, and hence failed to meet all the classic criteria for habituation (Barry et al., 1992). In the light of this failure, the response decrement process evidenced by NlOO cannot be identified as habituation. The present demonstration indicates that such a failure is not merely a correlate of the short IS1 ERP paradigm, since the electrodermal OR demonstrates dishabituation. Rather, it suggests that such a failure to demonstrate habituation is a reflection of the intrinsic nature of the NlOO. If this is the case, the NlOO as elicited in this short ISI paradigm cannot be considered to be an index of the OR, since habituation is a fundamental characteristic of that system. There remains, however, the possibility that some subtle aspect of this paradigm, perhaps reflecting an interaction with specific temporal parameters of the NlOO mechanism, may explain the present discrepancy between the SCR and NlOO. Such a possibility is currently under investigation in our laboratory. This study provides a clear demonstration that the paradigm gap, currently separating workers using autonomic and central measures to explore similar aspects of cognitive neuroscience, may be
bridged by the examination of autonomic data collected in an ERP paradigm. We hope that our novel measurement techniques, used here in an may serve as directionexploratory fashion, markers for future integrative efforts.
REFERENCES Barry, R.J. and James, A.L. (1981) Fractionation of phasic responses in a dishabituation paradigm. Physiol. Behal,., 26: 69-75. Barry, R.J. (1990) Scoring criteria for response latency and habituation in electrodermal research: A study in the context of the Orienting Response. Psychophysiology, 27: 94-100. Barry, R.J., Cocker, K.I., Anderson, J.W., Gordon, E. and Rennie, C. (1992) Does the NlOO evoked potential really habituate? Evidence from a paradigm appropriate to a clinical setting. ht. J. Psychophysiol., 13: 9-16. Kenemans, J.L., Verbaten, M.N., Sjouw, W. and Slangen, J.L. (1988) Effects of task relevance on habituation of visual single-trial ERPs and the skin conductance orienting response. ht. .I. Psychophysiol., 6: 51-63. Sokolov, E.N. (1963) Perception and the conditioned reyex, Pergamon Press, Oxford. Thompson, R.F., Groves, P.M., Teyler, T.J. and Roemer, R.A. (1973) Dual-process theory of habituation: Theory and Behavior. In H.V.S. Peeke and M.J. Herz (Eds.) Habituation: Behacioral studies and physiological substrates, Vol. 1, Academic Press, New York.