Biological Psychology, 3,1975,
CLASSICAL
101-112.
CONDITIONING
IN ALERT AND DROWSY DAVID G. MCDONALD*
@ North-Holland Publishing Company
OF AUTONOMIC
RESPONSES
SUBJECTS
and LAVERNE
C. JOHNSON
U.S. Naly Medical Neuropsychiatric Research Unit, San Diego, California, U.S.A.
Accepted for publication 15 April 1975
The primary purpose of this experiment was to study classically conditioned autonomic responses of alert versus drowsy subjects. Sixty-nine subjects in conditioning and pseudoconditioning groups were subdivided into alert and drowsy groups on the basis of EEG recordings during conditioning trials, giving a total of four groups: conditioning-alert, conditioning-drowsy, pseudoconditioning-alert and pseudoconditioning-drowsy. Using the latency criterion to define responses, significant conditioning occurred only in the conditioning-alert group for the following measures: skin resistance anticipatory and UCSomission responses, and finger plethysmograph anticipatory response. Both conditioning groups showed evidence of conditioned UCS-omission plethysmograph responses. Discussion centered primarily on (1) the implications of the results on the question of the feasibility of learning during sleep; and (2) the importance of including EEG recordings in studies of classically conditioned autonomic responses. The relationship between the various responses was also evaluated.
1. Introduction McDonald, Johnson and Hord (1964) have reported that certain changes occur in the orienting responses (ORs) of subjects who show electroencephalographic (EEG) signs of light sleep. Specifically, it was found that cardiovascular (heart rate and vasomotor) components of the OR increased significantly on later trials of an habituation series in those subjects whose EEGs indicated onset of drowsiness. The skin resistance OR, on the other hand, showed the same rate of habituation in both drowsy and alert groups. There were at least two major implications of these findings. First, there was the suggestion that there may be significant changes in cardiovascular components of the OR in other, more well-defined, stages of sleep, such as rapid eye movement (REM) sleep or stage 4 sleep. Indeed this has since been documented by Hord, Lubin and Johnson (1966) and by Johnson and Lubin (1967). Second, it was also clear that, in experiments on habituation of the OR, at least one channnel of EEG recording should be included in order that experimenters might have the capability of monitoring the extent of drowsiness in *Address for correspondence: D. G. McDonald, Department of Psychiatry, University of Missouri School of Medicine, Columbia, Missouri 65201, U.S.A. 101
102
D. G. McDonald
and L. C. Johnson
their subjects and therefore assessing the impact of this often overlooked variable on their results. In support of this need to include the EEG, we will show in this report that there are also significant differences in classically conditioned autonomic responses of alert versus drowsy subjects, in addition to the differences in ORs. The implications for learning during sleep will also be discussed below. 2. Method 2.1. Subjects The subjects were white male volunteers, all students in the U.S. Naval Hospital Corps School, San Diego. The results are based on a total Nof 69 subjects, although some additional subjects were excluded primarily because of equipment failures. The mean age of the total group was 19.1 yr with a range of 1724 yr; there were no differences in age between the four subgroups described below. 2.2. Apparatus A relatively detailed description of the apparatus and recording procedures has been previously reported by McDonald et al. (1964). Briefly, the recordings consisted of four channels of EEG (right and left occipital and parietal), skin resistance, finger plethysmograph, heart rate and respiration, recorded with an Offner type R dynagraph. 2.3. Procedure After an initial determination of each subject’s shock threshold, there followed a rest interval of 5-10 min. Subjects were instructed to try to stay awake, and they were also given a time-estimation task to reduce the incidence of drowsiness. This task was to try, as best as possible, to tap on a microphone once every 30 set and three times every 1.5min, except during presentation of a stimulus. All subjects were then given an habituation series which consisted of 10 presentations of the conditioned stimulus (CS), a 10 set 500 Hz tone, at 75-80 dB (reference = 0.0002 pbar). Immediately following the habituation trials, subjects in the conditioning group (N= 36) received 30 conditioning trials which consisted of the IO set CS followed immediately by a 0.5 set leg shock unconditioned stimulus (UCS). Shock intensity was varied randomly between 0.5 and 2.0 mA above the subject’s shock threshold. In addition, the UCS was presented on only 60 %, of the conditioning trials and was omitted on 12 of the 30 trials in a random order. This variation of shock intensity and partial presentation of the UCS were used in an attempt to reduce habituation to the UCS. Subjects in the pseudoconditioning control group (N- 33) received 32 unpaired presentations of either tone or shock alone. There were 20 tones and I2 shocks in the following order:
Conditioning in alert and drowsy subjects
103
TSTSTTSTSTSTTSTSTTSTSTSTTSTTSTTT. The intertrial intervals for all habituation, conditioning and pseudoconditioning trials were either 60 or 90 set in a random sequence. We were thus able to equate the two groups for (a) average intertrial interval, and (b) total time of the experimental session, but only by presenting the CS and UCS a somewhat different total number of times between the two groups, a problem recognized and discussed by Prokasy and Kumpfer (1973, pp. 168-169). 3. Results 3.1. Alert versus drowqj groups Following the procedures previously reported by McDonald et al. (1964), the EEG recordings of all subjects were visually scanned, and each of the conditioning or pseudoconditioning trials was rated for presence or absence of sleep. Subjects who showed any EEG evidence of sleep on any trial were classified as drowsy, and the remainder of the subjects were classified as alert. The term ‘drowsy’ has been chosen because the actual sleep stage varied from trial to trial, as well as across subjects, and no one standard label seemed appropriate. Generally, the EEG recordings were characterized by a significant loss of EEG alpha activity between trials, with frequent alpha enhancement observed with onset of the CS. This would most often correspond to initial descending sleep stage 1, although some stage 2 sleep was observed (cf. Rechtschaffen and Kales, 1968). Both the incidence and depth of sleep were very similar to that described in the earlier report (McDonald et al., 1964). Of the 36 conditioning subjects, 19 were classified alert and 17 drowsy; and of the 33 pseudoconditioning subjects, 12 were classified as being alert, with 21 judged to be drowsy. This gave us four groups: conditioning-alert (C-A), conditioning-drowsy (C-D), pseudoconditioning-alert (PC-A), and pseudoconditioning-drowsy (PC-D). A chi-square test of the differences in Ns between groups was not significant. 3.2. DeJnitions cfresponses Skin resistance responses and finger plethysmograph responses were categorized on the basis of latency of response onset, partly following the model of Stewart, Stern, Winokur and Fredman (1961), and partly on the basis of methods described by Davidoff and McDonald (1964) and McDonald and Johnson (1965). Skin resistance responses (SRRs) which occurred from 1.3 to 2.9 set after onset of the CS were categorized as orienting responses (ORs), from 3.0 to 11 .O set as anticipatory responses (ARs), and from 11.1 to 12.4 set on CS-only trials (the UCS-omission response, or response to shock in the absence of shock) the CR2. For finger plethysmograph responses (FPRs), the corresponding intervals were 2.0-6.0, 6. l-l 1.9, and 12.0-l 6.0 set respectively. Trials which were accompanied by irregularities in respiration were routinely excluded.from the analysis.
104
D. G. McDmald
and L. C. Johnson
3.3. Statistical methods The dependent measure for both SR- and FP-ORs was magnitude of response, the former in units of log conductance change, and the latter in millimeters of d.c. baseline shift. Both the ARs and CR,s were simply counted and the dependent measure was the percent of trials on which responses occurred for each subject; these two measures were transformed to the arc sine for purposes of the analyses of variance in the case of both SRR and FPR. We are reporting the results of 2 x 2 least squares analyses of variance for conditioning trials only. There were no differences between the four groups during habituation trials on any of the measures, except the FP- and HR-ORs which increased on later habituation trials in the two drowsy groups, corresponding to the earlier report (McDonald et al., 1964). The 2 x’ 2 analyses of the percent scores for each subject were necessary in part because of (a) unequal trials between conditioning and pseudoconditioning groups, and (b) the fact that ARs and CR,s were simply counted; further, we were more interested in demonstrating conditioning and alert versus drowsy effects than trials effects. Lastly, because of heterogeneity of variances in some of the measures we elected to compute MannWhitney U tests in those cases where the analyses of variance were significant as an additional precaution. Only the meaningful U tests were computed, as opposed to all possible combinations. The reader will note that there was indeed very good agreement between the results of the analyses of variance and the ranking techniques. 3.4. Skin resistance responses In the case of the SR-OR there were no significant differences, all of the Fs being less than I .OO. This means that there was no conditioning of the SR-OR, nor any difference between alert and drowsy groups, nor any interaction. Results for the SR-AR and CR,, on the other hand, were highly reliable and are presented in table I. Inspection of table I, as well asfigs. 1 and 2, clearly indicates that significant conditioning of these two responses occurred almost exclusively in the CA group, although there were modest differences between the C-D and PC-D groups. 3.5. Finger pleth~~smograph responses Analysis and interpretation of the FP-ORs was confounded by the fact that one might predict increases on the basis of both conditioning and drowsiness effects, or conversely that the PC-A group would show the least response. This in fact occurred: the means for all subjects over all conditioning trials for the C-A, C-D, PC-A and PC-D groups were 6.69,5.69, I .82 and 6.53, respectively. The conditioning by drowsiness interaction was significant (F = 1 I .80, SS = 13 1.99), the OR pattern being the opposite of the AR and CR,. Results of the analyses of FP-AR and CR, are summarized in table 2. While both main effects were significant for the AR, only the conditioning effect was
Co~djt~o~jng in alert aad drowsy subjects
105
Table I. Summary
of analyses of skin resistance measures.
response
ss
df
F
A. SR-AR Conditioning (C) Drowsiness (D) CxD
0.97 0.61 0.34
1 1 1
13.71* 8.61t 4.155
B. Sri-C& Conditioning Drowsiness CxD
2.58 1.75 0.84
1 1 1
27.88* 18.87* 9.11t
C. ‘U’ tests SR-AR
SR-CR2
comparison
U
Comparison
u
CA-CD CA-PCA CD-PCD PCA-PCD
50*
CA-CD CA-PCA CD-PCD PCA-PCD
799 44? 115$ 119
1:s; 92
*p < 0.001; tp i 0.01; $p < 0.05
73 -
0-l
-CONDITIONING-ALERT e~wCONDITIONING-DROWSY -PSEUDOCONDITIONING-ALERT ~_~PSEUDOCONDlTlONiNG-DR~SY
Habithim
’
Fig. 1. Mean percent of SR-AR responses in each group during habitLiation and conditioning trials.
106
D. G. McDonald
and L. C. Johnson
Table Summary
of analyses
A. FP-AR Conditioning Drowsiness CxD
(C) (D)
B. FP-CR2 Conditioning Drowsiness CxD c.
2.
of finger plethysmograph measures.
ss
df
F
0.18 0.20 0.01
I I I
9.22* 9.9g*
1.57 0.03 0.01
1
CLASSICAL
101-112.
CONDITIONING
IN ALERT AND DROWSY DAVID G. MCDONALD*
@ North-Holland Publishing Company
OF AUTONOMIC
RESPONSES
SUBJECTS
and LAVERNE
C. JOHNSON
U.S. Naly Medical Neuropsychiatric Research Unit, San Diego, California, U.S.A.
Accepted for publication 15 April 1975
The primary purpose of this experiment was to study classically conditioned autonomic responses of alert versus drowsy subjects. Sixty-nine subjects in conditioning and pseudoconditioning groups were subdivided into alert and drowsy groups on the basis of EEG recordings during conditioning trials, giving a total of four groups: conditioning-alert, conditioning-drowsy, pseudoconditioning-alert and pseudoconditioning-drowsy. Using the latency criterion to define responses, significant conditioning occurred only in the conditioning-alert group for the following measures: skin resistance anticipatory and UCSomission responses, and finger plethysmograph anticipatory response. Both conditioning groups showed evidence of conditioned UCS-omission plethysmograph responses. Discussion centered primarily on (1) the implications of the results on the question of the feasibility of learning during sleep; and (2) the importance of including EEG recordings in studies of classically conditioned autonomic responses. The relationship between the various responses was also evaluated.
1. Introduction McDonald, Johnson and Hord (1964) have reported that certain changes occur in the orienting responses (ORs) of subjects who show electroencephalographic (EEG) signs of light sleep. Specifically, it was found that cardiovascular (heart rate and vasomotor) components of the OR increased significantly on later trials of an habituation series in those subjects whose EEGs indicated onset of drowsiness. The skin resistance OR, on the other hand, showed the same rate of habituation in both drowsy and alert groups. There were at least two major implications of these findings. First, there was the suggestion that there may be significant changes in cardiovascular components of the OR in other, more well-defined, stages of sleep, such as rapid eye movement (REM) sleep or stage 4 sleep. Indeed this has since been documented by Hord, Lubin and Johnson (1966) and by Johnson and Lubin (1967). Second, it was also clear that, in experiments on habituation of the OR, at least one channnel of EEG recording should be included in order that experimenters might have the capability of monitoring the extent of drowsiness in *Address for correspondence: D. G. McDonald, Department of Psychiatry, University of Missouri School of Medicine, Columbia, Missouri 65201, U.S.A. 101
102
D. G. McDonald
and L. C. Johnson
their subjects and therefore assessing the impact of this often overlooked variable on their results. In support of this need to include the EEG, we will show in this report that there are also significant differences in classically conditioned autonomic responses of alert versus drowsy subjects, in addition to the differences in ORs. The implications for learning during sleep will also be discussed below. 2. Method 2.1. Subjects The subjects were white male volunteers, all students in the U.S. Naval Hospital Corps School, San Diego. The results are based on a total Nof 69 subjects, although some additional subjects were excluded primarily because of equipment failures. The mean age of the total group was 19.1 yr with a range of 1724 yr; there were no differences in age between the four subgroups described below. 2.2. Apparatus A relatively detailed description of the apparatus and recording procedures has been previously reported by McDonald et al. (1964). Briefly, the recordings consisted of four channels of EEG (right and left occipital and parietal), skin resistance, finger plethysmograph, heart rate and respiration, recorded with an Offner type R dynagraph. 2.3. Procedure After an initial determination of each subject’s shock threshold, there followed a rest interval of 5-10 min. Subjects were instructed to try to stay awake, and they were also given a time-estimation task to reduce the incidence of drowsiness. This task was to try, as best as possible, to tap on a microphone once every 30 set and three times every 1.5min, except during presentation of a stimulus. All subjects were then given an habituation series which consisted of 10 presentations of the conditioned stimulus (CS), a 10 set 500 Hz tone, at 75-80 dB (reference = 0.0002 pbar). Immediately following the habituation trials, subjects in the conditioning group (N= 36) received 30 conditioning trials which consisted of the IO set CS followed immediately by a 0.5 set leg shock unconditioned stimulus (UCS). Shock intensity was varied randomly between 0.5 and 2.0 mA above the subject’s shock threshold. In addition, the UCS was presented on only 60 %, of the conditioning trials and was omitted on 12 of the 30 trials in a random order. This variation of shock intensity and partial presentation of the UCS were used in an attempt to reduce habituation to the UCS. Subjects in the pseudoconditioning control group (N- 33) received 32 unpaired presentations of either tone or shock alone. There were 20 tones and I2 shocks in the following order:
Conditioning in alert and drowsy subjects
103
TSTSTTSTSTSTTSTSTTSTSTSTTSTTSTTT. The intertrial intervals for all habituation, conditioning and pseudoconditioning trials were either 60 or 90 set in a random sequence. We were thus able to equate the two groups for (a) average intertrial interval, and (b) total time of the experimental session, but only by presenting the CS and UCS a somewhat different total number of times between the two groups, a problem recognized and discussed by Prokasy and Kumpfer (1973, pp. 168-169). 3. Results 3.1. Alert versus drowqj groups Following the procedures previously reported by McDonald et al. (1964), the EEG recordings of all subjects were visually scanned, and each of the conditioning or pseudoconditioning trials was rated for presence or absence of sleep. Subjects who showed any EEG evidence of sleep on any trial were classified as drowsy, and the remainder of the subjects were classified as alert. The term ‘drowsy’ has been chosen because the actual sleep stage varied from trial to trial, as well as across subjects, and no one standard label seemed appropriate. Generally, the EEG recordings were characterized by a significant loss of EEG alpha activity between trials, with frequent alpha enhancement observed with onset of the CS. This would most often correspond to initial descending sleep stage 1, although some stage 2 sleep was observed (cf. Rechtschaffen and Kales, 1968). Both the incidence and depth of sleep were very similar to that described in the earlier report (McDonald et al., 1964). Of the 36 conditioning subjects, 19 were classified alert and 17 drowsy; and of the 33 pseudoconditioning subjects, 12 were classified as being alert, with 21 judged to be drowsy. This gave us four groups: conditioning-alert (C-A), conditioning-drowsy (C-D), pseudoconditioning-alert (PC-A), and pseudoconditioning-drowsy (PC-D). A chi-square test of the differences in Ns between groups was not significant. 3.2. DeJnitions cfresponses Skin resistance responses and finger plethysmograph responses were categorized on the basis of latency of response onset, partly following the model of Stewart, Stern, Winokur and Fredman (1961), and partly on the basis of methods described by Davidoff and McDonald (1964) and McDonald and Johnson (1965). Skin resistance responses (SRRs) which occurred from 1.3 to 2.9 set after onset of the CS were categorized as orienting responses (ORs), from 3.0 to 11 .O set as anticipatory responses (ARs), and from 11.1 to 12.4 set on CS-only trials (the UCS-omission response, or response to shock in the absence of shock) the CR2. For finger plethysmograph responses (FPRs), the corresponding intervals were 2.0-6.0, 6. l-l 1.9, and 12.0-l 6.0 set respectively. Trials which were accompanied by irregularities in respiration were routinely excluded.from the analysis.
104
D. G. McDmald
and L. C. Johnson
3.3. Statistical methods The dependent measure for both SR- and FP-ORs was magnitude of response, the former in units of log conductance change, and the latter in millimeters of d.c. baseline shift. Both the ARs and CR,s were simply counted and the dependent measure was the percent of trials on which responses occurred for each subject; these two measures were transformed to the arc sine for purposes of the analyses of variance in the case of both SRR and FPR. We are reporting the results of 2 x 2 least squares analyses of variance for conditioning trials only. There were no differences between the four groups during habituation trials on any of the measures, except the FP- and HR-ORs which increased on later habituation trials in the two drowsy groups, corresponding to the earlier report (McDonald et al., 1964). The 2 x’ 2 analyses of the percent scores for each subject were necessary in part because of (a) unequal trials between conditioning and pseudoconditioning groups, and (b) the fact that ARs and CR,s were simply counted; further, we were more interested in demonstrating conditioning and alert versus drowsy effects than trials effects. Lastly, because of heterogeneity of variances in some of the measures we elected to compute MannWhitney U tests in those cases where the analyses of variance were significant as an additional precaution. Only the meaningful U tests were computed, as opposed to all possible combinations. The reader will note that there was indeed very good agreement between the results of the analyses of variance and the ranking techniques. 3.4. Skin resistance responses In the case of the SR-OR there were no significant differences, all of the Fs being less than I .OO. This means that there was no conditioning of the SR-OR, nor any difference between alert and drowsy groups, nor any interaction. Results for the SR-AR and CR,, on the other hand, were highly reliable and are presented in table I. Inspection of table I, as well asfigs. 1 and 2, clearly indicates that significant conditioning of these two responses occurred almost exclusively in the CA group, although there were modest differences between the C-D and PC-D groups. 3.5. Finger pleth~~smograph responses Analysis and interpretation of the FP-ORs was confounded by the fact that one might predict increases on the basis of both conditioning and drowsiness effects, or conversely that the PC-A group would show the least response. This in fact occurred: the means for all subjects over all conditioning trials for the C-A, C-D, PC-A and PC-D groups were 6.69,5.69, I .82 and 6.53, respectively. The conditioning by drowsiness interaction was significant (F = 1 I .80, SS = 13 1.99), the OR pattern being the opposite of the AR and CR,. Results of the analyses of FP-AR and CR, are summarized in table 2. While both main effects were significant for the AR, only the conditioning effect was
Co~djt~o~jng in alert aad drowsy subjects
105
Table I. Summary
of analyses of skin resistance measures.
response
ss
df
F
A. SR-AR Conditioning (C) Drowsiness (D) CxD
0.97 0.61 0.34
1 1 1
13.71* 8.61t 4.155
B. Sri-C& Conditioning Drowsiness CxD
2.58 1.75 0.84
1 1 1
27.88* 18.87* 9.11t
C. ‘U’ tests SR-AR
SR-CR2
comparison
U
Comparison
u
CA-CD CA-PCA CD-PCD PCA-PCD
50*
CA-CD CA-PCA CD-PCD PCA-PCD
799 44? 115$ 119
1:s; 92
*p < 0.001; tp i 0.01; $p < 0.05
73 -
0-l
-CONDITIONING-ALERT e~wCONDITIONING-DROWSY -PSEUDOCONDITIONING-ALERT ~_~PSEUDOCONDlTlONiNG-DR~SY
Habithim
’
Fig. 1. Mean percent of SR-AR responses in each group during habitLiation and conditioning trials.
106
D. G. McDonald
and L. C. Johnson
Table Summary
of analyses
A. FP-AR Conditioning Drowsiness CxD
(C) (D)
B. FP-CR2 Conditioning Drowsiness CxD c.
2.
of finger plethysmograph measures.
ss
df
F
0.18 0.20 0.01
I I I
9.22* 9.9g*
1.57 0.03 0.01
1
