Pain 95 (2002) 111–118

www.elsevier.com/locate/pain

The role of operant conditioning in chronic pain:

an experimental investigation
Herta Flor a,*, Bärbel Knost b, Niels Birbaumer b,c
a
Department of Neuropsychology at the Ruprecht-Karls-University of Heidelberg, Central Institute of Mental Health, J 5, D-68159 Mannheim, Germany
b
Institute of Medical Psychology and Behavioral Neurobiology, University of Tübingen, Gartenstrasse 29, D-72074 Tübingen, Germany
c
Department of General Psychology, University of Padua, Via Venezia 8, I-35139 Padua, Italy
Received 9 October 2000; received in revised form 28 June 2001; accepted 24 July 2001

Abstract
The role of operant conditioning for the development and maintenance of chronic pain was examined in 30 chronic back pain patients
(CBP) and 30 matched healthy controls. Half of each group was reinforced for increased, half for decreased pain reports while EEG, EOG,
heart rate, skin conductance and muscle tension levels were recorded. Both groups showed similar learning rates, however, the CBP patients
displayed slower extinction of both the verbal and the cortical (N150) pain response. In addition, the CBP group displayed prolonged
elevated electromyogram levels to the task. These data suggest that CBP patients are more easily influenced by operant conditioning factors
than healthy controls and this susceptibility may add to the maintenance of the chronic pain problem. q 2002 International Association for
the Study of Pain. Published by Elsevier Science B.V. All rights reserved.
Keywords: Operant factors; Learning; Evoked potentials; Electromyogram

1. Introduction instead of the ischemic pain model. For the up-training

condition, a significant increase in pain ratings as compared
The operant model of chronic pain (Fordyce, 1976) to baseline was observed, whereas the subjects in the control
suggests that positive and negative reinforcement of acute group, that were not reinforced positively or punished for
pain behaviors such as moaning, inactivity or medication altered pain ratings, showed no change. Pain report
intake may lead to the development of chronic pain. increases were, however, higher when punishing responses
Although this model has had a significant impact on the (requests to better concentrate) were provided. Skin conduc-
treatment of chronic pain, little experimental evidence for tance responses also showed a significant increase in the
its validity is available. Linton and Götestam (1985) used experimental group. However, the replication of this study
calibrated painful ischemic stimuli and reinforced healthy failed. In an up- and down-conditioning paradigm with an
subjects for increases (up-training) or decreases (down- additional unreinforced control group neither the up-condi-
training) of verbal pain reports. Although the authors tioning nor the down-conditioning of pain reports could be
reported a within-subject differentiation between increases established. SCR changes were found to be more associated
and decreases of these reports, the study must be interpreted with arousal than with subjective pain intensity. The authors
with caution since individual learning curves were very suggested that the modified (punishing statements were now
variable and the ratings in the decrease pain condition free of reference to attention) verbal feedback prevented the
were higher than the baseline ratings. Furthermore, the appearance of conditioning effects. Therefore, the
subjects could easily predict the physical intensity of the previously shown conditioning effects might have been
stimuli since their intensity was dependent on the blow-up due to enhanced attention rather than operant conditioning.
time of the cuff. Replication studies by Lousberg and colla- In addition to verbal pain reports and autonomic measures
borators (Lousberg et al., 1996) failed to show clear operant the amplitude of the somatosensory evoked potential (SEP)
conditioning effects. These authors used electric stimuli related to painful stimuli has also been the target of operant
conditioning experiments. The N150–P260 peak to peak
amplitude of the SEP has been found to be positively corre-
* Corresponding author. Tel.: 149-0621-1703-922/-918; fax: 149-0621-
1703-932. lated with subjective pain intensity (Bromm, 1984; Chap-
E-mail address: [email protected] (H. Flor). man and Jacobson, 1984; Chen et al., 1979; Miltner et al.,
0304-3959/02/$20.00 q 2002 International Association for the Study of Pain. Published by Elsevier Science B.V. All rights reserved.
PII: S 0304-395 9(01)00385-2
112 H. Flor et al. / Pain 95 (2002) 111–118

1988a,b). Rosenfeld et al. (1985) reported that half of their allergy to plaster, (h) left-handedness and (i) use of centrally
subjects showed a strong positive correlation between the acting analgesics, muscle relaxants or tranquilizers. The
reinforced SEP-amplitude changes and pain sensitivity and experimental groups were matched with respect to gender,
that the other half showed strong negative correlations age and education (see Table 1). The healthy controls were
between the two variables. Miltner et al. (1988b) reported paid (10 DM/h) for participation and the chronic pain
increased pain reports to painful intracutaneous electric patients could choose between payment (N ¼ 21) and
stimulation of the finger when subjects were reinforced for psychological counseling (N ¼ 9). Half of the subjects in
increasing the SEP peak-to-peak amplitude of the N150/ each group were randomly assigned to a down-training
P260. A study by Dowman (1996) found no effect of rein- group and the other half to an up-training group.
forced changes in P200 amplitude on pain sensitivity. The
author rewarded the subjects for upward or downward 2.2. Experimental design
changes of the P200 amplitude during sural nerve stimula-
tion without explicit instructions how they could alter the Upon entering the laboratory, subjects were informed that
P200 amplitude. The subjects were asked to find some the purpose of the experiment was to determine the ability
psychological state that changed the amplitude. After the of pain patients and healthy controls to estimate various pain
conditioning procedure was completed, somatosensory intensities. The study was approved by the local ethics
sensitivity was tested by intensity ratings of 80 stimuli of committee and adhered to the Declaration of Helsinki in
eight different intensities varying between just above its revised version of Tokyo. The subjects signed informed
compound nerve action potential threshold and just below consent and completed several psychological assessment
subjective tolerance. Because the pain ratings were recorded instruments prior to the laboratory session. After the attach-
after the training, rapid extinction can be responsible for the ment of the EEG, EOG, EMG and SCL electrodes, a gold
lack of the conditioning effect on the subjective level. None electrode was inserted into a small opening in the upper
of the studies tested chronic pain patients in order to deter- layer of the skin of the middle finger of the left hand as
mine to what extent they might already have undergone an described by Bromm and Meier (1984). An 8-min adapta-
operant conditioning process and therefore show facilitated tion period and a 2-min resting baseline period followed.
learning. Pain threshold and pain tolerance for the arm and the back
The aim of the present study was to examine the effects of muscles were determined in two ascending and descending
operant conditioning of verbal pain expressions in both series of electric stimuli. The duration of a single electric
chronic back pain patients and healthy controls. It was stimulus was 50 ms, the stimulation frequency was 1 Hz.
hypothesized that chronic pain patients might have an The subjects indicated verbally when pain and tolerance
extended history of operant conditioning of pain-related threshold were reached. The mean amperage of the four
responses and might therefore learn the required responses intensity values obtained for the pain threshold was used
faster and might show greater resistance to extinction. In as individual pain threshold, the mean amperage of the
addition to verbal reports, central and peripheral correlates two tolerance values was used as pain tolerance level.
of pain were assessed to determine to what extent the verbal Four individual shock intensities were employed: pain
conditioning effect would generalize. threshold, 25, 50, and 75% of the distance between pain
threshold and tolerance. During habituation, 15 electric
stimuli at each of the four intensity levels (N ¼ 60) were
2. Methods delivered. The training phase consisted of the application of
20 electric stimuli of each intensity (N ¼ 80). During
2.1. Subjects extinction, 15 stimuli of each level were presented
(N ¼ 60). After shock delivery, the subjects rated the inten-
Thirty chronic back pain patients and 30 matched healthy sity of the electric stimulus on a visual analogue scale
volunteers recruited by newspaper advertisements that provided on the computer screen. Additional feedback
sought patients with chronic low back pain interested in (computer smiley and monetary gain) was offered to the
learning more about their problem participated in the subjects in the training phase 1 s after they had completed
study. The chronic pain patients had been suffering from their rating.
continuous low back pain for at least 6 months. The patients In the up-training group, subjects were given positive
had mainly descriptive diagnoses such as non-specific back feedback when their actual pain rating was higher than the
pain, back pain of muscular origin, musculoskeletal pain, average rating of the last ten trials for each intensity level in
etc. pertaining to IASP category XXVI-9. Exclusion criteria the habituation period (baseline rating), lower ratings were
for all participants were (a) inflammatory cause of pain (e.g., followed by negative feedback. In the down-training group
rheumatoid arthritis), (b) neurological complications (e.g., the positive/negative feedback assignment was defined
loss of sensory or motor function, nerve root inflammation), conversely. Positive as well as negative feedback was
(c) predominance of an unrelated pain syndrome, (d) cardiac arranged in four steps according to the degree of deviance
pacemaker, (e) major psychiatric illness, (f) drug abuse, (g) from the baseline ratings. Neutral feedback was provided
 H. Flor et al. / Pain 95 (2002) 111–118 113

Table 1
Demographic and clinical characteristics of the groups

Healthy Controls Patients

Up M (SD) Down M (SD) Up M (SD) Down M (SD)

Age (in years) 39.33 (11.58) 39.27 (10.31) 43.14 (10.81) 43.27 (11.95)
Sex
Male (N) 5 5 6 5
Female (N) 10 10 8 10
Education
#10 school years (N) 9 6 9 9
.10 school years (N) 6 9 5 6
MPI a 1 (0–6)
Pain severity – – 3.07 (1.06) 3.38 (1.32)
Interference – – 3.39 (1.30) 2.98 (1.39)
Life control – – 4.00 (1.20) 3.62 (1.32)
Affective distress – – 2.58 (1.28) 2.89 (1.68)
Support – – 3.36 (1.50) 3.29 (1.81)
MPI 2 (0–6)
Punishing responses – – 1.07 (1.25) 0.69 (0.65)
Solicitous responses – – 2.50 (1.69) 2.95 (1.57)
Distracting response – – 2.90 (1.69) 3.03 (1.44)
MPI 3 (0–6)
General activity – – 2.94 (0.74) 2.68 (0.71)
BSQ b
General stress 1.31 (0.91) 1.36 (0.65) 2.22 (0.95) 2.36 (0.79)
PRSS c
Catastrophizing 0.59 (0.67) 0.66 (0.63) 1.50 (0.98) 1.81 (0.94)
Coping 3.52 (1.04) 3.99 (0.72) 3.14 (0.62) 3.17 (0.80)
BDI d 5.86 (4.98) 3.87 (3.44) 9.87 (7.05) 12.27 (6.92)
Somatic-Complaints Scale 12.93 (7.82) 12.67 (10.45) 20.07 (8.37) 21.47 (11.36)
Perception Threshold in mA 1.25 (0.66) 1.20 (0.74) 0.72 (0.47) 0.65 (0.65)
Pain threshold in mA 2.27 (0.94) 2.14 (1.29) 1.68 (1.21) 1.56 (1.17)
Pain tolerance in mA 3.06 (1.26) 3.00 (1.71) 2.31 (1.44) 2.41 (1.66)
a
West-Haven Yale Multidimensional Pain Inventory.
b
Brief Stress Questionnaire.
c
Pain-Related Self Statements.
d
Beck Depression Inventory.

when the actual rating was equal to the baseline rating. control, and significant others’ responses to pain behaviors
Feedback consisted of a computer smiley that moved the of the patients as well as the general activity levels. This
corner of the mouth up (positive feedback) or down (nega- measure was used to describe the clinical characteristics of
tive feedback) depending on the degree of reinforcement. the chronic back pain patients. The frequency of coping and
Neutral feedback was represented by a straight line repre- catastrophizing pain-related self-statements, as measured by
senting the mouth. Additionally, a counter was presented the PRSS, was recorded in order to determine the relation-
that added (positive feedback) or subtracted (negative feed- ship of deficient coping and operant conditioning of pain
back) the amount of money earned in addition to the usual ratings. The BDI was used as a measure of depressed mood,
payment in four steps. In the case of neutral feedback, the the Brief Stress Scale as a measure of general stress. Both,
counter remained unchanged. mood and stress, may affect the perception of painful elec-
tric stimuli and, therefore, the acquisition rate of the condi-
2.3. Psychological assessment tioned responses.

Patients completed the German version of the West 2.4. Physiological recordings and instrumentation
Haven-Yale Multidimensional Pain Inventory (MPI, Flor
et al., 1990; Kerns et al., 1985), the Pain-Related Self-State- Electroencephalographic (EEG) data were recorded from
ment Scale (PRSS, Flor et al., 1993; Flor and Turk, 1988), 29 scalp locations using the International 10–20–System
the Beck Depression Inventory (BDI, Beck et al., 1961), the (Jasper, 1958) and extending it to a 10–10 system (Fpz,
Brief Stress Scale (Flor, 1991), and the Somatic-Complaints F7, F3, Fz, F4, F8, FC5, FC1, FC2, FC6, T3, C3, Cz, C4,
Scale (B-L, von Zerssen, 1976). The MPI assesses pain T4, CP5, CP1, CP2, CP6, T5, P3, Pz, P4, T6, PO1, PO2, O1,
severity, interference, affective distress, social support, life Oz, O2). The EEG channels were recorded based on the
114 H. Flor et al. / Pain 95 (2002) 111–118

guidelines for EEG recording of Pivik et al. (Pivik et al., nents consisted of the negatively deflecting N150 (peaking
1993). In addition, lateral and vertical EOGs were assessed. in the latency range 80–180 ms post stimulus, reference
Scalp recordings were monopolar referenced to Cz and peak latency at Cz), the positively deflecting P260 (peaking
converted off-line into linked earlobe-referenced record- in latency range 180–280 ms, reference peak latency at Cz)
ings. EOG recordings were bipolar. EEG and EOG were and the positively deflecting P300 (peaking in the latency
recorded by Neuroscan Synamps amplifiers and sampled range 280–400 ms, reference peak latency at Pz). The peaks
at a rate of 200 Hz (dc to 70 Hz low pass filter). The first were determined by visual inspection. As an additional
phase of data acquisition began 1000 ms prior to the electric ERP-parameter, the amplitude difference between the
stimulus and was terminated 4 s after the delivery of the N150- and the P260-component (N150/P260) was analyzed
electric stimulus. During acquisition, a second measurement as a potential cortical correlate of subjective pain experience
phase was used that began 1 s prior to the feedback and (Bromm, 1984; Chapman and Jacobson, 1984; Miltner et
continued for 5 s after feedback onset. For the peripheral al., 1988b).
data, the beginning of the first data acquisition phase was 4 s For pain ratings and ERP parameters, Greenhouse–Geis-
prior to the electric stimulus. Intertrial intervals varied ser corrected repeated-measures analyses of variance
randomly between 4 and 8 s. (Jennings and Wood, 1976) were computed using the two
Electromyographic activity was recorded bilaterally from between factors ‘group’ (chronic back pain patients versus
two muscle sites: M. flexor communis digitorum and M. healthy controls) and ‘direction‘ (up versus down training),
erector spinae. Electrode placements were based on the and the within factors ‘intensity’ (four intensity levels) and
recommendations of Fridlund and Cacioppo (1986) and ‘phase’ (habituation, acquisition, extinction). In addition, all
Lippold (1967). A Coulbourn S75-01 bioamplifier with a experimental phases were segmented into blocks of five
bandpass filter from 90 to 250 Hz was used to collect the trials for all four intensities. For each phase, single
EMG data. This signal was rectified and integrated on-line ANOVAs for the pain ratings were carried out with the
by a Coulbourn S76-01 contour following integrator with a between factors ‘group’ and ‘direction’ and the within
time constant of 20 ms. A sampling rate of 50 Hz was used factors ‘blocks’ (three blocks in habituation and extinction,
for the integrated EMG. four blocks in acquisition) and ‘intensity’. For the pain
The electrocardiogram was assessed using three 8-mm ratings, additional comparisons were made for the last
electrodes: left and right lower ribcage and clavicula. A block of the preceding and the first block of the respective
Schmitt-trigger was employed to determine the R-waves phase. For the ERP parameters, learning curves were
for the computation of heart rate. Skin conductance level analyzed by two ANOVAs with the factors ‘group’, ‘direc-
(SCL) was recorded by two 11 mm-electrodes from the tion’, ‘phase’ (habituation vs. acquisition and acquisition vs.
thenar and hypothenar eminence, according to the recom- extinction) and ‘blocks’ (first and last block of the analyzed
mendations of Fowles et al. (1981). A Coulbourn S71-22 phases). Post-hoc Tukey-tests were performed using
skin conductance coupler, which processed the signal with Bonferroni-adjusted alpha levels.
DC coupling and an amplification factor of 50 mV/ms,
provided a constant voltage of 0.5 V across electrodes. All
peripheral data were assessed via Coulbourn amplifiers and 3. Results
processed by a personal computer. ECG electrode impe-
dances were less than 10 kohms, all other electrode impe- 3.1. Psychological assessment and stimulation data
dances (except for SCL) were maintained below 5 kohms. A
ground electrode connected with all peripheral channels and A summary of the psychological assessment data is
spanning the whole wrist was placed on the left arm. The shown in Table 1. There were no significant differences in
subjects were seated in a dimly illuminated and sound-atte- any of the psychological assessment data between the up-
nuated room. and down training groups neither for the chronic back pain
patients nor for the healthy controls. Comparisons between
2.5. Data reduction and analysis the chronic back pain patients and the healthy controls were
as follows: As expected, chronic back pain patients showed
EOG artifact was corrected using an algorithm adapted significantly more catastrophizing (tð1Þ ¼ 24:845,
from Gratton et al. (1983). For all experimental phases and P , 0:001) and less coping pain-related self-statements
intensity levels, the dependent physiological variables (tð1Þ ¼ 2:867, P ¼ 0:006), they experienced significantly
(somatosensory evoked potentials (SEP), EMG, SCL, and more general stress (tð1Þ ¼ 24:467, P , 0:001), showed a
heart rate (HR) were averaged separately. After baseline higher level of depressed mood (tð1Þ ¼ 24:122, P , 0:001)
correction (referenced to 100 ms preceding the electric and stated significantly more somatic complaints
stimulus) and averaging within conditions, three peak (tð1Þ ¼ 23:245, P ¼ 0:002) as compared to the healthy
amplitudes (150, 260 and 300 ms) and the respective laten- group. There were no significant correlations between
cies were obtained. For the SEP components later than 400 stimulation-related pain ratings during habituation, acquisi-
ms no clear peak could be identified. The early SEP compo- tion or extinction and the psychological assessment data.
 H. Flor et al. / Pain 95 (2002) 111–118 115

The patients had significantly lower pain thresholds than the The analysis of the blocks for acquisition revealed a
healthy controls (tð56Þ ¼ 23:27, P ¼ 0:002) and tended to significant interaction between direction and block
show also lower perception thresholds (tð56Þ ¼ 21:96, (Fð3; 165Þ ¼ 16:93, P , 0:001, 1 ¼ 0:58) indicating that
P ¼ 0:055) and pain tolerances (tð56Þ ¼ 21:71, subjects’ pain ratings moved in the expected directions.
P ¼ 0:093, see Table 1). Post hoc contrasts showed that the pain ratings in the
down-training decreased significantly for each subsequent
block (all P values ,0.05). However, in the up-training, the
3.2. Pain ratings
pain ratings showed a significant increase only between
The overall ANOVA yielded a significant interaction blocks 1 and 3 (tð29Þ ¼ 22:12, P ¼ 0:04), and blocks 1
between direction and phase (Fð2; 110Þ ¼ 41:53, and 4 (tð29Þ ¼ 22:53, P ¼ 0:02). The comparison between
P , 0:001, 1 ¼ 0:78) indicating higher pain ratings in the the last block of the habituation phase and the first block of
up- as compared to the down-training in the acquisition the acquisition phase also yielded significantly higher pain
(tð1Þ ¼ 23:51, P ¼ 0:001) and extinction (tð1Þ ¼ 23:95, ratings for the up-training condition (Fð1; 29Þ ¼ 15:54,
P , 0:001) but not in the habituation phase P , 0:001).
(tð1Þ ¼ 20:46, ns, See Fig. 1). The main effect phase The ANOVA for the extinction phase revealed a signifi-
(Fð2; 110Þ ¼ 5:87, P ¼ 0:008,1 ¼ 0:78) indicated that the cant interaction between group and block
pain ratings were significantly lower in the habituation as (Fð2; 110Þ ¼ 3:328, P ¼ 0:043, 1 ¼ 0:63). During extinc-
compared to the acquisition phase (tð58Þ ¼ 22:66, tion, the ratings of the healthy controls decreased continu-
P ¼ 0:01) and significantly higher in the acquisition phase ously independent of the training condition (decrease from
as compared to the extinction phase (tð58Þ ¼ 2:41, block 1 to 2: tð29Þ ¼ 2:55, P ¼ 0:016); block 2–3: tð29Þ ¼
P ¼ 0:02). Furthermore, there was a significant main effect 2:68; P ¼ 0:012); block 1–3: tð29Þ ¼ 3:37, P ¼ 0:002),
of intensity level (Fð3; 165Þ ¼ 94:89, P , 0:001, 1 ¼ 0:40) whereas the ratings of the chronic back pain patients did
indicating that during all experimental phases the subjects not show extinction (all P values ns).
differentiated among the four intensity levels.
3.3. Somatosensory-evoked potentials

N150. The overall ANOVA revealed that the N150-

component of the somatosensory evoked potential (SEP)
was in general significantly enhanced in the chronic back
pain patients as compared to the healthy controls (effect
group: Fð1; 55Þ ¼ 5:34, P ¼ 0:025, see Fig. 2). There was
no significant intensity effect (Fð3; 165Þ ¼ 0:459, ns).
Furthermore, the overall ANOVA showed significant
phase differences (Fð2; 110Þ ¼ 8:100, P ¼ 0:001,
1 ¼ 0:78) indicating a general habituation of the N150
component of the SEP over the course of the experiment
(see Fig. 3).
Closer analysis of the phases of the experiment revealed a
significant interaction between group and direction during
extinction (Fð1; 55Þ ¼ 3:94, P ¼ 0:05), indicating signifi-

Fig. 2. N150 amplitude (SEM as error bar) for the chronic back pain
Fig. 1. Pain ratings in the habituation, acquisition and extinction phase of patients and the healthy controls in the habituation, acquisition and extinc-
the up- and down-training groups for the chronic back pain patients (top) tion phases (averaged across up- and down training groups and across
and the healthy controls (bottom). intensity levels).
116 H. Flor et al. / Pain 95 (2002) 111–118

P260.The overall ANOVA for P260 revealed that ampli-

tudes decreased significantly over the three phases
(Fð2; 110Þ ¼ 81:960, P # 0:001, 1 ¼ 0:72) indicating a
general habituation effect (see Fig. 3). Furthermore, there
was a significant intensity effect (Fð3; 165Þ ¼ 4:22,
P , 0:009). Post hoc tests revealed that P260 amplitudes
were significantly higher at intensity level 4 than at levels
1 (tð58Þ ¼ 22:82, P ¼ 0:007) and 2 (tð58Þ ¼ 22:58,
P ¼ 0:012). Level 3 elicited a significantly higher P260-
amplitude than level 1 (tð58Þ ¼ 22:33, P ¼ 0:024). The
ANOVA for the blocks of the habituation and acquisition
phase revealed no significant effects.
Peak-to-peak amplitude N150/P260.For the N150/P260
Fig. 3. ERP waveforms for the three experimental phases averaged across
all subjects and four intensity levels. amplitude, the overall ANOVA showed a significant phase
effect (Fð2; 110Þ ¼ 90:69, P , 0:001, 1 ¼ 0:70) indicating
habituation over the time course of the experiment (see Fig.
cant extinction of the N150-amplitude in the up-training 3). Post hoc tests revealed decreasing amplitudes between
condition in the healthy controls (tð14Þ ¼ 24:40, P ¼ habituation and acquisition (tð58Þ ¼ 9:56, P , 0:001) and
0:001 see Fig. 4) but not the chronic back pain patients between acquisition and extinction (tð58Þ ¼ 6:70,
(tð14Þ ¼ 20:80, ns). P , 0:001). Furthermore, higher intensity levels elicited
The ANOVA for the first and last blocks of the habitua- larger N150/P260 amplitudes (Fð3; 165Þ ¼ 4:96,
tion and acquisition phases revealed a significant group P ¼ 0:005, 1 ¼ 0:82). Post hoc tests showed that N150/
main effect (Fð1; 55Þ ¼ 4:43, P ¼ 0:040). Furthermore, the P260 amplitudes differentiated significantly between inten-
three-way interaction between direction, blocks and phase sity levels 1 and 4 (tð58Þ ¼ 10:96, P ¼ 0:002), levels 2 and 4
was significant Fð1; 55Þ ¼ 5:483, P ¼ 0:023). For the down- (tð58Þ ¼ 4:63, P ¼ 0:036) as well as between levels 1 and 3
training, the N150 amplitude during the last training block (tð58Þ ¼ 7:73, P ¼ 0:007). The analysis for the segmented
was significantly lower than in the habituation blocks (first extinction phase revealed no significant effects.
block of the habituation vs. last down-training: P300. No significant group or interaction effects could be
tð29Þ ¼ 23:155, P ¼ 0:004; last habituation block vs. last observed. Again, there was a significant general habituation
down-training block: tð29Þ ¼ 24:20, P , 0:001). For the effect (Fð2; 110Þ ¼ 129:011, P , 0:001) (see Fig. 3) and a
up-training, the N150 amplitude during the first block of significant amplitude difference among the four intensity
the habituation was enhanced as compared to the first levels (Fð3; 165Þ ¼ 30:276, P , 0:001).
block of the acquisition (tð28Þ ¼ 22:75, P ¼ 0:01) but no None of the latency analyses became significant suggest-
significant amplitude differences between the habituation ing that stimulus latencies were not affected by the condi-
phases and the last block of the up-training could be tioning procedure. There were no significant group or
found. The ANOVA for the blocks of the extinction phase condition differences in the phase before the feedback.
showed no significant effects. These data are therefore not presented.

3.4. Peripheral measures

There were no baseline differences in the bilateral m.

communis flexor digitorum and the bilateral m. erector
spinae EMGs. Furthermore, no significant group or direc-
tion differences could be detected for the electromyographic
activity of the right forearm. The ANOVA with the between
factor ‘group’ and the within factor ‘time’ course (baseline,
0–500 ms, 500–1000 ms) for the left flexor communis digi-
torum revealed a significant two-way interaction between
group and time course (Fð2; 114Þ ¼ 18:460, P , 0:001,
1 ¼ 0:9160) and a significant main effect of time course
(Fð2; 114Þ ¼ 24:884, P , 0:001). Post hoc tests showed
that the chronic back pain patients had significantly higher
EMG-levels than the healthy controls in the late period after
Fig. 4. N150 amplitude (SEM as error bar) for the chronic back pain stimulus onset (500–1000 ms) (tð1Þ ¼ 24:467, P , 0:001).
patients and the healthy controls for the up-and down-training conditions Within group tests showed that there were significantly
in the extinction phase. enhanced EMG-levels in the first 500 ms as compared to
 H. Flor et al. / Pain 95 (2002) 111–118 117

the baseline in both groups (P , 0:05), however, only the as well as primary and secondary somatosensory cortex
chronic back pain group showed a prolonged enhanced (Desmedt and Tomberg, 1989; Raichle, 1994) and the
EMG (P , 0:001). The ANOVA for the left m. erector modifiability of this component by the reinforcement of
spinae revealed a significant interaction between group verbal pain reports has important implications for the
and time course (Fð1; 55Þ ¼ 24:402, P , 0:001). Closer processing of painful stimuli in chronic pain patients.
analysis of this effect showed that EMG activity was These data suggest that external reinforcement of pain beha-
reduced in the period 500–1000 ms after electric stimulus vior enhances the subsequent processing of painful stimuli.
onset compared to the early period (1–500 ms) To what extent the affective or the sensory pain components
(tð29Þ ¼ 3:558, P ¼ 0:001) in the healthy controls. In (Rainville et al., 1997) are modified by operant conditioning
contrast, the chronic back pain patients showed enhanced cannot be determined from this study because source
EMG activity in the period 500–1000 ms as compared to 1– analyses were precluded by reduced signal to noise ratios
500 ms (tð28Þ ¼ 23:739, P ¼ 0:001). in the patient group.
The ANOVA for SCL did not reveal any significant P200, which reflects late evaluative processes in the pain
effects. The analysis of heart rate showed a significant experience, was not influenced by operant conditioning nor
main effect of phase Fð2; 110Þ ¼ 6:193, P ¼ 0:003) indicat- was the N150/P260 peak to peak amplitude as a general
ing significantly enhanced HR levels in the extinction as indicator of pain processing. These data suggest that it
compared to the habituation phase (tð58Þ ¼ 23:332, may be the early processing of pain reflecting sensory rather
P ¼ 0:002). than later cognitive-evaluative processes that are influenced
by operant learning. Operant conditioning of components
earlier than 100 ms might give additional information on
4. Discussion the stages of processing of pain responses.
The analysis of the peripheral measures suggests that the
It can be concluded from the data that pain reports in both altered verbal pain reports do not generalize to general indi-
healthy controls and chronic back pain patients can be cators of pain and arousal such as heart rate, skin conduc-
brought under operant control. Whereas the healthy controls tance or muscle tension. However, the muscle tension data
showed fast extinction, the chronic back pain patients main- confirm previous reports of stress-related muscle tension
tained their elevated pain ratings throughout the extinction increases in chronic back pain patients (Flor et al., 1992;
phase and thus counteracted the general habituation of the Moulton and Spence, 1992). Only the patients responded
pain ratings that was present in the data. The early SEP- with prolonged increased muscle tension levels to the oper-
component of the chronic pain patients was generally ant conditioning task. The stronger response of the left side
elevated and the pain threshold was reduced. This may of the body reflects (a) that the painful stimulus was applied
reflect an enhanced reactivity to painful stimuli in chronic to the left finger and (b) confirms previous reports of more
back pain patients as shown in previous MEG and EEG pronounced pain responses of the right cerebral hemisphere
studies of our group (e.g., Flor and Birbaumer, 1994; Flor (Brennum et al., 1989; Jensen et al., 1992). The increase of
et al., 1994; Flor et al., 1997a,b) indicating cortical hyper- heart rate in the extinction phase for both the patients and
reactivity to painful electric stimulation in chronic pain the healthy controls may reflect the uncertainty induced by
patients. If operant conditioning of overt pain responses the removal of reinforcement (Kelsey, 1991).
leads to the formation of new cortical and subcortical cell Taken together, the results of this study show that the
assemblies (e.g., Flor and Birbaumer, 1994; Flor et al., operant conditioning of verbal pain reports leads to delayed
1994), then pain-related EEG responses should reflect this extinction in chronic pain patients and generalizes, more-
conditioning process as well. over, to the cortical level. The lack of significant correlation
The up-down differentiation of the verbal reports was, between resistance to extinction and chronicity suggests that
however, not mirrored in the N150 component during acqui- the prolonged operant conditioning effects might precede,
sition although this effect was overlaid by a general habitua- not follow, the onset of chronic pain and thus constitute a
tion of the N150 component. The significant decrease of predisposing factor.
N150 during down training and the lack of habituation of
the N150 in the up training may be viewed as indicative of a
conditioning process. However, compared to subjective Acknowledgements
pain ratings, the cortical pain response showed delayed
conditioning with the up/down differentiation occurring This study was supported by grants of the Deutsche
only in the fourth as compared to the first acquisition Forschungs-gemeinschaft to N.B. and H.F.
block. During extinction, N150 remained elevated in the
CBP patients indicating that the conditioning effect of the References
chronic back pain patients remained present on the cortical
level. N150 has been related to activation of the anterior Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J. An inventory for
cingulate (Dowman, 1994a,b; Dowman and Darcey, 1994) measuring depression. Archives Gen Psychiatry 1961;56:561–571.
118 H. Flor et al. / Pain 95 (2002) 111–118

Brennum J, Kjeldsen M, Jensen K, Jensen TS. Measurements of human Fordyce WE. Behavioral concepts in chronic pain and illness. In: Davidson
pressure-pain thresholds on fingers and toes. Pain 1989;38:211–217. PO, editor. The behavioral treatment of anxiety, depression and pain,
Bromm B. Pain measurement in man-neurophysiological correlates of pain, New York: Bruner and Mazel, 1976. pp. 147–188.
Amsterdam: Elsevier, 1984. Fowles DC, Christie MJ, Edelberg R, Grings WW, Lykken DT, Venables
Bromm B, Meier W. The intracutaneous stimulus: A new pain model for PH. Publication recommendations for electrodermal measurements.
algesimetric studies. Methods Findings Exp Clin Pharmacol Psychophysiology 1981;18:232–239.
1984;6:405–410. Fridlund AJ, Cacioppo JT. Guidelines for human electromyographic
Chapman CR, Jacobson RC. Assessment of analgesic states: Can evoked research. Psychophysiology 1986;23:567–589.
potentials play a role. In: Bromm B, editor. Pain measurement in man. Gratton G, Coles MM, Donchin A. A new method of off-line removal of
Neurophysiological correlates of pain, Amsterdam: Elsevier, 1984. pp. ocular artifact. Electroencoph Clin Neurophysiol 1983;55:468–484.
233–256. Jasper HH. The ten-twenty electrode system of the International Federation.
Chen ACN, Chapman CR, Harkins SW. Brain evoked potentials are func- J Electroenceph Clinical Neurophysiol 1958;20:371–375.
tional correlates of induced pain in man. Pain 1979;6:365–374. Jennings JR, Wood CC. The e-adjustment procedure for repeated-measures
Desmedt JE, Tomberg C. Mapping early somatosensory evoked potentials analyses of variance. Psychophysiology 1976;13:277–278.
in selective attention: Critical evaluation of control conditions used for Jensen R, Rasmussen BK, Pedersen B, Lous I, Olesen J. Cephalic muscle
titrating by difference the cognitive P30, P40, P100 and N140. Electro- tenderness and pressure pain threshold in a general population. Pain
enceph Clin Neurophysiol 1989;74:321–346. 1992;48:197–203.
Dowman R. SEP topographies elicited by innocuous and noxious sural Kelsey RM. Electrodermal lability and myocardial reactivity to stress.
nerve stimulation. I. Identification of stable periods and individual Psychophysiology 1991;28:619–631.
differences. Electroenceph Clin Neurophysiol 1994a;92:291–302. Kerns RD, Turk DC, Rudy TE. The West Haven-Yale Multidimensional
Dowman R. SEP topographies elicited by innocuous and noxious sural nerve Pain Inventory (WHYMPI). Pain 1985;23:345–356.
stimulation. II. Effects of stimulus intensity on topographic pattern and Linton SJ, Götestam KG. Controlling pain reports through operant condi-
amplitude. Electroenceph Clin Neurophysiol 1994b;92:303–315. tioning: A laboratory demonstration. Percep Motor Skills 1985;60:427–
Dowman R. Effects of operantly conditioning the amplitude of the P200 437.
peak of the SEP on pain sensitivity and the spinal nociceptive with- Lippold OCJ. Electromyography. In: Venables PH, Martin I, editors. A
drawal reflex in humans. Psychophysiology 1996;33:252–261. manual of psychophysiological methods, Amsterdam: North Holland,
Dowman R, Darcey TM. SEP topographies elicited by innocuous and 1967. pp. 245–291.
noxious sural nerve stimulation. III. Dipole source localization analysis.
Lousberg R, Groenman NH, Schmidt AJ, Gielen AA. Operant conditioning
Electroencephal Clin Neurophysiol 1994;92:373–391.
of the pain experience. Percep Motor Skills 1996;83:883–900.
Flor H. Psychobiologie des Schmerzes. Empirische Untersuchungen zur
Miltner W, Larbig W, Braun C. Attention and event-related potentials
Psychophysiologie, Diagnostik und Therapie chronischer Schmerzsyn-
elicited by intracutaneous electrical stimulation of the skin. J Psycho-
drome der Skelettmuskulatur, Bern: Huber, 1991.
physiol 1988a;2:269–276.
Flor H, Birbaumer N. Acquisition of chronic pain: Psychophysiological
Miltner W, Larbig W, Braun C. Biofeedback of somatosensory event-
mechanisms. Am Pain Soc J 1994;32:119–127.
related potentials: Can individual pain sensations be modified by
Flor H, Turk DC. Chronic back pain and rheumatoid arthritis: Predicting
biofeedback-induced self-control of event-related potentials? Pain
pain and disability from cognitive variables. J Behav Med
1988b;35:205–213.
1988;11:251–265.
Moulton B, Spence SH. Site-specific muscle hyper-reactivity in musicians
Flor H, Birbaumer N, Turk DC. The psychobiology of chronic pain.
with occupational upper limb pain. Behav Res Ther 1992;30:375–386.
Advances Behavior Res Ther 1990;12:47–84.
Pivik RT, Broughton RJ, Coppola R, Davidson RJ, Fox N, Nuwer MR.
Flor H, Birbaumer N, Schugens MM, Lutzenberger W. Symptom-specific
psychophysiological responses in chronic pain patients. Psychophysiol- Guidelines for the recording and quantitative analysis of electroence-
ogy 1992;29:452–460. phalographic activity in research contexts. Psychophysiology
Flor H, Behle DJ, Birbaumer N. Assessment of pain-related cognitions in 1993;30:547–558.
chronic pain patients. Behaviour Res Ther 1993;31:63–73. Raichle ME. Images of the mind: Studies with modern imaging techniques.
Flor H, Birbaumer N, Fürst M. Pavlovian conditioning in chronic pain: Annu Rev Psychol 1994;45:333–356.
Dissociation between peripheral and central responses. Psychophysiol- Rainville P, Duncan GH, Price DD, Carrier B, Bushnell MC. Pain affect
ogy 1994;31:47. encoded in human anterior cingulate but not somatosensory cortex.
Flor H, Braun C, Elbert T, Birbaumer N. Extensive reorganization of Science 1997;277:968–971.
primary somatosensory cortex in chronic back pain patients. Neurosci Rosenfeld JP, Silvia R, Weitkunat R, Dowman R. Operant control of human
Lett 1997a;224:5–8. somatosensory evoked potentials alters experimental pain perception.
Flor H, Knost B, Birbaumer N. Processing of pain-related information in In: Fields HL, Dubner R, Cervero F, editors. Advances in pain research
chronic pain patients: electrocortical and peripheral correlates. Pain and therapy, New York: Raven Press, 1985. pp. 343–349.
1997b;73:413–421. Zerssen D von. Die Beschwerden–Liste. Weinheim: Beltz, 1976.