Differences in pain-related fear acquisition and generalization: An experimental study comparing...

Research Paper

Differences in pain-related fear acquisition andgeneralization: an experimental study comparingpatients with fibromyalgia and healthy controlsAnn Meuldersa,b,*, Anne Jansa, Johan W. S. Vlaeyena,b,c

AbstractAnomalies in fear learning, such as failure to inhibit fear to safe stimuli, lead to sustained anxiety, which in turn may augment pain. Inthe same vein, stimulus generalization is adaptive as it enables individuals to extrapolate the predictive value of 1 stimulus to similarstimuli. However, when fear spreads in an unbridled way to novel technically safe stimuli, stimulus generalization becomesmaladaptive and may lead to dysfunctional avoidance behaviors and culminate in severe pain disability. In a voluntary movementconditioning paradigm, we compared the acquisition and generalization of pain-related fear in patients with fibromyalgia (FM) andhealthy controls. During acquisition, participants received predictable pain in 1 context (ie, 1 movement predicts pain, whereasanother does not), and unpredictable pain in another (ie, pain never contingent upon movement). Fear generalization to novelmovements (resembling the original painful or nonpainful movement) was tested in both contexts. Results indicated that the FMgroup showed slower differential acquisition of pain-related fear in the predictable context, and more contextual pain-related fear inthe unpredictable context. Fear of movement-related pain spreads selectively to novel movements similar to the original painfulmovement, and not to those resembling the nonpainful movement in the healthy controls, but nondifferential fear generalization wasobserved in FM. As expected, in the unpredictable context, we also observed nondifferential fear generalization; this effect wasmorepronounced in FM. Given the status of overgeneralization as a plausible transdiagnostic pathogenic marker, we believe that thisresearch might increase our knowledge about pathogenesis of musculoskeletal widespread pain.

Keywords: Pain-related fear, Fear conditioning, Fear generalization, Voluntary movement paradigm, Fibromyalgia, Learning

1. Introduction

Fibromyalgia (FM) syndrome is a chronic widespread paincondition characterized by high comorbidity with anxiety,fatigue, cognitive impairment, and depression.55 The relation-ship between FM and anxiety is particularly firm,1 and evidencesuggests that anxiety sensitivity often paves the way for theonset of chronic musculoskeletal pain.2 Specifically, pain-related fear is considered to play a pivotal role in the originand maintenance of chronic pain disability,15,60,61 a notion thatis steadily receiving support.21,33,56,57 Associative learningmechanisms are shown to be crucially involved in developmentof pain-related fear.42 For instance, in a pain-related fearconditioning procedure, 1 neutral movement (conditionedstimulus [CS1]) that is paired with pain (unconditioned stimulus,pain-[US]) starts to elicit fear and avoidance, whereas anothermovement (CS2) that is never paired with the pain-US does not.

Fear learning has an important adaptive advantage, the ability toidentify cues that signal threat (eg, pain) can initiate appropriatedefensive responses (eg, avoidance) that protect us from harm,whereas identifying safety signals allows for goal-directedapproach behaviors. Anomalies in fear learning, such as failureto inhibit fear to safety signals (CS2), lead to sustained,generalized anxiety,3,14,24,45 which in turn may augmentpain.43,49 Moreover, a meta-analysis on fear conditioningstudies39 and recent experimental findings20,28,34,41 suggestthat pathological anxiety is associated with impaired inhibitionrather than disproportionate fear to danger signals (CS1).

Even adaptive learners face the challenge of how to deal withvariations in the appearances of signaling stimuli. Stimulusgeneralization22,31 is an adaptive mechanism as it enablesindividuals to extrapolate the predictive value of 1 stimulus tosimilar stimuli and minimizes the necessity to learn everythinganew. However, unbridled (over)generalization may lead towidespread fear, dysfunctional avoidance behaviors, and culmi-nate in severe pain disability.38,40 Using a voluntary joystickmovement paradigm,42 we recently tested fear generalizationtoward novel movements in 2 different contexts. In the predict-able context, 1 arm movement (eg, left) was followed by painfulelectrical stimulation and another was not (eg, right). In theunpredictable context, the pain-US was never paired with themovements (eg, upward/downward) but was presented duringthe context (intertrial interval [ITI]). As expected, healthy individ-uals were more afraid of the movements that resembled theoriginal CS1 than those resembling the CS2 (ie, differential feargeneralization) in the predictable context. No such difference

Sponsorships or competing interests that may be relevant to content are disclosed

at the end of this article.

a Research Group on Health Psychology, University of Leuven, Leuven, Belgium,b Center for Excellence on Generalization Research in Health and Psychopathology,

University of Leuven, Leuven, Belgium, c Department of Clinical Psychological

Science, Maastricht University, the Netherlands

*Corresponding author. Address: Department of Psychology, University of Leuven,

Tiensestraat 102, Box 3726, 3000 Leuven, Belgium. Tel.: 132 (0)16 32 60 38;

fax:132 (0)16326144.E-mail address: [email protected] (A.Meulders).

PAIN 156 (2015) 108–122

© 2014 International Association for the Study of Pain

http://dx.doi.org/10.1016/j.pain.0000000000000016

108 A. Meulders et al.·156 (2015) 108–122 PAIN®

Copyright � 2014 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.

occurred in the unpredictable context, but a nondifferentialincrease in fear was observed in this context.44

In this study, we used the same approach to compare theacquisition and generalization of both cued and contextual pain-related fear in a chronic pain population (FM) with healthycontrols (HC). Investigating (over)generalization—possiblya transdiagnostic marker of pathology—might further ourunderstanding of the spreading of pain-related fear and painin widespread musculoskeletal pain. We predicted (1) slower/less (adaptive) differential cued pain-related fear acquisition dueto impaired safety learning, (2) more contextual pain-related fearduring acquisition and generalization, (3) less differential(adaptive) fear generalization, and (4) lower pain thresholds inFM than in HC.

2. Methods and materials

2.1. Participants

This study used a convenience sample of 48 female participantsincluding 2 age-matched diagnostic groups: 24 FM patients(mean 6 SD [range] age 5 39 6 13 [19-60] years) and 24 HC(mean6 SD [range] age5 386 13 [18-61] years). Note: we didnot use absolute age-matched groups, but 5-year ranges tomatch the HC to the FM group.We do not think that the capacityto give verbal fear ratings or the fear-potentiated startlemeasures would significantly differ within the proposed ageranges. The most important inclusion criterion for the FM groupwas to be diagnosed with FM and experiencing some inter-ference in their daily life because of this condition. All patientssatisfied the American College of Rheumatology new diagnosticcriteria for FM66 based on the combinedWidespread Pain Index(range, 0-19) and Symptom Severity Score (range, 0-12)

(Table 1). The inclusion criterion for the HC group was to nothave FM. Exclusion criteria for both groups were the following:any other chronic pain conditions, diagnosed dyslexia oranalphabetism, pregnancy, current or history of cardiovasculardisease, chronic or acute respiratory disease (eg, asthma,bronchitis), neurological diseases (eg, epilepsy), uncorrectedhearing problems, having pain at the dominant hand, wrist, orarm that hinders to move a joystick painlessly, cardiac pacemakeror the presence of any other electronic medical devices, and thepresence of any other severe medical conditions. An additionalexclusion criterion only for the HC group was the following: anycurrent or past psychiatric disorder including clinical depressionand panic or other anxiety disorders. Because of its high com-orbiditywith depression, othermooddisorders, and anxiety,35,58,65

this additional criterion was omitted in the FM group. Patients withFMwere recruited from pain clinics in the regions Flemish-Brabantand Limburg (Belgium) and through a call for participants on theWeb site of the Flemish League for FM Patients. Following theprocedure of Schneider et al.,51 patients were asked to bringa physically and mentally healthy friend of their own age (range) tothe study. The experimental protocol was approved by theEthical Committee of the Faculty of Psychology and EducationalSciences of the University of Leuven (registration number:S-55100) and the Medical Ethical Committee of the UniversityHospital of the University of Leuven (registration number:ML9402). All participants signed the informed consent form,which explicitly stated that they were allowed to declineparticipation at any time during the experiment. As expected,patients with FM had lower educational level, were more likely tobe unemployed and were taking more medication than the HC.More detailed demographic and clinical characteristics can befound in Table 1.

Table 1

Demographic and clinical characteristics for the FM group (n 5 24) and the HC (n 5 24) group separately.

Total N 5 48

FM group HC group

t dfMean SD Mean SD

Selected pain intensity level, mA* 26.46 11.91 37.46 14.70 2.85 46

Selected self-reported pain intensity

(range, 1-10)

8.13 0.61 8.38 0.71 1.31 46

Age, y 39.29 13.06 38.38 13.50 20.24 46

WPI (range, 0-19)† 11.08 3.56 1.46 1.56 12.13 46

SS (range, 0-12)† 8.04 2.01 2.42 1.44 11.14 46

Highest education level, %

Primary school 4 0

Vocational secondary education 4 4

Technical secondary education 42 13

General secondary education 21 13

Bachelor’s degree 17 54

Master’s degree 12 21

Occupation, %

Working 25 71

Studying 12 25

Unemployed/invalidity allowance/retired 63 4

Medication (yes–no) 88 12

Type of medication, %

Antidepressants 50 0

Anxiolytics 17 0

Analgesics 92 8

Other 67 0

Pain treatment (yes–no) 100 29

* P , 0.01.

† P , 0.001.

FM, fibromyalgia; HC, healthy controls; SS, Symptom Severity Score: 0 5 no symptoms, 12 5 very much pain symptoms. Other medication includes sleep medication, muscle relaxants, hormones, antihypertension,

antiarrhythmic, proton-pump inhibitor, leukotriene receptor antagonist, and prevention for migraine medication; WPI, Widespread Pain Index: the higher the score, the more pain complaints on different sites of the body during

the past week.

January 2015·Volume 156·Number 1 www.painjournalonline.com 109


2.2. Pre-experimental measures

Before data collection, all participants completed a battery ofquestionnaires using a Web survey tool. The scores on thesequestionnaires can be found in Table 2. (1) Pain severity: TheChronic Pain Grade Scale63 assesses pain intensity andinterference with normal daily activities using 7 items (eg,How would you rate your pain at this moment?). Answers on 6of the 7 items range from 0 to 10, indicating “no pain” and “painas bad as it could be,” respectively. The 1 remaining itemrequires filling in the number of days that pain has keptrespondents from their typical daily activities in the last 6months (range, 0-180). Based on the pain intensity score, thedisability points (based on the disability score and the days ofdisability) of the respondents are classified in 4 grades ofchronic pain: grade I, low disability–low intensity; grade II, lowdisability–high intensity; grade III, high disability–moderatelylimiting; and grade IV, high disability–severely limiting. (2) Paincognitions: the Pain Cognition List59 consists of 39 itemsdivided into 5 subscales (Catastrophizing, Limitation, Opti-mism, Internal control, and Trust). Each item presents a specificpain cognition statement (eg, “My thoughts are always

concentrated on the pain”) and the respondent is asked toindicate (dis)agreement on a 5-point Likert scale. Items arescored from 1, “totally disagree,” to 5, “totally agree,” and a sumscore is obtained per subscale (Catastrophizing: range, 16-80;Limitation: range, 7-35; Optimism: range, 7-35; Internalcontrol: range, 5-25; Trust: range, 4-20). (3) Fear of movement:the Tampa Scale of Kinesiophobia50 comprises 17 itemsintended to assess fear of movement and fear of (re)injury.Respondents are asked to indicate to what extent each of thestatements (eg, “My body tells me that there is something

seriously wrong with it”) reflects a true description of theassumed association between movement and (re)injury ona 4-point Likert scale, ranging from 1, “strongly disagree,” to 4,“strongly agree,” with the total score ranging from 17 to 68. (4)Pain disability: the Fibromyalgia Impact Questionnaire (FIQ)11

assesses the impact of FM on the respondent’s daily activities.The FIQ is composed of 10 items. The first item contains 11

questions (eg, “Can you independently do the dishes”) relatedto physical functioning, each question is rated on a 4-pointLikert type scale. Items 2 and 3 ask the respondent to mark thenumber of days they felt well and the number of days they wereunable to work (including housework) because of painsymptoms. Items 4 through 10 are horizontal linear scalesmarked in 10 increments on which the respondent rates workdifficulty, pain, fatigue, morning tiredness, stiffness, anxiety,and depression. After normalization, the total score ranges from0 to 100, with 0 indicating no impairment at all and 100maximum impairment. (5) Affect: the trait version of the Positiveand Negative Affect Schedule18,64 consists of 20 items dividedinto 2 subscales. Participants are asked to indicate to whatextent, in their normal daily life, they experience the feelingsdefined by the 20 descriptors using a 5-point response scaleranging from “very little” to “a lot.” Ten items describe positivefeelings and assess positive affectivity (range, 10-50), and 10items describe negative feelings and assess negative affectivity(range, 10-50). (6) Depression and anxiety: the Hospital AnxietyDepression Scale53,67 consists of 14 items divided into 2subscales (Anxiety and Depression). Respondents are asked toindicate for each item (eg, “I still enjoy the things I used to enjoy”)which answer reflects best how they felt during last week.Answers are scored from 0 to 3. The scores for the Depressionsubscale and the Anxiety subscale range from 0 to 21.

2.3. Stimulus material and experimental measures

The experiment was run on a Windows XP computer (DellOptiplex 755) with 2GBRAMand an Intel Core2Duoprocessor at2.33 GHz and an ATI Radeon 2400 graphics card with 256 MB ofvideo RAM, using Affect 4.0.54 We used 4 proprioceptive stimuli(ie, moving a Logitech Attack 3 joystick in the horizontal plane[left/right] and in the vertical plane [upward/downward]) as CSs.The generalization stimuli (GSs) were diagonal movements(ie, left–top, right–top, left–bottom, right–bottom). An electro-cutaneous stimulus (2-millisecond duration) delivered by a com-mercial constant current stimulator (DS7A, Digitimer, WelwynGarden City, England) served as the pain-US. This stimulation

Table 2

Questionnaire scores for the fibromyalgia (FM) group and healthy controls (HC) group separately.

Total sample N 5 48

FM group (n 5 24) HC group (n 5 24)

t dfMean SD Mean SD

CPGS—pain intensity* 70.00 11.38 19.44 14.80 13.27 46

CPGS—pain disability* 60.97 19.72 5.69 7.26 12.89 46

CPGS—no. of days disability* 76.67 73.52 1.08 2.55 5.03 46

PCL—catastrophizing* 47.17 13.32 33.54 10.42 3.95 46

PCL—limitation* 28.00 3.60 17.42 4.37 9.15 46

PCL—optimism* 22.96 5.77 28.29 4.50 23.57 46

PCL—internal control† 17.00 3.53 19.63 2.95 22.80 46

PCL—trust‡ 14.00 2.25 15.54 2.83 22.09 46

TSK—total score* 38.88 6.70 31.46 6.71 3.83 46

FIQ—total score* 61.50 15.34 16.53 8.95 12.40 46

PANAS—positive affect* 30.25 7.81 37.96 5.03 24.06 46

PANAS—negative affect† 25.54 8.63 18.50 5.53 3.37 46

HADS—anxiety* 9.75 4.90 4.29 2.29 4.94 46

HADS—depression* 6.92 4.37 1.83 2.44 4.97 46

Based on the CPGS scales, 2% (1/24) of patients with FM were classified as grade I (low disability–low intensity), 8% (2/24) as grade II (low disability–high intensity), 38% (9/24) as grade III (high disability–moderately limiting),

and 50% (12/24) as grade IV (high disability–severely limiting).

* P , 0.001.

† P , 0.01.

‡ P , 0.05.

CPGS, Chronic Pain Grade Scale: pain intensity (items 1-3), pain disability (items 4-6), and days of disability (item 7); FIQ, total score on the Fibromyalgia Impact Questionnaire; HADS, Hospital Anxiety Depression Scale:

subscales are calculated for anxiety and depression; PANAS, Positive and Negative Affect Schedule: subscales are calculated for positive affect and negative affect; PCL, Pain Cognition List: subscales are calculated for

catastrophizing, limitation, optimism, internal control, and trust; TSK, total score on the Tampa Scale of Kinesiophobia.



was administered through surface SensorMedics electrodes(8 mm) filled with K-Y gel that were attached to the wrist of thedominant hand. During the calibration procedure, participantsreceived a series of electrocutaneous stimuli of increasingintensity and were asked to indicate how intense or painful eachstimulus was on a scale of 1 to 10, where “1” means: “you feelsomething but this is not painful, it is merely a sensation”; “2”means: “this sensation starts to be painful, but it is still a very

moderate pain”; and up to “10,” which means: “this is the worstpain you can imagine.” A subjective stimulus intensity of “8,”which refers to a stimulus that is “significantly painful and

demanding some effort to tolerate,” was targeted. We useda standard protocol to increase the intensity level at each step ofthis calibration procedure. For all participants, the first electro-cutaneous stimulus presentation had an intensity of 1 mA, the nextstimulus presentations, respectively, had an intensity of 2 mA, 4mA, 6mA, and8mA, then the stepswere increased from2mA to 3mA increases between each stimulus presentation (until 20 mA),finally the steps are again increased from 3 mA to 4 mA increasesbetween each stimulus presentation, until the participant indicatesthat the stimulus is “significantly painful anddemanding someeffortto tolerate” which corresponds to an 8 on the 11-point intensityscale. Themean subjective stimulus intensitywas 8.13 (SD5 0.61,range5 7-10) for the FMgroup and 8.38 (SD5 0.71, range5 6-9)for the HC group. Themean physical stimulus intensity (in mA) was26.46 (SD 5 11.91, range 5 6-48) for the FM group and 37.46(SD 5 14.70, range 5 14-68) for the HC group.

Conditioned pain-related fear was measured through self-reports and a psychophysiological measure of fear learning, thatis, the eyeblink startle response. The eyeblink startle response isa component of the reflexive cross-species, full-body defensiveresponse mobilization that is triggered by startle-evoking stimuli(eg, acoustic startle probe) and can bemeasured by the tension inthe muscles underneath the eye. Startle modulation refers to thepotentiation of the startle reflex during fear states elicited byanticipation of an aversive stimulus (eg, an electrocutaneousstimulus). In the present setup, the startle probe was a 100 dBAburst of white noise with instantaneous rise time presentedbinaurally for 50 milliseconds through headphones (PhilipsSHP2500). Eyeblink startle responses elicited by startle probesdelivered during the CS/GS movements served as an index ofcued pain-related fear. Eyeblink startle responses elicited bystartle probes during the ITI served as an index of contextual pain-related fear.60

2.4. Procedure

The procedure is largely the same as described by Meulders andVlaeyen.44 Important procedural adjustments were that (1) theoriginal CS movements were also performed during generaliza-tion, with reinforced CS1 movements. This procedure wasfollowed to prevent extinction and to facilitate the transfer of theoriginal acquisition task setting as much as possible to thegeneralization test, (2) we did not include latency measures asa proxy for avoidance behavior because we expected that FMwould be overall slower for several reasons (eg, medication,impaired working memory, attentional bias), which would make itdifficult to interpret these data. The experiment was conductedduring a 2-hour session and consisted of a preparation phase,a practice phase, a habituation phase, an acquisition phase,a transfer of acquisition phase, and a generalization phase. In amixed design (Table 3), both participants of the FMgroup and theHC group (between-subjects factor) received (1) predictable painstimuli in 1 context and (2) unpredictable pain stimuli in anothercontext (within-subjects factor). Half of the participants in eachgroup moved the joystick in the horizontal plane (left/right) in thepredictable context, and in the vertical plane (upward/downward)in the unpredictable context. The other half of the participants ineach group received the reverse combination. In the predictablecontext, 1 movement (CSp

1) was consistently followed by thepain-US and the other movement (CSp

2) was never followed bythe pain-US. Note that the direction of joystick movement thatserved as the CSp

1 and the CSp2 in the predictable context was

counterbalanced across participants as well. In the unpredictablecontext, however, the pain-USwas never delivered contingent oneither of the joystick movements (CSu1 and CSu2) but waspresented during the context (ITI). During acquisition, participantsfreely chose on each trial in which direction they were going tomove the joystick. During the transfer of acquisition, however,they could no longer choose the order of the movementsthemselves, but the movement direction was signaled. Duringgeneralization, the same signaling procedure was used to testthe spreading of conditioned fear to novel diagonal move-ments (GSs).

2.4.1. Preparation phase

On arrival to the laboratory, participants were informed that theexperiment involved repeated presentation of electrocutaneousstimuli (pain-US) and short loud noises (acoustic startle probes).

Table 3

Summary of the experimental design.

Context Practice Habituation Acquisition Transfer of acquisition Generalization

Predictable 4 3 CSp1 only 6 probes 12 3 CSp

1 4 3 CSp1 4 3 CSp

1

4 3 CSp2 12 3 CSp

2 4 3 CSp2 4 3 CSp

2

2 3 GSp1u1

2 3 GSp2u1

2 3 GSp1u2

2 3 GSp2u2

Unpredictable 4 3 CSu1 6 probes 12 3 CSu1 4 3 CSu1 4 3 CSu14 3 CSu2 12 3 CSu2 4 3 CSu2 4 3 CSu2

12 3 US (during ITI) 4 3 US (ITI) 2 3 GSp1u1

2 3 GSp2u1

2 3 GSp1u2

2 3 GSp2u2

CSp1 5 movement in predictable condition that was consistently followed by the painful stimulus (US); CSp

2 5 movement in predictable condition that was never followed by the painful stimulus (US); CSu1 and CSu1 5movements in unpredictable context that were never followed by the painful stimulus (US). GSs5 novel diagonal movements that share proprioceptive characteristics with the original CSp

1 (GSp1u1 and GSp

1u2) or the original

CSp2 (GSp

2u1 and GSp

2u2). For both experimental groups (FM and HC), the predictable and unpredictable pain conditions were run within subjects in blocks of 8 movements in a semirandomized order (no more than 2

consecutive blocks could be in the same context). GS movements are never reinforced. The suffix “only” is used to indicate nonreinforcement of the CSp1 movement (ie, during the practice phase).

CS, conditioned stimulus; GS, generalization stimulus; US, unconditioned stimulus.



Participants were also told that they were free to declineparticipation at any time without any negative consequences.After providing informed consent, electrodes for eyeblink startleresponses and the administration of the electrocutaneousstimulus were attached and the calibration procedure of thepain-US was initiated (see 2.3. Stimulus material and experimen-tal measures section).

2.4.2. Practice phase

Before starting the practice phase, participants received exten-sive written instructions about the experimental task. In eachblock, participants were requested to move the joystick 8 timesas quickly and accurately as possible when prompted bya starting signal “1” (fixation cross presented in the middle ofthe computer screen), in whatever order they freely chose. Theposition of counter bars on the computer screen indicated inwhich movement plane (horizontal vs vertical) they were to move.The counter bars, each divided into 4 equal segments, alwaysappeared on 2 sides of the computer screen (left/right or top/bottom). In a horizontal block, these bars were displayed on theleft side and right side of the computer screen, whereas in thevertical block, these bars appeared at the top and the bottom ofthe computer screen. Successful movements always resulted incoloring 1 segment of the corresponding counter bar blue. Thus,participants could instantly ascertain how many movements ineach direction still were to be performed. In total, 2 blocks of 8trials were run: 1 block (4 left/right) in the horizontal plane and 1block (4 upward/downward) in the vertical plane or vice versa.During the practice phase, no acoustic startle probes or pain-USswere presented, and the experimenter gave online verbalfeedback about the task performance.

2.4.3. Startle habituation phase

Because it is expected that the first responses to the startleprobes are significantly larger than the latter ones, we inserteda startle probe habituation phase to correct for such possibleconfound in the data collection. This phase consisted of 12 trials,each lasting 24 seconds (with a variable ITI of on average 5seconds [62 seconds]). During each trial, 1 startle probe waspresented either between the first and the second second(3 trials), at the 10th second (2 trials), or between the 15th and the17th second (3 trials) after trial-onset. During this phase, theparticipants wore headphones, and the lights in the experimentalroomwere dimmed. Note that no pain-USs were delivered duringthis phase.

2.4.4. Acquisition phase

This phase was basically the same as the practice phase (Fig. 1),but now, (1) pain-USs and startle probes were presented, (2) 6blocks (3 predictable and 3 unpredictable) of 8 trials were runinstead of 2 blocks, and (3) instructions emphasized to pay closeattention to the starting signal “1” and to respond as fast andaccurately as possible upon its presentation. Although a CSmovement was of variable length depending on the participants’movement speed, a trial typically included an ITI consisting ofa pre-CS interval of 3.5 seconds and a post-CS interval of 8seconds. The pain-US was presented within each CSp

1 trial inthe predictable context (100%contingency) and in half of the trialsduring the ITI in the unpredictable context. In the predictablecontext, the US occurred immediately after the CSp

1movement.In the unpredictable context, the pain-US occurred between 1and 3 seconds before the presentation of the starting signal “1”,

or between 0.5 and 2.5 seconds after the CSmovement. In eachblock with 8 CS movements, 4 of the startle probes werepresented during theCSmovements (2 duringCSp

1 or CSu1, andthe other 2 during CSp

2 or CSu2), and 4 during the ITI (2 probepositions before the CS and the other 2 probe positions after theCS). To avoid that startle facilitation during the ITI would beconfounded by direct responses to presentations of the pain-USin the unpredictable context, startle probes were presentedbefore the CSwhen the pain-USwas presented after the CS, andafter the CS when the pain-US was presented before the CS.Note that we did not inform the participants about thecontingencies between the joystick movements (CSs) and thepain-US. After each conditioning block, the participants ratedthe cued pain-related fear elicited by each of the CSmovements.

2.4.5. Transfer of acquisition

Transfer of acquisition trials were identical as those during theacquisition phase. Yet, CSmovements were no longer voluntarilyinitiated but signaled. That is, 3000 milliseconds after trial onset,a red asterisk “*” appeared for 500 milliseconds at 1 of the CSmovement directions (left/right in a horizontal block; top/bottomin a vertical block), indicating in which direction participants wererequested to move. Before actually performing the signaledmovement, participants rated their US-expectancy and cuedpain-related fear. After completing the ratings, they waited forthe “1” starting signal to start moving into the signaled direction.After successfully performing the signaled CSmovement, a post-CS-ITI of 8 seconds followed (compare with timing acquisition).During the transfer phase, 2 blocks (1 predictable and 1unpredictable) of 8 trials were run. Startle probes presentationfollowed the same timing schedule as during acquisition.

2.4.6. Generalization phase

The procedure of the generalization phase was mainly the sameas the transfer of acquisition phase. The difference was thatparticipants now had to perform 4 novel diagonal movements(GSs), which each have either a feature in common with theoriginal CSp

1 or the original CSp2, namely left–top, right–top,

left–bottom, or right–bottom. Again, 3000 milliseconds after thetrial onset, a red asterisk “*” appeared for 500 milliseconds in 1 ofthe corners of the screen to signal which movement had to beperformed in a randomized order after answering the US-expectancy and cued pain-related fear questions. After success-fully performing amovement, a post-GS-ITI of 8 seconds followed(compare with timing acquisition). On each trial, a startle probewas delivered during the GS movement, the original CS1

movement was still reinforced as during the previous phases,but none of the GSs was followed by the pain-US. During thisphase, no ITI probes were delivered. The 4 GS movements wereperformed 2 times in each experimental context (predictable vsunpredictable), the order was randomized across participants.

2.5. Manipulation checks and outcome variables

2.5.1. Manipulation checks

2.5.1.1. Contingency awareness

As a manipulation check, we assessed participants’ contingencyawareness after the entire experiment using a retrospective US-expectancy rating. Participants indicated for both CSmovementsin each context how much they expected the painful stimulus tooccur on an 11-point Likert scale (range, 0-10) with labels “not atall” to “very much.”



2.5.1.2. Unconditioned stimulus (US)-expectancy duringtransfer of acquisition

During the transfer of acquisition phase, participants indicatedbefore each movement to what extent they expected the painfulstimulus to occur when performing the “signaled” movements(CSs) on an 11-point Likert scale (range, 0-10) with labels “not at all”to “very much.” Thus, we could assess whether the contingencieslearned during the acquisition phase using the voluntarymovementsetup transfer to the signaled movement setup.

2.5.2. Self-reported fear of movement-related pain

After each block, participants answered the following question:“How afraid were you to perform the left/right/upward/downwardmovement?” on an 11-point Likert scale ranging from 0 to 10 withanchors “not fearful at all” to “the worst fear you can imagine.” Notethat these questions applied to the previous block only (ie, after

a horizontal block ratings for the left/right movements and aftera vertical block ratings for the upward/downward movements).During the transfer of acquisition and generalization phases,participants rated before each movement how afraid they were toactually perform the “signaled” movements (CSs/GSs).

2.5.3. Unconditioned stimulus (US)-expectancy duringgeneralization

During the generalization phase, participants rated before eachmovement towhat extent they expected thepainful stimulus to occurwhen performing the “signaled” movements (CSs/GSs) on an 11-point Likert scale (range, 0-10) with labels “not at all” to “very much.”

2.5.4. Eyeblink startle modulation

Orbicularis Oculi electromyographic activity (EMG) was recordedwith 3 Ag/AgCl SensorMedics electrodes (4 mm) filled with

Figure 1. Schematic overview of the experimental task and trial timing during the acquisition phase for both the predictable context (A) and the unpredictablecontext (B) and during the generalization phase for both predictable context (C) and the unpredictable context (D). Note: the white “1” serves as the starting signalto initiate the voluntary movements during the acquisition phase (A and B). During the transfer of acquisition phase and the generalization phase (panel C and D),the order of themovements is no longer chosen freely, a red “*”indicates which CS/GS (diagonal) movement ought to be performed. Blue colored segments of thecounter bars represent the number of performed movements and white colored segments of the counter bars indicate the movements that still ought to beperformed. The white arrow represents the CS movement and the drawing of a lightning bolt represents the pain-US presentation. CSp1 and CSp- respectivelyrefer to themovement that is followed by the pain-US (ie, left) and themovement that is never followed by the pain-US (ie, right) in the predictable context, whereasCSu1 and CSu2 both refer to movements that are never followed by the pain-US in the unpredictable condition (ie, upwards and downwards). GSmovements arenovel diagonal movements that have either one feature in common with the original CSp1 (GSp1u1 and GSp1u2) or the original CSp2 (GSp2u1 and GSp2u2).During acquisition and transfer of acquisition, 1 startle probe is presented on each trial, in 50% of the trials during the intertrial interval (ITI) and on 50% of the trialsduring the CS movement. In the unpredictable context, ITI probes are presented before the movement (pre-CS-ITI) if the unpredictable pain-US is delivered afterthe CS movement (post-CS-ITI), and the ITI probes are presented after the CS movement (post-CS-ITI) if the unpredictable pain-US is presented before the CSmovement (pre-CS-ITI). During the generalization phase, only CS/GS startle probes were presented.



electrolyte gel. After cleaning the skin with exfoliating peelingcream to reduce inter-electrode resistance, electrodes wereplaced on the left side of the face according to the sitespecifications proposed by Blumenthal et al.6 The raw signalwas amplified by a Coulbourn isolated bioamplifier with a band-pass filter (LabLinc v75–04). The recording bandwidth of the EMGsignal was between 90 Hz and 1 kHz (63 dB). The signal wasrectified online and smoothed by a Coulbourn multifunctionintegrator (LabLinc v76–23 A) with a time constant of 20milliseconds. The EMG signal was digitized at 1000 Hz from200 milliseconds before the onset of the auditory startle probeuntil 1000 milliseconds after probe onset.

2.6. Experimental setting

Participants were seated in an armchair (0.6-m screen distance)in a sound-attenuated and dimmed experimental room,adjacent to the experimenter’s room. Further verbal communi-cation was possible through an intercom system; the experi-menter observed the participants and their physiologicalresponses online by means of a closed-circuit TV installationand computer monitors.

2.7. Response definition and data analysis overview

2.7.1. Response definition of the startle modulation

Using psychophysiological analysis (PSPHA),16 a modular script-based program, we calculated the peak amplitudes defined as themaximum of the response curve within 21 to 175 milliseconds afterthe startle probeonset. All startlewaveformswere visually inspectedoff-line, and technical abnormalities and artifacts were eliminatedusing the PSPHA software. Every peak amplitude was scored bysubtracting its baseline score (averaged EMG level between 1 and20 milliseconds after the probe onset). The raw scores weretransformed to z-scores to account for interindividual differences inphysiological reactivity. To optimize the visualization of the startledata and avoid negative values on the y-axis, T-scores, a lineartransformation of the z-scores, were used in the figures. Averageswere calculated for responding during CS/GS movements and ITIseparately for the predictable and the unpredictable contexts.Participants who failed to reach elevated peak amplitudescompared with baseline on more than 50% of probed trials wereconsidered as nonresponders and were excluded from furtheranalyses. A total of 6 HC and 9 patients with FM from our samplewere excluded because of the absence of reliable startle eyeblinkresponses, therefore the statistical analyses of the psychophysio-logical data were run on a total sample of 33 participants.

2.7.2. Data analysis overview

We performed a series of repeated measures (RM) analyses ofvariance (ANOVAs) to examine the differences between patientswith FM and HC on the respective dependent measures in theunpredictable pain context and predictable pain context sepa-rately. Because we had clear a priori hypotheses, we furtheranalyzed the data using planned comparisons. We appliedHolm–Bonferroni corrections,30 instead of Bonferroni adjustments,to correct for multiple comparisons and to reduce the probability ofa type I error because the Bonferroni method is considered to betoo conservative and inflates the probability of a type II error.47 Theeffect size indication h2

p is reported for all omnibus ANOVA effects.Statistical analyses for all dependent measures were run withStatistica 12 software (StatSoft, Inc, Tulsa, OK).

3. Results

3.1. Unconditioned stimulus characteristicsand questionnaires

As expected, patients with FM selected an electrocutaneousstimulus of a lower intensity than the participants in the HC group(t[46]5 2.85, P, 0.01), however, the selected stimulus was notrated as less painful by the FM group than by the HC group(t[46] 5 1.31, P 5 0.20) (Table 1). In addition, the groupssignificantly differed on all measures (Table 2). In comparisonwith the HC group, the FM group had (1) higher pain intensity(t[46]5 13.27, P, 0.0001), greater pain disability (t[46]5 12.89,P , 0.0001), and more days of being disabled during the last 6months (t[46] 5 5.03, P , 0.0001) (Chronic Pain Grade Scale),(2) higher scores on the Catastrophizing (t[46] 5 3.95, P ,0.001), Limitation (t[46]5 9.15, P, 0.0001), and Internal control(t[46] 5 22.80, P , 0.01) subscales, but lower scores on theOptimism (t[46] 5 23.57, P , 0.001) and Trust (t[46] 5 22.09,P , 0.05) subscales (Pain Cognition List), (3) more fear ofmovement and (re)injury (t[46]5 3.83, P, 0.001) (Tampa Scaleof Kinesiophobia), (4) more impairment in their daily life activitiesdue to the pain (t[46] 5 12.40, P , 0.0001) (FIQ), (5) lowerpositive affect (t[46] 5 24.06, P , 0.001) and higher negativeaffect (t[46] 5 3.37, P , 0.01) (Positive and Negative AffectSchedule), (6) higher scores on both the depression (t[46] 54.97, P , 0.0001) and anxiety (t[46] 5 4.94, P , 0.0001)subscales (Hospital Anxiety Depression Scale).

3.2. Manipulation checks

3.2.1. Contingency awareness

Mean retrospective US-expectancy ratings in the predictable paincontext and the unpredictable pain context were analyzed asmanipulation checks using 2 separate 2 3 2 [group (FM/HC) 3stimulus type (CS1/CS2)] RM ANOVAs. Note: throughout thisarticle, the notations CS1 and CS2 used in the descriptions of thestatistical analyses and the figures, respectively, refer to the CSp1and the CSp2 in the predictable context, and to the unreinforcedCSs, (i.e., CSu1 and CSu2 in the unpredictable context). Withrespect to the GSs in the figures and analyses including thegeneralization data, for the sake of simplicity, GSp1u1 and GSp1u2

and, GSp2u1 andGSp2u2 are referred to respectively as GS1 andGS2. In the predictable context, ratings for the CS1 movementswere higher than for the CS2 movements (main effect of stimulustype: F[1, 46] 5 9.79, P , 0.01, h2

p 5 0.18) and patients with FMgave higher US-expectancy ratings irrespective of the stimulustype (main effect of group: F[1, 46] 5 5.15, P, 0.05, h2

p 5 0.10).The stimulus type3group interactionwas not significant (F, 1,h2

p

5 0.0003). In the unpredictable context, there was no significantmain effect of Stimulus Type (F, 1, h2

p 5 0.01), and no significantstimulus type3group interaction (F, 1,h2

p 5 0.003). PatientswithFM tended to give higher US-expectancy ratings for bothunpredictable movements, but the main effect of group just failedto reach significance (F[1, 46]5 3.64, P5 0.06, h2

p 5 0.07). Theseresults indicate that at least at the end of the experiment, bothgroups were aware of the CS-US contingencies that werepresented during the joystick task.

3.2.2. Unconditioned stimulus (US)-expectancy duringtransfer of acquisition

To checkwhether participants transferred their knowledgeabout thestimulus contingencies acquired in the voluntary to the signaledmovement setup, we collected US-expectancy ratings before eachmovement during the transfer of acquisition phase (Fig. 2). We



analyzed the mean online US-expectancy ratings during thepredictable pain context and the unpredictable pain context using2 separate 23 23 4 (group [FM/HC]3 stimulus type [CS1/CS2]3trial [T1-T4]) RM ANOVAs. In the predictable context, this analysisyielded a significant main effect for stimulus type, F(1, 46) 5 28.46,P,0.001,h2

p 50.38, aswell as ah2p stimulus type3 trial interaction,

F(3, 138) 5 14.78, P , 0.001, h2p 5 0.24. Although this 2-way

interaction was not modulated by group (group 3 stimulus type 3trial interaction, F , 1, h2

p 5 0.02), we conducted plannedcomparisons to examine the transfer effects within each group. Asexpected, in the predictable context, both groups gave higherUS-expectancy ratings for the CS1 movement than for the CS2

movement at the end of the transfer phase (T4) (FM: F[1, 46] 516.08, P , 0.05; HC: F[1, 46] 5 18.33, P , 0.05). Interestingly,however, although the HC showed differential learning immediatelyfrom the first transfer trial (T1) (HC: F[1, 46]56.33,P,0.05), patientswith FM did not (FM: F , 1). The analysis on the US-expectancyratings in the unpredictable context revealed a significant stimulustype 3 group 3 trial interaction, F(3, 138) 5 2.92, P , 0.05, h2

p 50.06. None of the groups showeddifferences inUS-expectancies forboth CSmovements (FM: F[1, 46]5 1.56, P5 0.22; HC: F[1, 46]51.64, P5 0.21). Taken together, it seems that the patients with FMhave more difficulties transferring the previously learned contingen-cies to a novel task context (ie, signaled vs voluntary movementsetup).

3.3. Outcome variables

3.3.1. Self-reported fear of movement-related pain

3.3.1.1. Acquisition effects

We conducted 2 separate 23 23 3 (group [FM/HC]3 stimulustype [CS1/CS2] 3 block [ACQ1-3]) RM ANOVA on the meanpain-related fear ratings for the CS movements after the 3acquisition blocks in both the predictable and the unpredictablepain context (Fig. 3). In the predicable context, there wasa significant main effect of stimulus type, F(1, 46) 5 21.92, P ,0.001, h2

p 5 0.32, and a marginally significant main effect ofgroup, F(1, 46)5 3.27, P5 0.08, h2

p 5 0.07, indicating that pain-related fear ratings were overall slightly higher in the patient group

than in the HCgroup. Importantly, there was a significant stimulustype 3 block interaction, F(2, 92) 5 3.77, P , 0.05, h2

p 5 0.08.Although, this 2-way interaction was not modulated by group, wecontinued to test our a priori hypotheses. Planned comparisonsrevealed that in the predictable context, both groups reportedmore pain-related fear at the end of the acquisition phase (ACQ3)in response to the CS1 movement than in response to the CS2

movement (FM: F[1, 46] 5 11.93, P , 0.01; HC: F[1, 46] 57.31, P , 0.05). Interestingly, this differential fear learningwas already acquired in the first block (ACQ1) in the HC group,F(1, 46) 5 7.78, P , 0.05, but not in the FM patient group,F(1, 46) 5 1.20, P 5 0.28. Furthermore, pain-related fear inresponse to the painful (CS1) movement did not seem to differ forboth groups across the acquisition phase, F(1, 46) 5 1.36,P5.25, but pain-related fear in response to the nonpainful (CS2)movement was elevated in the FM group, F(1, 46) 5 6.36, P ,0.05. The analysis in the unpredictable context yielded a border-line significant main effect of group, F(1, 46)5 2.99, P5.09, h2

p 50.06. As expected, no differences in pain-related fear wereobserved between both unreinforced CS movements in either ofthe groups (both F, 1). Interestingly, however, in the unpredict-able context, FM tended to report more pain-related fear inresponse to both movements as compared with HC in thebeginning of the acquisition phase (ACQ1), F(1, 46) 5 5.24,P 5.027; after Holm–Bonferroni corrections, this effect was nolonger statistically significant (P . 0.0125). At the end ofacquisition (ACQ3), there was no difference in pain-relatedfear in response to unpredictable movements between bothgroups, F(1, 46) 5 1.41, P 5.24.

3.3.1.2. Transfer of acquisition effects

To check whether participants transferred the pain-related fearacquired in the voluntary to the signaled movement setup, wecollected fear ratings before each movement during the transferof acquisition phase (Fig. 4). We analyzed the mean pain-relatedfear ratings in both contexts separately using 2 3 2 3 4 (group[FM/HC]3 stimulus type [CS1/CS2]3 trial [T1-T4]) RMANOVAs.In the predictable context, this analysis generated a significantmain effect for stimulus type, F(1, 46) 5 14.08, P , 0.001,h2

p 50.23, as well as a significant stimulus type 3 trial interaction,

Figure 2. Mean US-expectancy ratings (6SE) for the CS movements in the predictable and the unpredictable context assessed before each trial during thetransfer of acquisition phase (T1-T4).



F(3, 138) 5 3.61, P , 0.05, h2p 5 0.07. Although, this 2-way

interaction was not modulated by group, we continued to test oura priori hypotheses. As expected, in the predictable context, bothgroups reported more pain-related fear in response to the CS1

movement than the CS2 movement at the end of the transferphase (T4) (FM: F[1, 46] 5 9.06, P , 0.05; HC: F[1, 46] 5 6.83,P, 0.05). Interestingly, however, although theHC transferred theacquired differential fear learning immediately to the first transfertrial (T1) (HC: F[1, 46]5 10.91, P, 0.05), the patients with FM didnot (FM: F , 1). In line with the data pattern observed duringacquisition, pain-related fear elicited by the painful (CS1)movement was similar in both groups across the transfer phase,

F , 1, but pain-related fear elicited by the nonpainful (CS2)movement tended to be higher in the FM group, F(1, 46)5 4.40,P5 0.041; after Holm–Bonferroni corrections, this effect was nolonger statistically significant (P . 0.0125). In the unpredictablecontext, analyses showed a significant main effect of group,F(1, 46) 5 4.15, P , 0.05, h2

p5 0.08, as well as a significantstimulus type 3 group 3 block interaction, F(3, 138) 5 2.69,P , 0.05, h2

p5 0.06. None of the groups showed differences inpain-related fear between both CS movements (FM: F , 1; HC:F[1, 46] 5 1.53, P 5 0.22). Consistent with the acquisition data,patients with FM again tended to report more pain-related fear forboth movements than the HC in the beginning of the transfer

Figure 3.Mean self-reported pain-related fear (6SE) in response to the CSmovements in the predictable and the unpredictable context assessed after each blockduring acquisition (ACQ1-3).

Figure 4.Mean self-reported pain-related fear (6SE) in response to the CSmovements in the predictable and the unpredictable context assessed before each trialduring the transfer of acquisition phase (T1-T4).



phase (T1), F(1, 46)5 4.42, P, 0.05, but after Holm–Bonferronicorrections, this effect was no longer statistically significant (P .0.0125). No differences in pain-related fear for both unpredictablemovements were observed at the end of the transfer phase (T4),F(1, 46) 5 2.72, P 5 0.11. Basically, it seems that the patientswith FM have more difficulties transferring the differential fearlearning to a novel task context (ie, signaled vs voluntarymovement setup) and, as a consequence, are generally moreanxious.

3.3.1.3. Generalization effects

To examine generalization of pain-related fear to the noveldiagonal GSmovements in the predictable and the unpredictablepain context, we ran two 2 3 4 (group [FM/HC] 3 stimulus type[CS1/GS1/GS2/CS2]) RM ANOVAs (Fig. 5). From this analysis,in the predictable context, a significant main effect of stimulustype, F(3, 138) 5 12.68, P , 0.001, h2

p5 0.22, and a marginallysignificant main effect of group emerged, F(1, 46) 5 3.37, P 50.07, h2

p 5 0.07. Although, the stimulus type3 group, interactionfailed to reach significance, F(3, 138) 5 1.73, P 5 0.16, h2

p 50.04, we continued to test our a priori hypotheses. In line withprevious research, planned comparisons confirmed that in theHC, GS1 movements (diagonal movements proprioceptivelyrelated to the original CS1) elicited higher fear of pain reportsduring the generalization test than the GS2movements (diagonalmovements proprioceptively related to the original CS2) whentested in the predictable context, F(1, 46) 5 6.00, P , 0.05. Incontrast, this differential pain-related fear generalization effectwas not observed in the FM group, F(1, 46) 5 2.03, P5 0.16. Inthe unpredictable context, this analysis showed a borderlinesignificant main effect of group, F(1, 46) 5 4.01, P 5 0.05, h2

p 50.08, and no significant main effect of stimulus type, F, 1, h2

p 50.01. The stimulus type 3 group interaction also was not

significant, F(3, 138) 5 1.41, P 5 0.24, h2p 5 0.03. Planned

comparisons revealed that the same diagonal GS1 and GS2

movements did not elicit differential fear of pain ratings in eitherof the groups (both F , 1). Interestingly, the novel movementstested in the unpredictable context tended to produce more fearin the FM group than in the HC group, F(1, 46)5 4.53, P5 0.039;but after Holm–Bonferroni corrections, this effect was no longerstatistically significant (P . 0.017). Basically, it seems that in thepredictable context, pain-related fear in patients with FM spreadsin a less stimulus-specific way than in theHC, and fear is generallyhigher in response to all novel movements in the unpredictablecontext.

3.3.2. Unconditioned stimulus (US)-expectancy duringgeneralization

We analyzed the mean US-expectancy ratings in both contextsduring the generalization test using 2 separate 2 3 4 (group[FM/HC] 3 stimulus type [CS1/GS1/GS2/CS2]) RM ANOVAs.This analysis in the predictable context revealed a significant maineffect for stimulus type, F(3, 138)5 16.78, P, 0.001, h2

p 5 0.28.Both the main effect of group and the stimulus type 3 groupinteraction, F(3, 138) 5 11.13, P , 0.001, h2

p 5 0.19, failed toreach significance. Notwithstanding the absence of the 2-wayinteraction, we further tested our hypotheses using plannedcomparisons. Unconditioned stimulus-expectancies for the GS1

movements were elevated compared with those for the GS2

movements in both groups, but to a lesser extent in the patientgroup (HC: F[1, 46]5 13.38, P, 0.05; FM: F[1, 46]5 5.89, P,0.05) (Fig. 6). In the unpredictable context, the analysis did notshow any significant main effects for group or stimulus type (bothF , 1), as well as no significant 2-way interaction, F(3, 138) 51.80, P 5 0.15, h2

p 5 0.02. Planned comparisons revealed that

Figure 5.Mean self-reported pain-related fear (6SE) in response to the original CS movements (CS1/CS2) and the novel diagonal GS movements (GS1/GS2) inthe predictable and the unpredictable context assessed before each trial during the generalization phase.



the same GSmovements did not generate any differences in US-expectancy ratings in either of the groups, both F , 1.

3.3.3. Eyeblink startle modulation

We analyzed the mean startle responses during acquisition andtransfer of acquisition with a 2 3 2 3 3 3 4 (group [FM/HC] 3context [predictable/unpredictable] 3 stimulus type [CS1/CS2/ITI]3 block [ACQ1-3, ACQ*]) RM ANOVA (Fig. 7) (because startleresponses elicited by a single probe are not sufficiently reliable;means were calculated for the transfer phase and included asa fourth acquisition block [ACQ*] in the startle analysis). Theanalysis comparing psychophysiological fear responding elicitedduring the CSs and the ITI in both contexts revealed significantmain effects for context, F(1, 31)5 13.99, P, 0.001, h2

p 5 0.31,and for block, F(3, 93)5 12.97, P, 0.001, h2

p 5 0.29, indicatinghabituation, that is, the startle responses declined graduallyacross blocks. Importantly, the stimulus type 3 contextinteraction was significant, F(2, 62)5 6.86,P, 0.001,h2

p 5 0.18,confirming that the startle responses elicited during the ITI and therespective CSs differed significantly in both the predictable andthe unpredictable contexts. None of the interactions with groupwas significant (all P. 0.05, all h2

p , 0.12). Planned comparisonsfurther demonstrated that in both groups, the mean startleamplitudes during the CS1 movement were higher than duringthe CS2 movement, F(1, 31) 5 6.10, P , 0.05. In theunpredictable context, no such difference was observed,F(1, 31)5 1.21, P5 0.28. Interestingly, startle amplitudes duringthe ITIs in the unpredictable context were more elevated than inthe predictable context, F(1, 31)5 21.22, P, 0.001, suggestingthat more contextual pain-related fear emerged in the unpredict-able context.

To examine generalization of psychophysiological fearfulresponding to the novel diagonal GS movements, we performeda 2 3 2 3 4 (group [FM/HC] 3 context [predictable/unpredictable]3 stimulus type [CS1/GS1/GS/CS2] RM ANOVA)(Fig. 8). The pattern of differential startle respondingwas similar tothat of the verbal fear ratings, but the analysis failed to showa significant context 3 stimulus type interaction or a significant3-way interaction, both F, 1, both h2

p 5 0.01). Because we hadclear a priori hypotheses, planned comparisons were furthercalculated. Startle amplitudes elicited during the GS1 move-ments tended to be elevated compared with those during theGS2 movements in the predictable context for the HC group,F(1, 31)5 4.85,P5 0.035; but after Holm–Bonferroni corrections,this effect was no longer statistically significant (P 5 0.0125).There was no difference between the startle responses duringboth GSmovements in the FM group, F, 1. No differential startleamplitudes were elicited by the same GS movements in theunpredictable context (both F , 1).

4. Discussion

We examined the differences in the acquisition of fear ofmovement-related pain in both a predictable and unpredictablecontext and whether FM and HC differ in the spreading of feartoward novel movements that more or less resemble the original(non)painful movements. We used a similar approach asMeulders and Vlaeyen44 in their study in HC, except for somemethodological improvements. We hypothesized that ascompared to the HC, the FM group would show (1) slower/less differential pain-related fear acquisition due to impairedsafety learning, (2) more contextual pain-related fear during

Figure 6.MeanUS-expectancy ratings (6SE) for the original CSmovements (CS1/CS2) and the novel diagonal GSmovements (GS1/GS2) in the predictable andthe unpredictable context assessed before each trial during the generalization phase.



acquisition and generalization, (3) less differential fear general-ization, and (4) lower pain thresholds. The results can besummarized as follows:

First, in the predictable context, we successfully demonstratedthe acquisition of pain-related fear to the painful CS1 movement,but not to the nonpainful CS2 movement, in both groups. Thiseffect was expressed by elevated fear reports, higher startle

amplitudes, and higher US-expectancies to the CS1 than to theCS2. As expected, no such differential fear learning emerged inthe unpredictable context. Interestingly, both groups did notacquire this differential fear learning at the same time; althoughHC showed this effect after only 1 training block, it took longer forthe FM to pick up the CS-US contingencies. Further, it appearsthat the fear to the CS2 was higher in the FM than in the HC

Figure 7.Mean eyeblink startle amplitudes (6SE’s) during the CSmovements and the ITI in the predictable and the unpredictable contexts during the acquisition.Note that for graphic purposes, T-scores were used.

Figure 8. Mean eyeblink startle amplitudes (6SE) during the original CS movements (CS1/CS2) and the novel diagonal GS movements (GS1/GS2) in thepredictable and the unpredictable contexts during the generalization phase. Note that for graphic purposes, T-scores were used.



group, indicating a lack of safety learning rather than excessivefear to the CS1.

Second, in both groups, we successfully establishedcontextual pain-related fear, as indicated by higher ITI startles,in the unpredictable than in the predictable pain context. Moreimportantly, our second hypothesis was also supported, that is,contextual pain-related fear was more substantial in the FMthan in the HC group. More indirect evidence was found in thepain-related fear reports: both unpredictable movements andthe nonpainful CS2 movement yielded higher fear ratings in theFM than in the HC group. This relationship might be explainedby a spill-over effect from the threatening/unpredictable context(ie, contextual fear) that is greater in the FM group. Generally,these results corroborate previous findings in anxiety disorders(eg, posttraumatic stress disorder, panic disorder) suggestingthat anxiety patients exhibit elevated contextual fear and/orimpaired safety learning.23,25,27,34,40,41,46

Third, we replicated differential fear generalization to the noveldiagonal movements in the predictable context in the HC. Morespecifically, the GS1 movements produced higher fear ratings,higher US-expectancy ratings, and marginally significant higherstartle responses than the GS2 movements. As predicted, thespreading of cued pain-related fear was not stimulus specific inthe FM group, that is, fear responding generalized equally to themovements resembling the original CS1 and CS2 movements.This finding was evident in both fear measures. In the US-expectancy however, there wasminimal evidence for differentialgeneralization of the acquired contingencies, suggesting thatthe major anomaly resides in the fear response system (eg,elevated baseline startle) rather than on the basic associativelearning level (dissociations between response systems havebeen reported earlier).4,52 As expected, in the unpredictablecontext, the same GS movements did not expose a differentialgeneralization effect in the HC. Neither the fear reports, nor US-expectancy ratings, nor startle responses differed between theGS1 and the GS2movements. The same generalization patternwas observed in the FM group, but again, fear elicited by the GSstimuli was generally higher in the unpredictable contextcompared with the HC group. Finally, patients with FM choselower pain-US intensity than that of the HC, corroborating theobservation that pain thresholds (pressure and heat) arereduced in patients with FM.5,36,37,48

Some other interesting observations deserve further attention.First, we switched from a voluntary movement setup to a signaledmovement setup; this signaled procedure is necessary to indicatethe novel diagonal movements that have to be performed duringthe generalization test. In line with previous findings, participantsin the HC transferred their CS-US knowledge easily from theacquisition to the transfer phase, whereas this trend was not thecase in the FM group. The US-expectancy and the fear ratingsassessed before the actual execution of the movements showthat patients do not show any differential conditioned respondingin the beginning of the transfer phase. Again, this effect is largelydriven by increased fear to the CS2 movement. Furthermore,although the difference in contextual pain-related fear dissipatedby the end of acquisition, during the transfer phase, elevatedpain-related fear ratings to both movements reappeared in theFM but not in the HC group. Altogether, these data suggest thatadaptive differential pain-related fear, and safety learning inparticular, transfers poorly to novel contexts in the FM group.Evenmore maladaptive is the fear generalization in the FM group,that is, fear seems to spread to all novel movements whether theyare related to the original painful or nonpainful movements. Thisovergeneralization is seemingly contradictory: on the 1 hand,

patients with FM overgeneralize their fear to all novel move-ments, but once the adaptive differential fear learning tookplace, it does not seem to generalize well to new contexts (ie,overdiscrimination). Especially, inhibitory learning to the CS2

movement seems to be fragile and context-dependent. Thisobservation can be understood in terms of the well-knowndissociation between first and second learned associations:acquisition learning (first learned association) normally general-izes easily across contexts, whereas extinction learning(second learned association that inhibits the behavioralexpression of the first learned association) does not, but ismore context-dependent.7–10 By analogy, safety learning mightbe more vulnerable during context switches, especially inindividuals characterized by fragile safety learning.

Second, based on the retrospective US-expectancy ratings,patients with FM acquired the CS-US contingencies, althoughthey fail to inhibit their psychophysiological and self-reported fearto the nonpainful movements during the task. These resultscontradict the findings of Jenewein et al.,32 demonstrating thatonly 50% of the patients with FM were contingency aware ascompared with 100% and 86% of the HC and the osteoarthritisgroup, respectively. This findingmight be associated with the lessclear-cut contingencies in their study. That is, a triangle (CS2) wasalways followed by a low-temperature heat stimulus and a square(CS1) was either followed by a high- or low-temperature heatstimulus. Because of this partial reinforcement scheme, patientswith FM might have had more difficulties to discriminateaccurately between cues that were sometimes or always followedby the low temperature.

Some limitations should be addressed as well. First, weexcluded 15 nonresponders from the startle analyses; generalinteractions might have failed to reach statistical significancebecause of reduced statistical power. The relatively high numberof nonrespondersmight be related to the inclusion of older adults,who typically haveweaker startle responses.19 Second, emotionsalso fade when getting older (ie, blunted affect), which is reflectedin the lower fear measures compared with healthy studentsamples. Alternatively, this result might be explained by socialdesirability in students or by the fact that patients’ persistent real-life pain is more extreme than the experimental pain. Third, thisstudy cannot draw any conclusion regarding the causal relation-ship between poor acquisition and overgeneralization of pain-related fear and the development of FM, therefore longitudinalstudies are needed to disentangle the causal status of fearlearning deficits in the development of FM. Nevertheless, we areconvinced that this learning deficit might be involved in themaintenance of FM, because when potential danger is notsuccessfully identified, it boils down to sustained anxiety, furtherspreading of fear, and proliferation of avoidance and defensivebehaviors. Fourth, the groups also differed with respect tomedication use12,17,29 and comorbidity with anxiety34,40,41 anddepression, thus we cannot exclude the possibility that thesefactors might have contributed to the observed differences in fearlearning and expression. Indeed, anxiolytics might affect theexpression of context conditioning,26 opioids have shown toimpair fear learning,17 and antidepressants may enhance cuedfear conditioning.13,29 However, given the possible oppositeeffects of the different drugs used in our patient group, it is ratherunlikely that themedication use explains all the observed variancebetween FM and HC. Fifth, although some of the omnibusANOVA interactionswere not significant, probably due to a lack ofstatistical power, we continued testing our a priori hypotheses.Bonferroni corrections for multiple testing have been criticized forbeing too conservative, for increasing the probability of type II



errors and undermining the statistical power,47 therefore we usedHolm–Bonferroni corrections.

Despite these limitations, the data seem to support the ideathat increased anxiety levels in patients with FM are likelyassociated with deficient safety learning, which in turn enhances(over)generalization of pain-related fear and the possibly mainte-nance of widespread pain. Future prospective studies witha longitudinal design are needed to further test this intriguingpossibility.

To conclude, this study is the first to demonstrate sloweracquisition due to impaired safety learning and overgeneralizationof pain-related fear in FM compared with HC. We believe thatinvestigating (over)generalization, possibly a transdiagnostic markerof pathology, might help shed light on pathogenesis of musculo-skeletal widespread pain and spreading of anxiety and pain.

Conflict of interest statement

The authors have no conflicts of interest to declare. A.Meulders isa postdoctoral researcher of the Research Foundation Flanders(FWO Vlaanderen), Belgium (grant ID: 12E3714N). This studywas also supported by the Odysseus Grant “The Psychology ofPain and Disability Research Program” to J.W.S.V. and fundedby the Research Foundation Flanders (FWO-Vlaanderen),Belgium (grant ID: G090208N), and an EFIC Grunenthal Re-search Grant (E-G-G ID: 169518451) to A.M. The data of thisstudy were presented at the topical workshop on “Generalizationprocesses in chronic pain: an associative learning approach” atthe 15th World Congress on Pain, Buenos Aires, Argentina,October, 2014.

Article history:Received 27 February 2014Received in revised form 22 September 2014Accepted 28 October 2014

References

[1] Asmundson GJ, Katz J. Understanding the co-occurrence of anxietydisorders and chronic pain: state-of-the-art. Depress Anxiety 2009;26:888–901.

[2] Asmundson GJ, Taylor S. Role of anxiety sensitivity in pain-related fearand avoidance. J Behav Med 1996;19:577–86.

[3] Baas JMP, van Ooijen L, Goudriaan A, Kenemans JL. Failure to conditionto a cue is associated with sustained contextual fear. Acta Psychol (Amst)2008;127:581–92.

[4] Beckers T, Krypotos A-M, Boddez Y, Effting M, Kindt M. What’s wrongwith fear conditioning? Biol Psychol 2013;92:90–6.

[5] Berglund B, Harju EL, Kosek E, Lindblom U. Quantitative and qualitativeperceptual analysis of cold dysesthesia and hyperalgesia in fibromyalgia.PAIN 2002;96:177–87.

[6] Blumenthal TD, Cuthbert BN, Filion DL, Hackley S, LippOV, Van Boxtel A.Committee report: Guidelines for human startle eyeblinkelectromyographic studies. Psychophysiology 2005;42:1–15.

[7] Bouton ME. Context and ambiguity in the extinction of emotionallearning: implications for exposure therapy. Behav Res Ther 1988;26:137–49.

[8] Bouton ME. Context, ambiguity, and classical conditioning. Curr DirPsychol Sci 1994;3:49–53.

[9] Bouton ME. A learning theory perspective on lapse, relapse, and themaintenance of behavior change. Health Psychol 2000;19(1 suppl):57–63.

[10] Bouton ME. Context, ambiguity, and unlearning: sources of relapse afterbehavioral extinction. Biol Psychiatry 2002;52:976–86.

[11] Burckhardt CS, Clark SR, Bennett RM. The Fibromyalgia ImpactQuestionnaire: development and validation. J Rheumatol 1991;18:728–33.

[12] Burghardt NS, Bauer EP. Acute and chronic effects of selective serotoninreuptake inhibitor treatment on fear conditioning: implications forunderlying fear circuits. Neuroscience 2013;247:253–72.

[13] Burghardt NS, Bush DE, McEwen BS, LeDoux JE. Acute selectiveserotonin reuptake inhibitors increase conditioned fear expression:blockade with a 5-HT(2C) receptor antagonist. Biol Psychiatry 2007;62:1111–8.

[14] Craske M, Hermans D, Vansteenwegen D. Fear and learning. From basicprocesses to clinical applications. Washington: American PsychologicalAssociation, 2006.

[15] Crombez G, Eccleston C, Van Damme S, Vlaeyen JW, Karoly P. Fear-avoidance model of chronic pain: the next generation. Clin J Pain 2012;28:475–83.

[16] de Clercq A, Verschuere B, de Vlieger P, Crombez G. Psychophysiologicalanalysis (PSPHA): a modular script-based program for analyzingpsychophysiological data. Behavior Research Methods 2006;38:504–10.

[17] Eippert F, Bingel U, Schoell E, Yacubian J, Buchel C. Blockade ofendogenous opioid neurotransmission enhances acquisition ofconditioned fear in humans. J Neurosci 2008;28:5465–72.

[18] Engelen U, De Peuter S, Victoir A, Van Diest I, Van Den Bergh O. Verderevalidering van de positive and negative affect schedule (PANAS) envergelijking van twee Nederlandstalige versies (further validation of thepositive and negative affect schedule [PANAS] and comparison of twoDutch versions). Gedrag & Gezondheid: Tijdschrift voor Psychologie enGezondheid 2006;34:89–102.

[19] Feng MC. Age-related affective modulation of the startle eyeblinkresponse: older adults startle most when viewing positive pictures.Psychol Aging 2011;26:752.

[20] Gazendam FJ, Kamphuis JH, Kindt M. Deficient safety learningcharacterizes high trait anxious individuals. Biol Psychol 2013;92:342–52.

[21] GheldofEL,CrombezG,VandenBusscheE, Vinck J,VanNieuwenhuyseA,Moens G, Mairiaux P, Vlaeyen JW. Pain-related fear predicts disability, butnot pain severity: a path analytic approach of the fear-avoidancemodel. EurJ Pain 2010;14:870.e871–9.

[22] Ghirlanda S, Enquist M. A century of generalization. AnimBehav 2003;66:15–36.

[23] Grillon C. Fear-potentiated startle conditioning to explicit and contextualcues in Gulf War veterans with posttraumatic stress disorder. J AbnormPsychol 1999;108:134.

[24] Grillon C. Associative learning deficits increase symptoms of anxiety inhumans. Biol Psychiatry 2002;51:851–8.

[25] Grillon C. Startle reactivity and anxiety disorders: aversive conditioning,context, and neurobiology. Biol Psychiatry 2002;52:958–75.

[26] Grillon C. The benzodiazepine alprazolam dissociates contextual fearfrom cued fear in humans as assessed by fear-potentiated startle. BiolPsychiatry 2006;60:760.

[27] Grillon C. Models and mechanisms of anxiety: evidence from startlestudies. Psychopharmacology (Berl) 2008;199:421–37.

[28] Grillon C, Falls WA, Ameli R, Davis M. Safety signals and human anxiety:a fear-potentiated startle study. Anxiety 1994;1:13–21.

[29] Grillon C, Levenson J, Pine DS. A single dose of the selective serotoninreuptake inhibitor citalopram exacerbates anxiety in humans: a fear-potentiated startle study. Neuropsychopharmacology 2007;32:225–31.

[30] Holm S. A simple sequentially rejective multiple test procedure. Scand JStat 1979;6:65–70.

[31] Honig WK, Urcuioli PJ. The legacy of Guttman and Kalish (1956): 25years of research on stimulus-generalization. J Exp Anal Behav 1981;36:405–45.

[32] Jenewein J, Moergeli H, Sprott H, Honegger D, Brunner L, Ettlin D,Grillon C, Bloch K, Brugger M, Schwegler K, Schumacher S, Hasler G.Fear-learning deficits in subjects with fibromyalgia syndrome? Eur JPain 2013;17:1374–84.

[33] Jensen JN, Karpatschof B, Labriola M, Albertsen K. Do fear-avoidancebeliefs play a role on the association between low back pain and sicknessabsence? A prospective cohort study among female health care workers.J Occup Environ Med 2010;52:85–90.

[34] Jovanovic T, Kazama A, Bachevalier J, Davis M. Impaired safety signallearning may be a biomarker of PTSD. Neuropharmacology 2012;62:695–704.

[35] Kurtze N, Svebak S. Fatigue and patterns of pain in fibromyalgia:correlations with anxiety, depression and co-morbidity in a female countysample. Br J Med Psychol 2001;74:523–37.

[36] Lautenbacher S, Rollman GB. Possible deficiencies of pain modulation infibromyalgia. Clin J Pain 1997;13:189–96.

[37] Lautenbacher S, Rollman GB, McCain GA. Multi-method assessment ofexperimental and clinical pain in patients with fibromyalgia. PAIN 1994;59:45–53.

[38] Lissek S, Grillon C. Overgeneralization of conditioned fear in the anxietydisorders. J Psychol 2010;218:146–8.



[39] Lissek S, Powers AS, McClure EB, Phelps EA, Woldehawariat G, Grillon C,Pine DS. Classical fear conditioning in the anxiety disorders: a meta-analysis. Behav Res Ther 2005;43:1391–424.

[40] Lissek S, Rabin S, Heller RE, LukenbaughD,GeraciM, PineDS,Grillon C.Overgeneralization of conditioned fear as a pathogenic marker of panicdisorder. Am J Psychiatry 2010;167:47–55.

[41] Lissek S, Rabin SJ, McDowell DJ, Dvir S, Bradford DE, Geraci M, Pine DS,Grillon C. Impaired discriminative fear-conditioning resulting from elevatedfear responding to learned safety cues among individuals with panicdisorder. Behav Res Ther 2009;47:111–8.

[42] Meulders A, Vansteenwegen D, Vlaeyen JW. The acquisition of fear ofmovement-related pain and associative learning: a novel pain-relevanthuman fear conditioning paradigm. PAIN 2011;152:2460–9.

[43] Meulders A, Vansteenwegen D, Vlaeyen JW.Women, but notmen, reportincreasingly more pain during repeated (un)predictable painfulelectrocutaneous stimulation: evidence for mediation by fear of pain.PAIN 2012;153:1030–41.

[44] Meulders A, Vlaeyen JWS. The acquisition and generalization of cued andcontextual pain-related fear: an experimental study using a voluntarymovement paradigm. PAIN 2013;154:272–82.

[45] Mineka S, Zinbarg R. A contemporary learning theory perspective on theetiology of anxiety disorders: it’s not what you thought it was. Am Psychol2006;61:10–26.

[46] Morgan CA. Exaggerated acoustic startle reflex in Gulf War veterans withposttraumatic stress disorder. Am J Psychiatry 1996;153:64.

[47] Perneger TV.What’s wrongwith Bonferroni adjustments. BMJ 1998;316:1236–8.

[48] Petzke F, Clauw DJ, Ambrose K, Khine A, Gracely RH. Increased painsensitivity in fibromyalgia: effects of stimulus type and mode ofpresentation. PAIN 2003;105:403–13.

[49] Rhudy JL, Meagher MW. Fear and anxiety: divergent effects on humanpain thresholds. PAIN 2000;84:65–75.

[50] Roelofs J, van Breukelen G, Sluiter J, Frings-Dresen MH, Goossens M,Thibault P, Boersma K, Vlaeyen JW. Norming of the Tampa scale forkinesiophobia across pain diagnoses and various countries. PAIN 2011;152:1090–5.

[51] Schneider C, Palomba D, Flor H. Pavlovian conditioning of muscularresponses in chronic pain patients: central and peripheral correlates.PAIN 2004;112:239–47.

[52] Soeter M, Kindt M. Dissociating response systems: erasing fear frommemory. Neurobiol Learn Mem 2010;94:30–41.

[53] Spinhoven PH, Ormel J, Sloekers PPA, Kempen GIJM, Speckens AEM,Van Hemert AM. A validation study of the hospital anxiety and depressionscale (HADS) in different groups of Dutch subjects. Psychol Med 1997;27:363–70.

[54] Spruyt A, Clarysse J, Vansteenwegen D, Baeyens F, Hermans D. Affect4.0: a free software package for implementing psychological andpsychophysiological experiments. Exp Psychol 2010;57:36.

[55] Staud R, Rodriguez ME. Mechanisms of disease: pain in fibromyalgiasyndrome. Nat Clin Pract Rheumatol 2006;2:90–8.

[56] Swinkels-Meewisse IE, Roelofs J, Oostendorp RA, Verbeek AL,Vlaeyen JW. Acute low back pain: pain-related fear and paincatastrophizing influence physical performance and perceiveddisability. PAIN 2006;120:36–43.

[57] Swinkels-Meewisse IE, Roelofs J, Schouten EG, Verbeek AL,Oostendorp RA, Vlaeyen JW. Fear of movement/(re)injury predictingchronic disabling low back pain: a prospective inception cohort study.Spine (Phila Pa 1976) 2006;31:658–64.

[58] Thieme K, Turk DC, Flor H. Comorbid depression and anxiety infibromyalgia syndrome: relationship to somatic and psychosocialvariables. Psychosom Med 2004;66:837–44.

[59] Van Breukelen GJ, Vlaeyen JW. Norming clinical questionnaires withmultiple regression: the pain cognition list. Psychol Assess 2005;17:336–44.

[60] Vansteenwegen D, Iberico C, Vervliet B, Marescau V, Hermans D.Contextual fear induced by unpredictability in a human fear conditioningpreparation is related to the chronic expectation of a threatening US. BiolPsychol 2008;77:39–46.

[61] Vlaeyen JW, Linton SJ. Fear-avoidance and its consequences inchronic musculoskeletal pain: a state of the art. PAIN 2000;85:317–32.

[62] Vlaeyen JW, Linton SJ. Fear-avoidancemodel of chronicmusculoskeletalpain: 12 years on. PAIN 2012;153:1144–7.

[63] Von Korff M, Ormel J, Keefe FJ, Dworkin SF. Grading the severity ofchronic pain. PAIN 1992;50:133–49.

[64] Watson D, Clark LA, Tellegen A. Development and validation of briefmeasures of positive and negative affect: the PANAS scales. J Pers SocPsychol 1988;47:1063–70.

[65] Weir PT, Harlan GA, Nkoy FL, Jones SS, Hegmann KT, Gren LH, Lyon JL.The incidence of fibromyalgia and its associated comorbidities:a population-based retrospective cohort study based on InternationalClassification of Diseases, 9th revision codes. J Clin Rheumatol 2006;12:124–8.

[66] Wolfe F, Clauw DJ, Fitzcharles MA, Goldenberg DL, Katz RS, Mease P,Russell AS, Russell IJ, Winfield JB, Yunus MB. The American College ofRheumatology preliminary diagnostic criteria for fibromyalgia andmeasurement of symptom severity. Arthritis Care Res (Hoboken) 2010;62:600–10.

[67] Zigmond AS, Snaith RP. The hospital anxiety and depression scale. ActaPsychiatr Scand 1983;67:361–70.



Differences in pain-related fear acquisition and generalization: An experimental study comparing...

Documents

Transcript of Differences in pain-related fear acquisition and generalization: An experimental study comparing...