ORIGINAL ARTICLE

Eye movements during visual search for emotional faces in individuals with chronic headache D.E. Schoth, H.J. Godwin, S.P. Liversedge, C. Liossi Academic Unit of Psychology, University of Southampton, UK

Correspondence Christina Liossi E-mail: [email protected] Funding sources This investigation was supported by a grant from the Economic & Social Research Council, grant reference RES-000-22-4128. Conflicts of interest None declared. Accepted for publication 7 August 2014 doi:10.1002/ejp.595

Abstract Background: Attentional biases for pain-related information have been frequently reported in individuals with chronic pain. Recording of participants’ eye movements provides a continuous measure of attention, although to date this methodology has received little use in research exploring attentional biases in chronic pain. The aim of the current investigation was to explore the specificity of attentional orienting bias using a novel visual search task while recording participant eye movement behaviours. This also allowed for the investigation of whether attentional biases for pain-related information exist in the presence of multiple stimuli competing for attention. Methods: Twenty-three participants with chronic headache and 24 pain-free, healthy control participants were engaged in a visual search task where pain, angry, happy and neutral faces were used as both target and distractor stimuli. While completing this task, participants’ eye movements were recorded. Results: Supporting the adopted hypothesis, participants with chronic headache, relative to healthy controls, demonstrated a significantly higher proportion of initial fixations to target pain expressions when the pain expressions were presented in displays containing neutral-distractor faces. No significant differences were found between groups in the time taken to fixate target pain expressions (localization time). Conclusions: Individuals with chronic headache show facilitated initial orienting towards pain expressions specifically when used as targets in a visual search task. This study adds to a growing body of research supporting the presence of pain-related attentional biases in chronic pain as assessed via different experimental paradigms, and shows biases to exist when multiple stimuli competing for attention are presented simultaneously.

1. Introduction Theoretical accounts of pain and attention (e.g., Pincus and Morley, 2001; Van Damme et al., 2010) predict that individuals with chronic pain will show attentional biases towards pain-related information. Although this prediction has received empirical support (Schoth et al., 2012; Crombez et al., 2013), one meta-analysis reported a small effect size (Crombez et al., 2013), and it has been argued that the © 2014 European Pain Federation - EFIC®

commonly used visual-probe task is not a robust measure of bias (Dear et al., 2011). Exploration of attentional biases with alternative paradigms is therefore essential, particularly as implications of bias are currently debated (Liossi, 2012; Sharpe, 2012; Van Ryckeghem and Crombez, 2014), and preliminary evidence supports the therapeutic benefits of attentional bias modification (ABM) in pain populations (Carleton et al., 2011; Sharpe et al., 2012; Schoth et al., 2013). Eur J Pain •• (2014) ••–••

1

Visual search for emotional faces in chronic headache

What’s already known about this topic? • Individuals with chronic pain show attentional biases for pain-related information. • Attentional biases may be implicated in the onset and maintenance of chronic pain. What does this study add? • This investigation explored attentional biases in chronic headache using a visual search task with concurrent recording of eye movement behaviours. • Individuals with chronic headache, relative to healthy controls, demonstrate facilitated initial orienting of attention to facial expressions of pain when engaged in visual search for emotionally salient stimuli.

D.E. Schoth et al.

Considering this, the present study further explored the specificity of bias in chronic pain using pain, angry, happy and neutral faces as both target and distractor stimuli in a visual search task while participants’ eye movements were recorded. Based upon the theoretical accounts of pain and attention, it was hypothesized that individuals with chronic headache, relative to healthy controls, would show biases in attentional orienting to pain expressions specifically as indicated via: (1) significantly higher proportion of initial fixations to target pain expressions and (2) significantly faster times to fixate target pain expressions (i.e., localization time).

2. Methods 2.1 Participants

Eye tracking provides an excellent method for the assessment of moment-to-moment visual processing during experimental tasks (Rayner, 1998; Liversedge and Findlay, 2000) and is a more direct measure of attentional deployment than manual response latencies alone (Duc et al., 2008). However, there has only been a limited degree of eye movement research examining biases in chronic pain (Yang et al., 2013; Liossi, Schoth, Goodwin, and Liversedge, 2014). Using a visual scanning task showing pain, angry, happy and neutral faces concurrently, Liossi et al. (2014) confirmed a bias in attentional orienting for pain-related information in individuals with chronic headache, compared with healthy controls, in the presence of multiple salient stimuli also competing for attention. It remains to be ascertained whether such patterns of bias are present when individuals with chronic pain are actively searching for specific salient information in their environment rather than passively observing. Addressing this question has important implications both for theoretical accounts of chronic pain and attention, as well as for our ability to estimate how chronic pain might influence participants’ engagement with the environment in everyday life (see Van Damme et al., 2010 for a motivational account of attention to pain). Visual search is a ubiquitous behaviour made necessary by the complexity and richness of information which typically exists in a visual scene (Chun, 2003). The specificity of bias in chronic pain has been highlighted (e.g., Liossi et al., 2011; Schoth and Liossi, 2013), but only a limited number of studies have used different emotional facial expressions in the same paradigm (i.e., Khatibi et al., 2009; Liossi et al., 2014). 2 Eur J Pain •• (2014) ••–••

Participants were recruited from the South of England via press advertisements and from the University of Southampton via posters. Inclusion criteria for the chronic headache group were: (1) having been diagnosed with primary tension-type headache or migraine by a general practitioner or consultant neurologist, and satisfying the criteria stated in the International Classification of Headache Disorders, 2nd edition, Headache Classification (Headache Classification Subcommittee of the International Headache Society, 2004), i.e., occurring 15 or more days per month for more than 3 months and in the absence of medication overuse; (2) aged 18 years or over; and (3) having normal or corrected-tonormal vision. Exclusion criteria were: (1) having a diagnosis or receiving treatment for any psychiatric disorder, either currently or within the past 5 years and (2) suffering from any other form of chronic pain including secondary headache (i.e., caused by another medical disorder). Inclusion criteria for the healthy control group were: (1) aged 18 years or over and (2) having normal or corrected-to-normal vision. Exclusion criteria were: (1) having a diagnosis or receiving treatment for any psychiatric disorder, either currently or within the past 5 years; (2) suffering from any form of chronic or recurrent pain (in terms of headache frequency having more than seven headaches per month); and (3) taking any psychotropic or analgesic medication regularly. Eligibility was established by the first author via telephone interview. Based upon these criteria, at this stage, two individuals were excluded who did not meet research criteria for chronic headache, and two individuals with visual impairments. Fifty-one individuals were invited to the lab to participate in the study. There were difficulties calibrating the eye tracker for four of them (two individuals with chronic headache, two healthy control individuals) and these individuals were therefore excluded from the investigation and not considered further. The final sample consisted of 47 participants (mean age = 34.26, SD = 15.54; range 19–65 years), including 23 participants with chronic headache (mean

© 2014 European Pain Federation - EFIC®

D.E. Schoth et al.

age = 35.39, SD = 16.35; range 19–65 years) and 24 painfree, healthy control participants (mean age = 33.17, SD = 15.00; range 20–65 years). The majority of participants were female (30; 64%). Chronic headache participants reported living with headache for a mean duration of 11.4 years (SD = 10.7, range 5 months to 35 years), with the majority (15; 65%) experiencing one headache per day. Nineteen (83%) participants had tension-type headache and four (17%) had migraine. Thirteen (57%) reported at least one relative to also suffer from regular headache. As indexed by their MIDAS (Migraine Disability Assessment) scores, 12 (52%) participants indicated severe disability as a consequence of their headaches. All but three (87%) were taking regular analgesic medication for the management of their headaches.

2.2 Measures The following questionnaires were used to assess participants’ pain characteristics and emotional and physical functioning: The Anxiety Sensitivity Index (ASI; Peterson and Reiss, 1992) is a 16-item questionnaire assessing the extent to which the individual fears anxiety-related sensations, stemming from the belief that these sensations are harmful. Questions are responded to on a 5-point Likert scale (0 = very little to 4 = very much). A total score is calculated by summing all 16 items with a range of 0–64. Higher scores indicate greater fear of anxiety-related sensations. The ASI has been widely used in chronic pain research, and data support its psychometric properties, including the test–retest reliability of the total score (0.72) (Rodriguez et al., 2004). Cronbach’s alpha in the current investigation was 0.86. The Hospital Anxiety and Depression Scale (HADS (Zigmond and Snaith, 1983) rates the severity of seven symptoms of anxiety and seven symptoms of depression experienced over the past week. Scores are summed for anxiety and depression subscales, each with a range of 0–21. The HADS has been extensively used in past research, and a review of its psychometric properties has supported the internal consistency of anxiety (0.68–0.93, mean 0.82) and depression (0.67–0.90, mean 0.82) subscales (Bjelland et al., 2002). Cronbach’s alpha in the current investigation for the anxiety and depression subscales was 0.78 and 0.77, respectively. The State-Trait Anxiety Inventory (STAI; Spielberger et al., 1970) is a widely used 40-item measure of state (i.e., how the respondent currently feels) and trait (i.e., how the respondent generally feels) anxiety. Answers are selected from a 4-point Likert scale, with possible scores ranging between 20 and 80 for both state and trait subscales. Higher scores represent more intense or more frequent feelings of anxiety. A review of 45 articles reporting psychometric properties of the STAI found high levels of internal consistency (0.91 and 0.89, respectively) and test–retest reliability (0.70 and 0.88, respectively; Barnes et al., 2002). Cronbach’s alpha in the current investigation for the state and trait subscales was 0.88 and 0.93, respectively.

© 2014 European Pain Federation - EFIC®

Visual search for emotional faces in chronic headache

The McGill Pain Questionnaire-Short Form (MPQ-SF; Melzack, 1987) features a 15-item checklist assessing the sensory (11 items) and affective (4 items) dimensions of pain, which are rated on a 4-point Likert scale (0 = no pain to 4 = severe). The measure also includes two single-item measures of current pain: a 10-cm visual analogue scale (‘no pain’ to ‘worst possible pain’) and the Present Pain Intensity Index, a 6-point Likert scale (0 = no pain to 5 = excruciating). The psychometric properties of the MPQ-SF have been supported including intra-class correlation coefficients for sensory and affective subscales and present pain intensity (0.95, 0.88 and 0.75, respectively; Grafton et al., 2005). Cronbach’s alpha in the current investigation for total, sensory and affective descriptors was 0.66, 0.71 and 0.50, respectively. The MIDAS Questionnaire (Stewart et al., 2001) assesses headache-related disability over the past 3 months, asking participants to score the number of days of activity limitations due to headache over this period. A total score is calculated, which can be categorized to yield four grades of increasing disability (‘little or no disability’ to ‘severe disability’). The psychometric properties of the MIDAS have been supported, which correlates with physicians’ clinical judgements (Lipton et al., 2001) and diary-based measures (Stewart et al., 2000). The MIDAS was administered to all participants with chronic headache, regardless of headache type, which is in line with current clinical practice and research (Harpole et al., 2005; Matchar et al., 2008). Cronbach’s alpha in the current investigation was 0.76.

2.3 Experimental stimuli Pain, angry, happy and neutral colour images of male and female Caucasian models were derived from the Montréal Pain and Affective Face Clips (Simon et al., 2008), a database of 1-s video clips of models depicting emotional facial expressions. Pain expressions corresponded to prototypical description provided by Williams (2002), and angry, happy and neutral faces corresponded to prototypical descriptions detailed in the Facial Action Coding System (Ekman et al., 2002). Images were created by extracting a single frame from each video clip at the moment that the emotional expression was deemed to be at its most intense. Each image was then cropped to capture the face only (i.e., cutting out the model’s hair and background content) using an oval-shaped template. Thus, each image comprised of a facial expression within a standardized sized oval frame (i.e., 123 pixels high by 100 pixels wide). The same method was used to create practice stimuli, which featured images of surprise and neutral faces from a female model from the NimStim set of facial expressions (Tottenham et al., 2009). Example stimuli are provided in Supporting Information Fig. S1. Analysis of the valence and arousal of the experimental stimuli used in this investigation was conducted with a computerized version of the Self-Assessment Manikin (SAM; Lang, 1980). Twenty independent healthy individuals, who were not under any medication or had any diagnosed

Eur J Pain •• (2014) ••–••

3

Visual search for emotional faces in chronic headache

medical or mental health condition at the time they provided the ratings or 5 years previously (10 male, 10 female, mean age 26.15; SD = 4.46), rated a large set of emotional expressions, including those used in the current investigation. All images were randomly presented to participants for 3 s each. Following each image, two 9-point SAM scales were presented: one for valence (1 = low pleasure to 9 = high pleasure) and one for arousal (1 = low arousal to 9 = high arousal). Participants were instructed to indicate how pleasant and aroused they felt while viewing each image using the computer mouse to provide their responses. One-way repeated-measures analyses of variance (ANOVAs) with four levels (image; pain, angry, happy, neutral) were conducted on valence and arousal scores. t-tests were used in post hoc analyses to clarify significant main effects, with Bonferroni correction applied for multiple comparisons. Effect sizes in ANOVA and t-tests were quantified using partial eta-squared ηp2 and dependent Cohen’s d, respectively. Dependent Cohen’s d [note that confidence intervals (CIs) can only be calculated when dependent d lies between −2 and 2] were calculated via Exploratory Software for Confidence Intervals (ECSI; Cumming, 2012). For all analyses, the alpha level was set at 0.05, two tailed. Mean results are available online in Supporting Information Table S1. For valence, a main effect of image was found, F(2, 37) = 50.62, p < 0.001, ηp2 = 0.73. Participants felt significantly less pleasant viewing pain expressions than angry [mean difference (MD) = 1.13, SE = 0.31, p = 0.01, d = 0.88, 95% CI of d (0.32, 1.42)], happy (MD = 3.93, SE = 0.43, p < 0.001, d = 3.06) and neutral expressions [MD = 1.74, SE = 0.38, p = 0.001, d = 1.54, 95% CI of d (0.71, 2.34)]. Participants also felt significantly less pleasant viewing angry expressions than happy expressions (MD = 2.80, SE = 0.35, p < 0.001, d = 2.95) and significantly more pleasant viewing happy expressions than neutral expressions (MD = 2.18, SE = 0.20, p < 0.001, d = 2.99). For arousal, a main effect of image was found, F(3, 57) = 11.13, p < 0.001, ηp2 = 0.37. Neutral expressions were significantly less arousing than pain [MD = 2.82, SE = 0.53, p < 0.001, d = 1.70, 95% CI of d (0.86, 2.51)], angry [MD = 1.42, SE = 0.32, p = 0.002, d = 0.94, 95% CI of d (0.42, 1.43)] and happy expressions [MD = 1.67, SE = 0.45, p = 0.009, d = 1.12, 95% CI of d (0.42, 1.80)]. No other significant differences in arousal and valence were found between image categories.

D.E. Schoth et al.

image, which were embedded among pain, angry and happy distractor expressions. Images of two models (one male and one female) were included in each block. Therefore, in each block, 20 trials featured images of a male model and 20 trials featured images of a female model. Two unique models were used for pain-target/neutral-distractor and neutral-target/ pain-distractor blocks, two unique models for angry-target/ neutral-distractor and neutral-target/angry-distractor blocks, and two unique models for happy-target/neutral-distractor and neutral-target/happy-distractor blocks (i.e., the same images were used for corresponding emotion-target and neutral-target blocks). This approach was taken to avoid practice effects: Specifically, if the same models were used in each of the six experimental blocks, identical neutral stimuli would have been presented on each occasion, potentially influencing search performance across time. Participants first completed 10 practice trials, requiring them to locate a surprise expression embedded among neutral-distractor faces. For the main task, participants completed six experimental blocks including three emotiontarget blocks (pain target/neutral distractor, angry target/ neutral distractor, happy target/neutral distractor) and three neutral-target blocks (neutral target/pain distractor, neutral target/angry distractor, neutral target/happy distractor), the order of which was randomized. Each trial began with the presentation of a central fixation cross. Once participants focused their gaze on this cross, this disappeared and was replaced with eight images presented in a circular array (an example is provided in Supporting Information Fig. S2). Trials were therefore gaze dependent and only began once the central cross had been fixated. Each image was equidistant from the central fixation cross. Targetpresent trials included one target image and seven identical distractor images, the locations of which were randomized. That is, trials in emotion-target blocks featured one emotional expression embedded among seven neutral distractors; trials in neutral-target blocks featured one neutral expression embedded among seven emotional expressions. Target-absent trials in all emotion-target and neutral-target blocks included eight identical distractor images. Prior to each block, participants were informed of the target image they were searching for, and using a two button response pad (left button = target absent; right button = target present) was required to indicate as quickly as possible whether the target image was present or absent. During each trial, the circular array of images was shown until participants provided a manual response.

2.4 Visual search task The visual search task was programmed and run using SR-Research Experiment Builder (SR Research Ltd. Experiment Builder [computer programme], Mississauga, ON, Canada). This visual search task consisted of six experimental blocks of 40 trials each. Three emotion-target blocks featured pain target, angry target and happy target expressions, embedded among neutral-distractor faces. Three neutraltarget blocks featured neutral expressions as the target

4 Eur J Pain •• (2014) ••–••

2.5 Apparatus and procedure Ethical approval for this investigation was obtained from the Research Ethics Committee of the Academic Unit of Psychology, University of Southampton. Prior to inclusion in the investigation, participants gave their informed consent in compliance with the regulations of the University and the guidelines of the Helsinki Declaration. The visual search task was presented on a 21-inch P227f Viewsonic monitor (View-

© 2014 European Pain Federation - EFIC®

D.E. Schoth et al.

Sonic Corporation, Walnut, CA, USA) with a 1024 × 768 pixel resolution and a 100 Hz refresh rate. Eye movement data were recorded via an SR Research Eyelink 1000 eyetracker system running at 1000 Hz. For the detection of fixations and saccades, the recommended settings for cognitive experiments in the EyeLink system were used (i.e., a velocity threshold of 30°/s, an acceleration threshold of 8000°/s2 and a motion threshold (i.e., the distance the eye needs to move before the motion is classified as a saccade) of 0.15°; Stampe, 1993). A minimal track status of 80% was used. A 9-point calibration procedure was used, which was accepted when the average calibration error was lower than 1° of visual angle, with none of the individual nine points having greater error than 1°. A drift check was conducted prior to each trial, and recalibration was performed when necessary. A desk-mounted chin rest was used to reduce participant’s head movements, with participant’s eyes approximately 17 inches from the monitor. Participants first completed the visual search task, followed by the self-report measures. To control for potential order effects, questionnaires were presented in a new random order to each participant. The total experimental duration was approximately 90 min.

2.6 Data reduction and analytic plan Fixation data were extracted using SR Research DataViewer and then processed and summarized using R (The R Development Core Team, 2011). All analyses were conducted in PASW Statistics 18.0 for Windows (SPSS Inc., PASW Statistics for Windows, Version 18.0 [computer software, released 2009], Chicago, IL, USA), except effect sizes and their associated 95% CIs, which were calculated with ECSI (Cumming, 2012). For the visual search task, practice and incorrect experimental trials were excluded from final analyses. Mean manual response times were calculated, and trials with any responses more than three standard deviations away from this individual mean were removed as outliers. Finally, fixations less than 60 ms or greater than 1200 ms were removed as outliers, which resulted in the removal of less than 1% of fixations. Between-groups differences for demographic characteristics and emotional functioning were explored via t-tests and χ2 for continuous and categorical variables, respectively. A series of 2 × 3 × 2 ANOVAs examined the visual search data from target-present trials, with group (chronic headache, healthy control) as a between-subjects independent variable and image condition (pain, angry, happy) and target type (emotional target, neutral target) as within-subjects independent variables. ANOVAs and t-tests were used where needed in post hoc analyses to clarify significant main effects and interactions. Effect sizes for ANOVA and t-tests were quantified using partial eta-squared ηp2 and Cohen’s d, respectively. Cohen (1988) considered d = 0.2, d = 0.5 and d = 0.8 to be small, medium and large effect sizes, respectively. Independent Cohen’s d were calculated for all

© 2014 European Pain Federation - EFIC®

Visual search for emotional faces in chronic headache

independent-samples t-tests (i.e., for chronic headache vs. healthy control comparisons), which is defined as the difference between the two means divided by the pooled standard deviation (Cohen, 1988; Rosenthal and Rosnow, 1991). Dependent Cohen’s d were calculated for all dependent t-tests (i.e., for image condition and target type comparisons), which is defined as the difference between the two means divided by the standard deviation of the MD, taking into account the correlation between the two variables (Cohen, 1988). One-sample Cohen’s d were calculated for one-sample t-tests (i.e., probability of initial fixations on target images compared with chance), which is defined as the difference between the observed mean and the hypothetical mean divided by the standard deviation (Cohen, 1988). For ANOVA analyses, the alpha level was set at 0.05, two tailed. The alpha level was set at 0.01 for correlation coefficients due to the large number of calculations conducted. Two groups of analyses were conducted. Performance analyses examined the behavioural responses of participants including: (1) response accuracy and (2) mean response time from array onset until manual button response. Behavioural responses were analysed and are presented here because eye movement data were only analysed from correct trials, and therefore it is important to demonstrate whether the pattern of incorrect responses differed according to participant group or target type. Attentional orienting analyses were conducted on: (1) proportion of initial fixations on the target face and (2) localization time (the time taken to fixate the target face following array onset).

3. Results 3.1 Group characteristics The chronic headache and healthy control groups did not differ significantly in sex ratio [chronic headache group: 17 (74%) female, control group: 13 (54%) female, χ2 = 1.98, p = 0.159] or age [chronic headache group: 35.39 (SD = 16.35), control group: 33.17 (SD = 15.00), t(45) = 0.49, p = 0.629, d = 0.14, 95% CI of d (−0.43, 0.71)]. As expected, significantly more headache days per month were reported by the chronic headache group than the control group [chronic headache group: 20.43 (SD = 5.62), control group: 1.54 (SD = 1.67), t(26) = 15.47, p < 0.001, d = 4.60, 95% CI of d (3.49, 5.70)]. Descriptive statistics (means and standard deviations) of questionnaire data are available in Supporting Information Table S2. A series of independent t-tests were conducted on measures administered to both groups. The chronic headache group reported significantly higher anxiety and depression compared with the control group. These variables were not included as covariates in the following analyses, however, as they did not correlate Eur J Pain •• (2014) ••–••

5

6 Eur J Pain •• (2014) ••–••

1014.24 (301.93) 952.68 (248.46) 0.111 (0.08) 0.160 (0.11) 1728.76 (405.67) 0.931 (0.06) 0.924 (0.07)

1684.33 (514.13)

1132.58 (367.10) 1050.01 (319.06) 0.121 (0.07) 0.104 (0.08) 1949.42 (604.30) 0.888 (0.08) 0.900 (0.11)

1900.61 (605.06)

864.77 (255.22) 839.38 (200.48) 0.141 (0.08) 0.146 (0.08) 1425.52 (366.67) 0.946 (0.08) 0.944 (0.05)

1411.00 (421.90)

847.70 (287.76) 832.51 (259.46) 0.146 (0.07) 0.140 (0.09) 1417.99 (403.78) 0.957 (0.05) 0.958 (0.06)

1437.43 (487.10)

1136.75 (504.15) 1122.02 (247.88) 0.113 (0.07) 0.104 (0.08) 1857.59 (751.14) 0.915 (0.07) 0.933 (0.06)

1788.06 (466.37)

581.18 (224.34) 0.200 (0.07) 0.319 (0.17) 1074.69 (318.73) 1031.22 (428.58) 0.965 (0.06) 0.983 (0.03)

Chronic headache Healthy control Chronic headache Healthy control Chronic headache Chronic headache

Healthy control

Mean response time (ms)

Note: Data are from target-present trials only.

The attentional orienting hypothesis was first explored by analysing the proportion of trials where the first

Pain target in neutral distractors Angry target in neutral distractors Happy target in neutral distractors Neutral target in pain distractors Neutral target in angry distractors Neutral target in happy distractors

3.3 Attentional orienting analyses

Response accuracy (proportion correct)

Descriptive statistics (means and standard deviations) for all behavioural and eye movement data are presented in Table 1. Examination of response accuracy revealed a significant main effect of image condition, F(2, 90) = 16.88, p < 0.001, ηp2 = 0.273; participants’ responses were less accurate in trials with angry expressions than trials with pain expressions [MD = 0.05, p < 0.001, d = 0.84, 95% CI of d (0.48, 1.19)] and happy expressions [MD = 0.034, p = 0.003, d = 0.57, 95% CI of d (0.23, 0.90)]. A significant main effect of target type was also found, F(1, 45) = 29.21, p < 0.001, ηp2 = 0.394; participants’ responses were more accurate in emotional-target trials than neutraltarget trials [MD = 0.03, d = 0.61, 95% CI of d (0.35, 0.86)]. No significant interactions were found. Analysis of manual mean response times revealed significant main effects of image condition, F(2, 66) = 123.22, p < 0.001, ηp2 = 0.732, and target type, F(1, 45) = 67.06, p < 0.001, ηp2 = 0.598. These effects were qualified by a significant image condition by target type interaction, F(2, 79) = 10.63, p < 0.001, ηp2 = 0.191. Post hoc t-tests showed participants were significantly faster to respond to emotional targets than neutral targets for trials with pain expressions, t(46) = 12.50, p < 0.001, d = 0.96, 95% CI of d (0.71, 1.20), and happy expressions, t(46) = 5.52, p < 0.001, d = 0.62, 95% CI of d (0.37, 0.87), while the difference for trials with angry expressions approached significance, t(46) = 1.99, p = 0.053, d = 0.17, 95% CI of d (−0.002, 0.33). To clarify this finding, MDs in response times were calculated between emotion-target and neutral-target trials. Differences were significantly shorter for trials with angry expressions compared with trials with pain expressions [MD = 263.03, t (46) = 2.63, p < 0.001, d = 0.91, CI of d (0.51, 1.32)] and happy expressions [MD = 177.54, t (46) = 1.65, p = 0.012, d = 0.51, CI of d (0.11, 0.90]. Aside from these results, no other significant interactions were found.

Table 1 Means (SD) for behavioural and eye movement data as a function of emotion type, target type and participant group.

3.2 Performance analyses

Proportion of initial fixations on target

Localization time of target (ms)

significantly with the dependent variables (Tabachnick and Fidell, 2012). Correlations between the self-report measures with performance and eye movement measures for chronic headache and healthy control groups separately are available online in Supporting Information Table S3. With an adopted alpha of 0.01, no significant correlations were found for either group.

620.77 (217.17)

D.E. Schoth et al.

Healthy control

Visual search for emotional faces in chronic headache

© 2014 European Pain Federation - EFIC®

D.E. Schoth et al.

image fixated was the target (proportion of initial fixations). Significant main effects were found for group, F(1, 45) = 5.75, p = 0.021, ηp2 = 0.113, image condition, F(2, 90) = 22.13, p < 0.001, ηp2 = 0.330, and target type, F(1, 45) = 8.50, p = 0.006, ηp2 = 0.16. These effects were qualified by a significant image condition by target type interaction, F(1, 45) = 17.02, p < 0.001, ηp2 = 0.274, and a significant image condition by target type by group three-way interaction, F(2, 90) = 6.49, p = 0.002, ηp2 = 0.126. To explore these interactions, two separate 3 × 2 ANOVAs were conducted for the chronic headache and healthy control groups separately. A significant image condition by target type interaction was found for the chronic headache group only, F(2, 44) = 17.55, p < 0.001, ηp2 = 0.444. To explore this further, separate one-way ANOVAs were conducted for emotion-target and neutral-target trials. A significant effect of image condition was found for emotional-target trials only, F(2, 44) = 26.80, p < 0.001, ηp2 = 0.55; pairwise comparisons revealed significantly more initial fixations towards target pain expressions than target angry expressions [MD = 0.22, p < 0.001, d = 1.35, CI of d (0.89, 1.81)] and target happy expressions [MD = 0.18, p < 0.001, d = 1.02, CI of d (0.57, 1.46)]. Furthermore, paired-samples t-tests were conducted comparing performance on emotion-target and neutral-target trials. For trials with pain expressions, chronic headache participants made significantly more initial fixations to emotional targets than neutral targets, t(22) = 4.27, p < 0.001, d = 1.01, CI of d (0.53, 1.48). To further explore the significant three-way interaction, independent-samples t-tests were conducted to compare performance between chronic headache and healthy control groups. Chronic headache participants made significantly more initial fixations to target pain expressions than healthy control participants, t(30) = 3.12, p = 0.004, d = 0.93, CI of d (0.32, 1.53). No other significant differences were found. In addition, probability scores were compared with chance for both chronic headache and healthy control groups separately. As there were eight images in each array, there was a 0.125 probability of participants initially fixating the target expression by chance alone. Proportions of initial fixations on target expressions were therefore compared with 0.125 via one-sample t-tests. Significantly greater bias towards target pain expressions than expected by chance was shown by both the chronic headache group, t(22) = 5.59, p < 0.001, d = 1.17, CI of d (0.63, 1.70), and the healthy control group, t(23) = 4.96, p < 0.001, d = 1.03, CI of d (0.53, 1.52). As reported above, this bias was significantly © 2014 European Pain Federation - EFIC®

Visual search for emotional faces in chronic headache

more pronounced in the chronic headache group than the healthy control group. Aside from these results, no other significant interactions were found. Analysis of localization time revealed significant main effects of image condition, F(2, 71) = 111.06, p < 0.001, ηp2 = 0.712, and target type, F(1, 45) = 45.96, p < 0.001, ηp2 = 0.505. These main effects were qualified by a significant image by target interaction, F(2, 78) = 24.64, p < 0.001, ηp2 = 0.354. A series of paired-samples t-tests were conducted between emotion-target and neutral-target trials for angry, pain and happy expressions. Participants were significantly faster to locate emotional targets than neutral targets for trials with pain expressions, t(46) = 11.75, p < 0.001, d = 1.12, CI of d (0.82, 1.42), and happy expressions, t(46) = 4.96, p < 0.001, d = 0.53, CI of d (0.29, 0.76). Aside from these results, no other significant interactions were found.

4. Discussion and conclusions This is the first study to explore pain-related attentional biases in chronic pain via a visual search task with concurrent recording of eye movement behaviours. Supporting our hypothesis, participants with chronic headache, compared with healthy controls, showed a significantly higher proportion of initial fixations to target pain expressions. The chronic headache group also showed significantly greater bias towards target pain expressions when compared with chance. These results show attentional biases for pain-related information influence visual search. Biases also exist in the presence of multiple stimuli competing for attention (i.e., displays of eight images), and therefore extend former visual-probe studies that have identified pain-related biases relative to a single neutral stimulus (e.g., Liossi et al., 2009; Sharpe et al., 2009; Schoth and Liossi, 2013), along with a recent visual scanning task providing evidence of attentional orienting bias for pain expressions presented concurrently with three other emotionally salient faces (Liossi et al., 2014). Attentional biases for pain-related information in individuals with chronic pain have been demonstrated in a range of experimental paradigms, supporting theoretical accounts of pain and attention (e.g., Pincus and Morley, 2001). In line with recent research (i.e., Schoth and Liossi, 2013; Liossi et al., 2014), evidence for the specificity of bias was found in the present study, as between-groups differences were found only for target pain expressions, and not target angry or happy expressions. Within-groups analysis also revealed that, for chronic headache participants only, Eur J Pain •• (2014) ••–••

7

Visual search for emotional faces in chronic headache

the proportion of initial fixations to target pain expressions was significantly higher than that to angry and happy target expressions. Evidence of bias specificity has theoretical and clinical implications, particularly for ABM, which is a theory-driven intervention that aims to train patient attention away from threatening information towards neutral information (Bar-Haim, 2010). While preliminary evidence is supportive of this intervention in chronic pain (Carleton et al., 2011; Sharpe et al., 2012; Schoth et al., 2013), the optimal form of ABM has yet to be established, including the most appropriate stimuli to train patients with. However, specifically tailored stimuli are considered essential in ABM to maximize treatment efficiency (Bar-Haim, 2010). Considering chronic pain populations without other medical or psychiatric comorbidities, the present results suggest training with painful expressions may be preferable to training with other threat-related stimuli such as angry expressions, as a lack of specific bias towards angry faces (i.e., the present study; Liossi et al., 2014) and angerrelated words (i.e., Liossi et al., 2011) in chronic pain groups relative to healthy control groups has been reported. A recent investigation using a linguistic visual-probe task with eye tracking found significantly shorter initial fixation durations on health catastrophe words in those with chronic pain compared with healthy controls, although no difference in initial orienting for sensory pain or health catastrophe words was reported between the two groups (Yang et al., 2013). Pain expressions alternatively are particularly salient, with both participant groups in the present study showing significantly greater attentional orienting to pain expressions than expected by chance (as noted, this effect was significantly greater in the chronic headache group). Facial expressions of pain convey to others messages of threat and harm, and are also a social signal for help (Williams, 2002). It is unsurprising therefore that pain expressions rapidly capture attention. Furthermore, numerous potential triggers have been connected with headache, and identification and avoidance of such triggers has long been associated with their clinical management (Martin and MacLeod, 2009). Heightened vigilance in those with chronic headache for external pain-related cues and information may also be expected based upon search behaviours such as these. Considering healthy individuals, a recent visual-probe study using the same stimulus set (i.e., Simon et al., 2008) also reported early attentional engagement with pain expressions (Baum et al., 2013). Analysis of localization time was not in line with our predictions, as no significant differences were found 8 Eur J Pain •• (2014) ••–••

D.E. Schoth et al.

between chronic headache and healthy control groups in the time taken to localize target pain expressions. This could be due to the fact that both groups were more likely to orient their initial fixations towards such expressions than expected by chance. Supporting this, overall results revealed that both groups combined were significantly faster to localize pain target expressions than neutral-target expressions embedded among pain expression distractors: This again supports the saliency of pain expressions for all participants. Furthermore, participants were also significantly faster to localize happy target expressions than neutraltarget expressions embedded among happy expression distractors, which supports a preference for happy over neutral faces. In line with these findings, participants were also significantly faster providing manual responses to emotional targets than neutral targets for both pain and happy trials. Emotional faces are a rich source of information, providing an important indication of others’ emotional states and intentions (Frischen et al., 2008). Emotional faces are therefore both biologically and socially significant, and likely to receive enhanced processing. In their review of face perception and attention, Palermo and Rhodes (2007) provide support for this assertion, showing faces, especially those depicting emotional expressions, to be rapidly detected, which may be an obligatory process. In contrast to the above, no significant differences in localization time and mean manual response times were found between angry target expressions and neutral-target expressions embedded among angry expression distractors (although this latter result approached significance). This may be attributable to reductions in accuracy: Overall, participants were significantly less accurate responding to trials with angry expressions than trials with happy and pain expressions. As threat captures attention, differences between angry and happy expressions are not surprising. Furthermore, pain and angry expressions are associated with different temporal and spatial distributions of brain responses (González-Roldan et al., 2011). While both portray threat, pain expressions are more ambiguous in terms of the source of potential danger (a noxious stimulus may be internal or external); in contrast, the source of danger from an angry expression is less ambiguous, being the individual expressing anger. As a result of these differences, it is possible that participants were more distracted by angry faces than pain or happy faces, resulting in reduced response accuracy. An alternative possibility is that participants simply found it more difficult to discriminate angry expressions from their neutral counterparts in this experiment, resulting in overall © 2014 European Pain Federation - EFIC®

D.E. Schoth et al.

reductions in accuracy across both participant groups. This is an important consideration for future visual search research, and clarification may be sought by asking participants to provide difficulty ratings for each search condition. The study has many strengths including the use of a novel cognitive paradigm combined with eye tracking technology. Furthermore, both threat-related (i.e., angry) and pleasant stimuli (i.e., happy) were included that were matched to pain stimuli on arousal value. This is necessary in order to demonstrate the specificity of attentional bias to pain in chronic pain. However, a number of limitations may also be raised. First, because only individuals with chronic headache and without comorbid psychiatric conditions were recruited, caution should be taken in generalizing these results to patients with other forms of chronic pain and comorbidities. Research has yet to address whether patterns of bias differ between patients with and without comorbid conditions, although this is likely an important consideration given the growing interest in therapeutic benefits of modifying attentional biases. Second, the facial expressions used were simulated rather than genuine (however, all expressions corresponded to prototypical descriptions), and were derived from video clips (Simon et al., 2008), selected when deemed at their most intense. Future research should therefore explore whether attentional biases for facial expressions of pain are also found with more subtle expressions, and whether the experience of chronic pain makes one more or less sensitive to pain expressions in others. A third limitation is the recruitment of a relatively small sample size, which precludes the statistical detection of small effect sizes. The results from Liossi et al. (2014) and the present study support the continued use of eye tracking in the exploration of attentional biases in chronic pain populations using facial expressions. As an emerging area, there are numerous potential avenues for future research. First, as discussed, pain-related attentional biases exist even in the presence of multiple stimuli competing for attention. To our knowledge, no previous research has used real-world scenes (e.g., images depicting individuals with injury and pain) to explore attentional biases, however, although the use of such stimuli in combination with eye tracking would provide more detailed information on the pattern and time course of bias. Second, research to date has only used static images of faces, whereas in real life facial expressions are dynamic whereby coordinated muscle contractions transpire over time. Research suggests that dynamic expressions are recognized more accurately and faster than static expressions (Recio et al., © 2014 European Pain Federation - EFIC®

Visual search for emotional faces in chronic headache

2011). Future investigation may therefore compare attentional biases for static and dynamic facial expressions of pain. The results of such research are likely to be informative for ABM interventions, a promising but still in its infancy therapeutic possibility for chronic pain patients. Should biases be more pronounced to dynamic faces, training with such stimuli while taking into account the pattern of biases (e.g., vigilance vs. avoidance) shown by specific clinical groups (e.g., acute vs. chronic), and the relationship of this pattern with pain outcomes, may yield greater therapeutic gains. In conclusion, the current study shows individuals with chronic headache to demonstrate facilitated attentional orienting towards pain expressions specifically when used as targets in a visual search task. Attentional biases influence visual search, and are present when multiple stimuli competing for attention are presented simultaneously. This study builds upon recent investigations, supporting the continued use of eye tracking to explore biases for pain-related information. Author contributions All listed authors contributed to the writing of this manuscript, and all discussed the results and commented on the manuscript in this form.

References Bar-Haim, Y. (2010). Research review: Attention bias modification (ABM): A novel treatment for anxiety disorders. J Child Psychol Psychiatry 51, 859–870. Barnes, L.L.B., Harp, D., Jung, W.S. (2002). Reliability generalization of scores on the Spielberger State-Trait Anxiety Inventory. Educ Psychol Meas 62, 603–618. Baum, C., Schneider, R., Keogh, E., Lautenbacher, S. (2013). Different stages in attentional processing of facial expressions of pain: A dotprobe task modification. J Pain 14, 223–232. Bjelland, I., Dahl, A.A., Tangen, T., Necklemann, D. (2002). The validity of the Hospital Anxiety and Depression Scale: An updated literature review. J Psychosom Res 52, 69–77. Carleton, R.N., Richter, A.A., Asmundson, G.J.G. (2011). Attention modification in persons with fibromyalgia: A double blind, randomized clinical trial. Cogn Behav Ther 40, 279–290. Chun, M.M. (2003). Scene perception and memory. In Cognitive Vision, D.E. Irwin, ed. (London: Academic Press) pp. 79–109. Cohen, J. (1988). Statistical Power Analysis for the Behavioural Sciences, 2nd edn (Hillsdale, NJ: Lawrence Earlbaum Associates). Crombez, G., Van Ryckeghem, D.M., Eccleston, C., Van Damme, S. (2013). Attentional bias to pain-related information: A meta-analysis. Pain 154, 497–510. Cumming, G. (2012). Understanding the New Statistics: Effect Sizes, Confidence Intervals, and Meta-analysis (New York: Routledge). Dear, B.F., Sharpe, L., Nicholas, M., Refshauge, K. (2011). The psychometric properties of the dot-probe paradigm when used in pain-related attentional bias research. J Pain 12, 1247–1254. Duc, A.H., Bays, P., Husain, M. (2008). Eye movements as a probe of attention. Prog Brain Res 171, 403–411.

Eur J Pain •• (2014) ••–••

9

Visual search for emotional faces in chronic headache

Ekman, P., Friesen, W.V., Hager, J.C. (2002). Facial Action Coding System (FACS) (Salt Lake City, UT: A Human Face). Frischen, A., Eastwood, J.D., Smilek, D. (2008). Visual search for faces with emotional expressions. Psychol Bull 134, 662–676. González-Roldan, A.M., Martínez-Jauand, M., Muñoz-García, M.A., Sitges, C., Cifre, I., Montoya, P. (2011). Temporal dissociation in the brain processing of pain and anger faces with different intensities of emotional expression. Pain 152, 853–859. Grafton, K., Foster, N., Wright, C. (2005). Test-retest reliability of the Short-Form McGill Pain Questionnaire: Assessment of intraclass correlation coefficients and limits of agreement in patients with osteoarthritis. Clin J Pain 21, 73–82. Harpole, L.H., Samsa, G.P., Matchar, D.B., Silberstein, S.D., Blumenfeld, A., Jurgelski, A.E. (2005). Burden of illness and satisfaction with care among patients with headache seen in a primary care setting. Headache 45, 1048–1055. Headache Classification Subcommittee of the International Headache Society (2004). The International Classification of Headache Disorders. Cephalalgia 24, 9–160. Khatibi, A., Dehghani, M., Sharpe, L., Asmundson, G.J.G., Pouretemad, H. (2009). Selective attention towards painful faces among chronic pain patients: Evidence from a modified version of the dot-probe. Pain 142, 42–47. Lang, P.J. (1980). Behavioral treatment and bio-behavioral assessment: computer applications. In Technology in Mental Health Care Delivery Systems, J.B. Sidowski, J.H. Johnson, E.A. Williams, eds. (Norwood, NJ: Ablex) pp 119–137. Liossi, C. (2012). Attentional biases in chronic pain: Do they exist and does it really matter? Pain 153, 9–10. Liossi, C., Schoth, D.E., Bradley, B.P., Mogg, K. (2009). Time course of attentional bias for pain-related cues in chronic daily headache sufferers. Eur J Pain 13, 963–969. Liossi, C., White, P., Schoth, D.E. (2011). Time-course of attentional bias for threat-related cues in patients with chronic daily headache–tension type: Evidence for the role of anger. Eur J Pain 15, 92–98. Liossi, C., Schoth, D.E., Godwin, H.J., Liversedge, S.P. (2014). Using eye movements to investigate selective attention in chronic daily headache. Pain 155, 503–510. Lipton, R.B., Stewart, W.F., Sawyer, J., Edmeads, J.G. (2001). Clinical utility of an instrument assessing migraine disability: The Migraine Disability Assessment (MIDAS) questionnaire. Headache 41, 854– 861. Liversedge, S.P., Findlay, J.M. (2000). Saccadic eye movements and cognition. Trends Cogn Sci 1, 6–14. Martin, P.R., MacLeod, C. (2009). Behavioral management of headache triggers: Avoidance of triggers is an inadequate strategy. Clin Psychol Rev 29, 483–495. Matchar, D.B., Harpole, L., Samsa, G.P., Jurgelski, A., Lipton, R.B., Silberstein, S.D., Young, W., Kori, S., Blumenfeld, A. (2008). The headache management trial: A randomized study of coordinated care. Headache 48, 1294–1310. Melzack, R. (1987). The short-form McGill pain questionnaire. Pain 30, 191–197. Palermo, R., Rhodes, G. (2007). Are you always on my mind? A review of how face perception and attention interact. Neuropsychologia 45, 75–92. Peterson, R.A., Reiss, S. (1992). Anxiety Sensitivity Index Manual, 2nd edn (Worthington, OH: International Diagnostic Systems). Pincus, T., Morley, S. (2001). Cognitive-processing bias in chronic pain: A review and integration. Psychol Bull 127, 599–617. Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research. Psychol Bull 124, 372–422. Recio, G., Sommer, W., Schacht, A. (2011). Electrophysiological correlates of perceiving and evaluating static and dynamic facial emotional expressions. Brain Res 1376, 66–75. Rodriguez, B.F., Bruce, S.E., Pagano, M.E., Spencer, M.A., Keller, M.B. (2004). Factor structure and stability of the Anxiety Sensitivity Index in a longitudinal study of anxiety disorder patients. Behav Res Ther 42, 79–91. Rosenthal, R., Rosnow, R.L. (1991). Essentials of Behavioral Research: Methods and Data Analysis, 2nd edn (New York: McGraw-Hill).

10 Eur J Pain •• (2014) ••–••

D.E. Schoth et al.

Schoth, D.E., Delgado Nunes, V., Liossi, C. (2012). Attentional bias towards pain-related information in chronic pain; a meta-analysis of visual-probe investigations. Clin Psychol Rev 32, 13–25. Schoth, D.E., Georgallis, T., Liossi, C. (2013). Attentional bias modification in people with chronic pain: A proof of concept study. Cogn Behav Ther 42, 233–243. Schoth, D.E., Liossi, C. (2013). Specificity and time-course of attentional bias in chronic headache: A visual-probe investigation. Clin J Pain 29, 583–590. Sharpe, L. (2012). Attentional biases and their role in the management of pain. Pain 153, 2307–2308. Sharpe, L., Dear, B.F., Schreiber, L. (2009). Attentional biases in chronic pain associated with rheumatoid arthritis: Hypervigilance or difficulties disengaging? J Pain 10, 329–335. Sharpe, L., Ianiello, M., Dear, B.F., Nicholson Perry, K., Refshauge, K., Nicholas, M.K. (2012). Is there a potential role for attention bias modification in pain patients? Results of 2 randomised, controlled trials. Pain 153, 722–731. Simon, D., Craig, K.D., Gosselin, F., Belin, P., Rainville, P. (2008). Recognition and discrimination of prototypical dynamic expressions of pain and emotions. Pain 135, 55–64. Spielberger, C.D., Gorsuch, R.L., Lushene, R.E. (1970). State Trait Anxiety Inventory (Palo Alto, CA: Consulting Psychologists Press). Stampe, D.M. (1993). Heuristic filtering and reliable calibration methods for video based pupil-tracking systems. Behav Res Methods Instrum Comput 25, 137–142. Stewart, W.F., Lipton, R.B., Dowson, A.J., Sawyer, J. (2001). Development and testing of the Migraine Disability Assessment (MIDAS) Questionnaire to assess headache-related disability. Neurology 56 (Suppl. 1), S20–S28. Stewart, W.F., Lipton, R.B., Kolodner, K.B., Sawyer, J., Lee, C., Liberman, J.N. (2000). Validity of the Migraine Disability Assessment (MIDAS) score in comparison to a diary-based measure in a population sample of migraine sufferers. Pain 88, 41–52. Tabachnick, B.G., Fidell, L.S. (2012). Using Multivariate Statistics, 6th edn (Boston, MA: Pearson). Tottenham, N., Tanaka, J.W., Leon, A.C., McCarry, T., Nurse, M., Hare, T.A., Marcus, D.J., Westerlund, A., Casey, B.J., Nelson, C. (2009). The NimStim set of facial expressions: Judgments from untrained research participants. Psychiatry Res 168, 242–249. Van Damme, S., Legrain, V., Vogt, J., Crombez, G. (2010). Keeping pain in mind: A motivational account of attention to pain. Neurosci Biobehav Rev 34, 204–213. Van Ryckeghem, D.M.L., Crombez, G. (2014). Attentional bias and chronic pain: Where to go from here? Pain 155, 6–7. Williams, A.C. (2002). Facial expression of pain: An evolutionary account. Behav Brain Sci 25, 439–455. Yang, Z., Jackson, T., Chen, H. (2013). Effects of chronic pain and painrelated fear on orienting and maintenance of attention: An eye movement study. J Pain 14, 1148–1157. Zigmond, A.S., Snaith, R.P. (1983). The hospital anxiety and depression scale. Acta Psychiatr Scand 67, 361–370.

Web references The R Development Core Team (2011). R: A language and environment for statistical computing. R Foundation for Statistical Computing [computer software] (Vienna: Austria). Retrieved from http://www.R-project.org/ (accessed 1 February 2011).

Supporting Information Additional Supporting Information may be found in the online version of this article at the publisher’s web-site: Figure S1. Pain, angry, happy and neutral counterpart images used in the visual search task.

© 2014 European Pain Federation - EFIC®

D.E. Schoth et al.

Figure S2. Example pain-target/neutral-distractor circular array showing a participant’s scan path and fixation durations beginning from the central drift correct (not displayed). Table S1. Mean (SD) valence and arousal image ratings from the Self-Assessment Manikin (n = 20). Table S2. Mean (SD) for questionnaire measures completed by chronic headache and healthy control groups.

© 2014 European Pain Federation - EFIC®

Visual search for emotional faces in chronic headache

Table S3. Pearson correlation coefficients between performance and eye movement measures with questionnaire measures for chronic headache (n = 23) and healthy control participants (n = 24). Appendix S1. Supplementary results 1.

Eur J Pain •• (2014) ••–••

11