J Behav Med (2007) 30:455–469 DOI 10.1007/s10865-007-9127-2

Anger Suppression, Ironic Processes and Pain Phillip J. Quartana Æ K. Lira Yoon Æ John W. Burns

Accepted: July 31, 2007 / Published online: 22 August 2007
Springer Science+Business Media, LLC 2007

Abstract Whether anger suppression exerts a causal influence on pain experience, and the mechanisms of such an influence, are not well understood. We report two experimental studies that examine the hypothesis that anger suppression paradoxically increases cognitive accessibility of anger, in turn coloring perceptions of succeeding pain in an anger-congruent fashion. The results of two experimental studies largely confirmed these predictions. Study 1 revealed that participants instructed to suppress emotions during anger-provocation experienced greater cold-pressor pain than those in the control condition. This difference was confined to perception of anger-specific qualities of pain. Study 2 replicated key findings of Study 1, but also provided partial evidence for increased cognitive accessibility of anger tied to anger suppression through self-report and modified dot-probe methodologies. Implications and limitations of these findings are discussed. Keywords Pain

Anger
Anger suppression
Ironic processes


Pain severity is related to negative emotional traits and states such as anxiety, depression, and anger (Bonica 1990; Keefe et al. 2004). Although relationships of pain with anxiety and depression have been widely studied, recent

P. J. Quartana (&)
J. W. Burns Department of Psychology, Rosalind Franklin University of Medicine and Science, 3333 Green Bay Road, North Chicago, IL 60064, USA e-mail: [email protected] K. L. Yoon Department of Psychology, University of Miami, Coral Gables, FL, USA

work suggests that the arousal of anger in particular may amplify pain and concomitant physiological reactivity to a greater extent than the arousal of other negative emotions (Berkowitz 1990, 1993; Berkowitz and Harmon-Jones 2004; Burns 2006a; Greenwood et al. 2003; Janssen et al. 2001; Fernandez and Turk 1995; Okifuji et al. 1999). Thus, the connection between anger and pain may be strong and unique, and not completely reducible to effects of a common substrate of negative affect. Whereas patients with various chronic pain disorders are characterized by high levels of trait anger and hostility, many findings implicate anger regulation—primarily the suppression or expression of anger—as a particularly robust determinant of chronic pain severity (e.g., Bruehl et al. 2002, 2003a, b; Kerns et al. 1994). As in the broader literature on the regulation of emotion (e.g., Gross 2002), much attention in pain research has focused on the suppression or inhibition of anger (Greenwood et al. 2003). Studies suggest that inhibited anger is more prevalent among chronic pain patients than among healthy individuals, and that, among pain patients, suppression of anger and hostility are related to poor adjustment, greater pain severity, exaggerated symptom-specific reactivity to stress and pain, and attenuated benefit from multidisciplinary pain treatment (Bruehl et al. 2003a, b; Burns 1997; Burns et al. 1998, 2006; Gelkopf 1997; Kerns et al. 1994; Pilowsky and Spence 1976; Janssen et al. 2001). It is important to keep in mind that a majority of studies of anger suppression and pain have relied implicitly on psychodynamic notions that habitual suppression of strong negative emotions may somehow be ‘‘converted’’ into physical symptoms, including pain (e.g., Engel 1959; Frued 1915/1957; Pilowsky and Spence 1976). Such theories do not provide well-specified accounts of cognitive, emotional, and physiological mechanisms by which suppressed

123

456

anger may affect pain. Further, recent studies have relied almost exclusively on self-reported measures of trait ‘‘anger-in,’’ (Spielberger et al. 1985) and then examined correlations between anger-in scores and either selfreported chronic pain severity or acute pain induced in the laboratory (e.g., Burns et al. 2006; Kerns et al. 1994; Janssen et al. 2001; Gelkopf 1997). This approach is limited in that measures of ‘‘anger-in’’ are related moderately to other trait negative affects, and so observed effects of ‘‘anger-in’’ on pain indexes may reflect variance held in common with measures of negative affectivity (e.g., neuroticism, Martin et al. 1999). Finally, most studies have not used experimental methods to manipulate anger arousal, anger regulation, and pain-induction. As a result, it remains unclear whether the actual suppression of anger in the context of anger arousal exerts an effect on perception of subsequent painful stimuli. To fill the many inferential gaps between purported cause (anger suppression) and effect (amplified pain) left open by existing research, a well-articulated theoretical and methodological framework is needed from which empirically testable hypotheses can be derived and by which results may be interpreted. We (Quartana and Burns 2007) recently proposed an ironic process model of anger suppression and pain that is based on Wegner’s (1994) ironic process theory of mental control. According to Wegner’s (1994) ironic process theory, efforts to suppress thoughts, emotions or behaviors require the operation of two cognitive search processes. First, a resource-dependent operating process works to eliminate unwanted stimuli from conscious awareness. Second, a monitoring process operates in an automatic, resourceindependent fashion, and scans for awareness of to-besuppressed stimuli. The irony of the monitoring process is that by searching ceaselessly and effortlessly for evidence of unwanted mental contents, those contents become highly accessible to conscious awareness. Evidence for ironic effects of suppression has been uncovered in multiple domains of psychological inquiry (for reviews, see Abramowitz et al. 2001; Wegner 1994; Wenzlaff and Wegner 2000). Further, highly accessible mental content appears to influence perceptions of succeeding events (DeSteno et al. 2004; Forgas 1995; Schwarz 1990), even when this material does not fully reach conscious awareness (Bargh and Pietromonaco 1982; Mayer and Merckelbach 1999; Zajonc 2000). Insofar as efforts to suppress a stimulus (e.g., anger) increase cognitive accessibility of themes related to the tobe-suppressed stimulus (e.g., yelling at a provocateur), such suppression should exert a measurable influence on perceptions of, and responses to subsequent events. For instance, attempts to suppress awareness of pain and suffering at one point appear to foster strong perceptions of

123

J Behav Med (2007) 30:455–469

physical discomfort and heightened autonomic responses during subsequent benign sensory events (Cioffi and Holloway 1993), stressful (non-pain) events (Burns 2006b) as well as stressful, painful events (Sullivan et al. 1997). Our ironic process model of anger suppression holds that increased accessibility of to-be-suppressed material represents a cognitive–affective mechanism that links the attempt to suppress anger with increased intensity of later pain. Thus, efforts to suppress anger in the context of anger arousal should render anger-related thoughts and feelings highly accessible to conscious awareness, thereby increasing sensitivity to a subsequent painful stimulus. Based on mood-congruency hypotheses (Forgas 1995; DeSteno et al. 2004; Schwarz 1990; Zajonc 2000), we further assert that highly accessible mental content may influence perception of events in a manner that is highly congruent with the to-be-suppressed material. Specifically, perceptions of pain following anger suppression should be colored by anger-specific qualities of the noxious stimulus (e.g., irritating) to a greater extent than the sensory (e.g., throbbing) and generally distressing (e.g., fearful) aspects (Quartana and Burns 2007). Moreover, it may be the case that anger-content highly accessible to consciousness may impact pain perception more so than other negative emotional content (e.g., anxiety). On one level, work by Berkowitz and colleagues suggests that anger and pain have a particularly strong relationship that may have features unique from links between pain and other negative emotions (Berkowitz 1990; 1993; Berkowitz and Thome, 1987; Berkowitz and Harmon-Jones 2004). On a second level, Quartana and Burns (2007) reported that individuals instructed to suppress awareness of thoughts and feelings during interpersonal harassment—which invoked anger and anxiety— reported greater subsequent pain intensity than individuals who suppressed thoughts and feelings during a manipulation that roused primarily anxiety. Finally, results show that trait anger management styles are related to acute and chronic pain intensity most reliably under conditions of anger arousal (Burns et al. 2004, in press). Taken together, findings suggest that, although pain is associated with other negative emotions, there appears to be a particularly strong and perhaps fundamental connection between anger and pain. Although we suggest that ironic processes may represent a cognitive mechanism that links anger suppression to exaggerated pain experience, a potentially more parsimonious model is Muraven and Baumeister’s (2000) limited resources model. This model postulates that a finite pool of resources is available for self-regulation. Thus, attempts to self-regulate at one point may result in fewer resources available to effectively cope and self-regulate during subsequent events. In the case of anger suppression and pain,

J Behav Med (2007) 30:455–469

this model would suggest that efforts to suppress anger may deplete the pool of self-regulatory resources available to assist in tolerating subsequent painful events. If the limited resource model is valid in this instance, then we would expect that efforts to suppress anger will amplify perception of all elements of pain to a similar degree; an outcome due to resource depletion associated with initial efforts to suppress, and not due to ironic increases in the cognitive accessibility of anger-themed content. In the present research, we conducted two studies designed to assess the validity of the propositions derived from our ironic process model of anger suppression and pain. Moreover, the design of these studies allowed us to evaluate differential hypotheses regarding our ironic process model and a limited resources model (Muraven and Baumeister 2000).

Study 1 Study 1 was designed to address the basic hypothesis of our model that suppression of anger in the context of anger arousal would enhance subsequent pain experience to a greater degree than not suppressing in the context of anger arousal. We also examined whether pain intensity was related to perceptions of pain confined primarily to the anger-specific element of pain (per ironic process model), or whether pain intensity was related to anger-specific, sensory and general distress dimensions of pain to a similar degree (per limited resource model). Participants underwent an anger-provoking experimental task with instructions to suppress their emotions (i.e., suppression) or to manage them in any way they chose (i.e., control). A 2-min cold-pressor pain-induction task followed. Participants provided numeric pain ratings at 40s intervals during the cold pressor, and provided ratings for 18 adjectives used to describe sensory (e.g., throbbing), general distress (e.g., cruel) and anger-specific (e.g., irritating) elements of pain (Quartana and Burns 2007). Support for the validity of our ironic process model of anger suppression and pain would be provided if: (a) participants in the suppression condition reported greater pain intensity during the cold pressor than those in the control condition, (b) suppression condition participants reported greater perception of anger-specific elements of pain than control participants, but both conditions reported generally comparable levels of generally distressing and sensory elements, and (c) group differences on pain intensity were accounted for (i.e., mediated) by perceptions of the angerspecific element of pain, but not ratings of the other pain dimensions. However, support for a limited resource model would be provided if, despite participants in the suppression condition reporting greater pain intensity than those in

457

the control condition, both groups reported similar levels of perception of anger-specific, generally distressing and sensory elements of pain.

Methods Participants Fifty-two undergraduates (26 women; 52%) from a Midwestern university participated in exchange for partial course credit. Exclusion criteria included (a) cardiovascular disease, (b) chronic pain conditions (e.g., migraine headaches), (c) Raynaud’s disease, (d) history of bipolar or psychotic disorders, and (e) current alcohol or other substance abuse. The sample reported a mean age of 18.72 years (SD = .99); 70.0% were Caucasian; 22% Asian; 2% Hispanic; 2% African-American; 2% Other; and 2% choose not to indicate an ethnic/racial category. Experimental Tasks Anger-induction Participants gave an impromptu speech during which they argued either for or against a 10% increase in tuition at their undergraduate institution. They were told that the speech would be 5 min in duration and that it would be evaluated for clarity and quality of content by the experimenter. During the speech, the experimenter sat with a clipboard and appeared to be intently listening to and watching the participant. At 1 min, the experimenter interrupted the participant and asked, ‘‘why would you argue that?’’ At 1 min, 30-s, the experimenter stated, ‘‘your speech is making no sense.’’ At 2 min, 30-s, the experimenter stated, ‘‘I still don’t follow what you are saying.’’ At 3 min, the experimenter stated, ‘‘you know what, that’s enough.’’ Pain-induction Pain was induced with a fixed latency (2 min) cold pressor of the non-dominant hand. The cold-pressor apparatus consisted of a standard 48-quart cooler half-filled with water and ice. A metal grate was used to separate the ice from the remainder of the tank to prevent direct ice-to-skin contact. Water temperature was maintained between 1C and 3C. Measures Self-reported Affect and Pain Intensity Participants responded to single items tapping anxious, (i.e., nervous), angry (i.e., mad), and positive affect (i.e.,

123

458

joy) prior to and immediately following the speech. Responses were provided on a 0 (‘‘not at all’’) to 10 (‘‘extremely’’) numeric rating scale. Participants rated cold-pressor pain intensity on a numeric rating scale with anchors of 0 = ‘‘no discomfort’’ and 10 = ‘‘extreme discomfort.’’ Pain intensity ratings were obtained at 40-s intervals during the cold pressor. Pain Adjective Rating Form Participants rated each of a series of adjectives that described the cold-pressor pain on a scale ranging from 0 (‘‘none’’) to 3 (‘‘severe’’). These pain descriptors were taken from the full (Melzack 1975) and short-form (Melzack 1987) versions of the McGill Pain Questionnaire, a wellvalidated and widely used measure to assess the differential qualities of both acute and chronic pain. We assessed ratings on sensory (throbbing, shooting, pricking, stabbing, sharp, cramping, tingling, hot-burning), general distress (punishing-cruel, frightening, tiring-exhausting, vicious, sickening), and anger-specific (irritating, annoying, irksome, infuriating, frustrating) dimensions of pain. Results of a principal components analysis of the descriptors used in this rating form (Quartana and Burns 2007) revealed a 3-component solution that consisted of sensory, general distress and anger-specific subscales of pain elements. Manipulation Check Participants rated the extent to which they tried to suppress their emotions during the speech task on a 0 (‘‘not at all’’) to 10 (‘‘very much’’) scale. Procedures Upon arrival at the laboratory, participants were screened for exclusion criteria and provided written informed consent. Participants then provided affect ratings based on how they felt ‘‘right now.’’ Participants were then given instructions for the speech task. Prior to the speech task, they were provided their respective condition instructions. Participants in the think anything-control condition received the following instructions: While giving your speech, you can expect to experience a number of different emotions. People deal with these emotions in different ways. For the purposes of this experiment, it is very important that you allow yourself to deal with any emotions you experience during the speech as you normally would. That is, feel free to deal with any emotions you experience during the speech in any manner that you would like.

123

J Behav Med (2007) 30:455–469

Participants in the suppression condition received the following instructions: While giving your speech, you can expect to experience a number of different emotions. People deal with these emotions in different ways. For the purposes of this experiment, it is very important that you try to suppress any emotions that you experience during the speech. That is, do not allow yourself to feel any emotions during the speech, no matter what. Participants then performed the speech task under their respective instructions. Immediately following the task, participants again provided affect ratings based on how they felt, ‘‘during the speech task,’’ and subsequently completed the manipulation check questionnaire. Participants were then instructed to immerse their nondominant hand into iced water for 2 min. During hand immersion, participants were asked to indicate their ‘‘discomfort’’ in 40-s intervals on the 0 (‘‘no discomfort’’) to 10 (‘‘extreme discomfort’’) pain intensity scale. Upon removal of the hand, participants completed the pain adjective rating form. Participants were then thoroughly debriefed. Data Reduction and Analytic Strategy Composite variables representing perceptions of angerspecific, general distress, and sensory components of pain were created by taking the mean of responses on relevant items for each subscale (see above). Manipulation check analyses were conducted to assess appropriate changes, and whether participants in the suppression condition attempted to suppress their emotions during the speech task to a greater extent than those in the control condition. Central analyses focused on betweengroups differences in pain indexes. To examine condition effects on pain intensity we conducted Condition · Period mixed design ANOVAs, followed by a between-groups linear contrast. To examine condition effects on ratings of anger-specific, general distress and sensory components of pain, we conducted separate independent samples t-tests. Lastly, to examine whether condition effects on pain intensity ratings were accounted for by perceptions of pain qualities, we examined zero-order correlations among the pain descriptor subscales and pain intensity ratings. Where appropriate preconditions for mediation were met (Baron and Kenny 1986; Shrout and Bolger 2002), we conducted a Condition · Period ANCOVA with pain intensity ratings adjusted for ratings on, for instance, the anger-specific composite. Evidence for mediation was considered present if controlling for pain descriptor subscale scores rendered previously significant Condition effects on pain intensity non-significant.

J Behav Med (2007) 30:455–469

459

Results Manipulation Checks To assess whether the speech task led to expected changes in affect, we conducted a series of Condition (suppression, control) · Period (baseline, speech task) repeated measures ANOVAs. Condition · Period effects were nonsignificant for self-reported anger, anxiety, and positive affect [F’s < 1]. However, the speech task led to significant increases in self-reported anger [F(1, 48) = 27.29, p < .001, g2 = .37; MBaseline = 1.10, SDBaseline = 1.26; MSpeech = 2.40, SDSpeech = 2.10] and anxiety [F(1, 48) = 9.97, p < .005, g2 = .18; MBaseline = 2.80, SDBaseline = 1.87; MSpeech = 3.73, SDSpeech = 2.38], and decreases in positive affect [F(1, 48) = 30.13, p < .001, g2 = .39; MBaseline = 3.47, SDBaseline = 1.95; MSpeech = 2.23, SDSpeech = 2.09]. Although the speech task led to significant increases in anxiety, the effect size for anger increases was twice as large as the effect size for anxiety increases, suggesting that the harassment manipulation aroused anger more strongly than anxiety. To determine whether participants in the suppression condition attempted to suppress emotions during the speech task to a greater extent than those in the control condition, we conducted an independent samples t-test on responses to the manipulation check questionnaire. Results revealed that participants in the suppression condition (M = 5.92, SD = 2.25) reported greater attempts to suppress emotions than those in the control condition (M = 3.26, SD = 2.03; t(48) = 4.28, p < .001, d = 1.24). Condition Effects on Pain Indexes Self-reported Pain Intensity To determine whether pain intensity ratings were strongest among suppression condition participants, responses on the pain discomfort scale were submitted to a 2 Condition (suppression, control) · 4 Period (0-s, 40-s, 80-s, 120-s) mixed design ANOVA. Mean (SD) pain intensity ratings by condition and period are provided in Table 1. This analysis yielded a significant Condition · Period interaction [F(3, 144) = 2.73, p < .05, g2 = .06]. A linear contrast

was significant [F(1, 48) = 5.67, p < .05, g2 = .11], indicating that increases in pain intensity during pain-induction were more rapid and robust among suppression compared to control participants. To illustrate this effect further, we conducted a Condition (suppression, control) betweengroups ANOVA on 40-s and 120 s pain ratings. The Condition effect was non-significant at 40-s [F(1, 48) < 1], but was significant at 120-s [F(1, 48) = 5.27, p < .05, g2 = .10]. In sum, efforts to suppress emotion during an anger-provoking task appeared to enhance pain intensity induced by the subsequent cold pressor. Pain Adjective Rating Form We submitted participants’ ratings of sensory, general distress, and anger-specific component ratings to independent samples t-tests. Table 2 provides Mean (SD) values for sensory, general distress, and anger-specific component ratings by Condition. For sensory and general distress component ratings, significant differences between suppression and control conditions did not emerge [t’s(48) £ 1.31, p ‡ .20, d £ .16]. As predicted, however, a significant group difference emerged for anger-specific component ratings [t(48) = 2.73, p = .009, d = .76] such that participants in the suppression condition provided greater ratings for adjectives tied to the anger-specific element of pain than those in the control condition. Accounting for Condition Differences in Pain Intensity Because the suppression and control conditions differed only on ratings of the anger-specific component of pain, this variable could potentially serve as a mediator of the link between experimental condition and pain intensity produced by the cold pressor, whereas the sensory and general distress ratings could not. However, examination of correlations among the general negative affect and sensory composite ratings and pain intensity ratings at 80- and 120s intervals revealed positive correlations for general negative affect (mean r = .52) as well as sensory (mean r = .28) composite ratings. We next examined whether the proposed mediator (anger-specific ratings) was related significantly to pain ratings. Indeed, pain ratings at 40-s, 80-s, and 120-s were correlated significantly with anger-specific ratings r = .35, .41, and .51 (p’s < .05), respectively.

Table 1 Mean (SD) pain intensity ratings by condition and time—Study 1 40-s

80-s

120-s

Control

4.91 (1.88)

6.38 (2.04)

6.65 (2.35)

Suppression

4.96 (2.37)

6.92 (2.46)

7.69 (2.46)

Note. Participants reported no discomfort at 0-s, and so the mean and SD were each zero

Table 2 Mean (SD) sensory, general distress, and anger-specific composite ratings by condition—Study 1 Sensory

General distress

Anger-specific

Control

1.01 (.47)

.66 (.44)

.78 (.62)

Suppression

1.03 (.66)

.58 (.57)

1.24 (.51)

123

460

Finally, we conducted a 2 Condition (suppression, control) · 4 Period (0-s, 40-s, 80-s, 120-s) mixed design ANCOVA with ratings on the anger-specific composite as the covariate. The previously significant Condition x Period effect on pain ratings was rendered non-significant, F(3, 141) < 1, g2 = .02, such that, after controlling for group differences in anger-specific composite ratings, adjusted pain intensity values indicated that participants in the control condition reported pain levels that were nearly equivalent to those reported by those in the suppression condition. For instance, covariate-adjusted 120-s mean pain intensity ratings were 7.30 and 7.10 for those in the suppression and control conditions, respectively. These results suggest that the greater pain intensity reported by those in the suppression condition was, at least in part, driven by participants’ greater perceptions of angerspecific elements of the cold-pressor pain.

Discussion Results of Study 1 lend additional support to an ironic process model of anger suppression and pain (Quartana and Burns 2007). Participants in the suppression condition reported greater pain intensity during the cold pressor and provided greater ratings of perceptions of anger-specific dimensions of pain than participants in the control condition. Suppression and control participants, however, did not rate the sensory or general distress elements of pain differently. Lastly, effects of suppression on levels of pain intensity during the cold pressor were largely accounted for by the effect of suppression on perceptions of those elements of pain tightly linked to anger, irritation, and annoyance. Taken together, findings suggest that attempts to suppress anger may have ironically rendered anger-related content highly accessible to conscious awareness, thereby contaminating (cf., Cioffi and Holloway 1993) perceptions of the succeeding painful event with angry thoughts and feelings; ultimately, the overall intensity of the pain experience was amplified. A crucial limitation of Study 1 was that we inferred the presence of suppression-induced cognitive accessibility of anger based on self-reported ratings of distinct dimensions of pain. Thus, we did not directly measure whether participants across experimental conditions differed in their levels of cognitive accessibility of anger-related content prior to or during a painful experience.

Study 2 The focus of Study 2 was to more thoroughly examine whether suppression-induced increases in the cogni-

123

J Behav Med (2007) 30:455–469

tive accessibility of anger indeed, in part, link anger suppression and pain intensity. Two modifications of Study 1 were introduced in Study 2. First, we sought to follow Wegner’s (1994) though suppression paradigm more closely than in Study 1, and so examined the effects of the suppression of thoughts and feelings about an angerprovoking event after the event had occurred. Doing so also allowed us to examine whether the effects of anger suppression on pain intensity tend to be confined to attempts to curtail anger during an anger-provoking event (Study 1; Quartana and Burns 2007), or whether such effects may also manifest following efforts to suppress thoughts and feelings after anger has been fully roused. Second, we employed two distinct methods in an attempt to tap cognitive accessibility of anger. First, we used a modified dot-probe paradigm. This kind of task is often used to examine attentional biases toward specific stimuli (e.g., pain-related content among high pain fearful individuals; Keogh et al. 2001). On a modified dot-probe task, highly accessible material should result in a rapid orienting response and sustained attentional allocation to probes that appear in the spatial location formerly occupied by material congruent with highly accessible material, thereby resulting in short probe detection latencies under such conditions. For instance, patients with high levels of pain anxiety have shown strong attentional biases toward pain-related stimuli (e.g., throbbing) compared to neutral stimuli (e.g., Keogh et al. 2001), and those with anxiety disorders tend to show heightened attentional bias to anxiety-relevant stimuli (e.g., MacLeod et al. 1986; Mogg et al. 1993). In terms of the ironic effects of anger suppression, evidence for suppression-induced increases in cognitive accessibility of anger content would be revealed by greater attentional bias towards anger-related stimuli relative to neutral stimuli among participants in an anger suppression versus a control condition. The modified dot-probe task used in Study 2 included trials with words semantically related to anger or anxiety, which were paired with category-neutral (i.e., household items) words. Word pairs were presented for 500 ms, a period sufficiently long to allow initial orienting and attention engagement processes (LeBerge 1995; Lees et al. 2005; MacLeod et al. 1986; Mogg et al. 1993). If suppression of anger renders anger more accessible than not suppressing anger, then attentional biases toward angerrelated stimuli should be strongest within the suppression condition. Keeping in mind that our anger-induction methods also appear to elicit increases in anxiety (see Study 1), we focused on determining whether participants attempting to suppress exhibited greater attentional bias toward anger-related words relative to anxiety-related words than those in a control condition. As a second, more direct method of assessing between-groups differences in anger accessibility, participants provided ratings of anger,

J Behav Med (2007) 30:455–469

anxiety, and positive affect following each epoch of the laboratory protocol. This procedure allowed us to examine whether self-reported levels of anger in particular were elevated among those in the suppression condition compared to participants in the control condition during suppression efforts as well as during pain-induction. In Study 2, participants underwent an anger-provoking speech task identical to that used in Study 1, and then completed the modified dot-probe task under suppression or ‘‘think anything’’ (i.e., control) conditions. All participants then underwent a cold pressor of the non-dominant hand. If our ironic process model of anger suppression and pain is valid, then participants in the suppression condition should reveal ‘‘hyperaccessibility’’ (cf., Wegner and Erber 1992) of anger evidenced via greater attentional biases to anger-related relative to anxiety-related stimuli on the modified dot-probe task, greater self-reports of anger during pain, greater pain intensity, and stronger perception of anger-specific elements than those in the control condition. Further, we expected that differences between suppression and control conditions on pain intensity would be mediated by relative attentional bias towards anger-related stimuli (i.e., versus anxiety-related stimuli), self-reported increases in anger during pain, and perception of anger-specific elements of pain. If a limited-resource model (Muraven and Baumester 2000) is valid, then participants in the suppression condition would reveal greater overall performance deficits (i.e., longer probe detection latencies across all trials) on the modified dot-probe task, similar levels of anger and anxiety during pain-induction as well as greater perception of all qualities of pain assessed (i.e., sensory, general distress, and anger) compared to those in the control condition. Moreover, we would anticipate that any effect of suppression on pain intensity would be mediated by increased negative affect in general during the cold pressor, or by overall performance deficits on the modified dot-probe task by participants in the suppression condition.

461

Experimental Tasks Anger- and pain-induction methods used in Study 2 are identical to those of Study 1 (see above). Modified Dot-probe This task consisted of 12 anger-neutral, 12 somatic-neutral, and 12 social evaluative-neutral word pairs. To test the hypotheses of the current study, mean reaction times for the somatic-neutral and social evaluative-neutral trials were combined to reflect a single ‘‘anxiety-neutral’’ composite. The anger- and anxiety-relevant stimulus words were chosen from a number of prior research studies (e.g., Clore et al. 1987; Faunce et al. 2004). All neutral stimuli were category-neutral and were related to household items (e.g., Mogg et al. 1993), and were matched in terms of word length and frequency of usage in the English language. An additional 10 neutral–neutral word pairs were created for the purposes of 10 practice trials (see below). None of these neutral–neutral word pairs were used during the experimental trials. There were 36 experimental trials, and so each word pair appeared only once. Trials for the modified dot-probe task proceeded in the following manner: 500 ms presentation of fixation point (+++); 500 ms presentation of a word pair; and presentation of probe (*) in the upper or lower spatial location. Participants were instructed to respond as quickly and accurately as possible by pressing a key on a keypad indicating whether the probe appeared in the ‘‘upper’’ or ‘‘lower’’ spatial location. The inter-trial interval was 1,000 ms across all trials. The position of the word pairs and dot-probe were counterbalanced so that they appeared in upper or lower spatial locations with equal frequency. The orders of the presentation of word pairs were completely randomized between participants. All words were presented in capital letters and in white and bold font against a black background. Measures

Methods

The measures employed in Study 2 are identical to those of Study 1.

Participants Procedure Ninety-six undergraduates (58 women; 60.4%) from a Midwestern university participated in exchange for partial course credit. Inclusion and exclusion criteria are identical to those of Study 1. The sample reported a mean age of 19.1 years (SD = 1.07); 58.3% were Caucasian, 28.1% Asian, 7.3% Hispanic, 3.1% African-American, 1% Other, and 2.1% choose not to indicate an ethnic/racial category.

Procedures for Study 2 are identical to those of Study 1 with the following exceptions. First, prior to the speech task, participants were given instructions for the modified dot-probe task. They then performed 10 practice trials so as to acclimate them to the task. This preparation allowed for a more rapid transition from the anger-provoking speech task to the experimental trials of the dot-probe task.

123

462

Second, participants were randomized to suppress or think anything-control conditions after the speech task was terminated. Specifically, prior to the experimental trials on the modified dot-probe task, participants received their respective think anything-control or suppression condition instructions. Participants in the control condition received the following instructions: I would now like you to do the computer task again. The task is exactly the same as before. Feel free to think about anything you would like during the computer task. You can think about the speech you just gave, or how it made you feel, but you do not have to. Remember to try your hardest on this task, and to think about anything you would like. Participants in the suppression condition received the following instructions: I would now like you to do the computer task again. The task is exactly the same as before. However, try your hardest not to think about the speech you just gave, or how it made you feel, during the computer task. Remember to try your hardest on this task, and not to think about anything to do with the speech you just gave, or how it made you feel. Participants then performed experimental trials on the modified dot-probe task under their respective instructions. Lastly, participants rated their affect (i.e., mad, joyful, nervous) immediately following the anger-provoking speech task, and following the modified dot-probe (i.e., ‘‘how you felt during the computer task’’) and cold-pressor tasks (i.e., ‘‘how you felt while your hand was in the iced water’’). Data Reduction and Analytic Strategy As in Study 1, subscales were created for ratings of items corresponding to anger-specific, general distress and sensory elements of pain. To reduce non-normality of the reaction time data distributions, dot-probe trials were discarded that had response errors or had reaction times less than 200 ms or greater than 1,000 ms. To further reduce non-normality of the reaction time distribution, raw values were converted to z-scores and values greater than 3.29 or less than –3.29 were replaced by the next highest true value in the distribution (Tabachnik and Fidell 1996). To examine the magnitude and direction of attention bias to anger- and anxiety-related word stimuli, we computed bias scores based on the formula provided by MacLeod and Mathews (1988) for use in dot-probe paradigms:

123

J Behav Med (2007) 30:455–469

Bias Score ¼1=2 ½ðProbe Upper=Emotion Word Lower  Probe Upper=Emotion Word UpperÞ þ ðProbe Lower=Emotion Word Upper  Probe Lower=Emotion Word LowerÞ Thus, positive values reflected an attentional bias toward the emotion-relevant word, and negative values reflected an attentional bias away from the emotion-relevant word. Mean attentional bias scores were calculated for the two groups (suppression and control) for anger-neutral and anxiety-neutral trials. Manipulation check analyses were conducted to assess appropriate changes in affect, and whether participants in the suppression condition attempted to suppress their thoughts and feelings of the speech during the modified dot-probe to a greater extent than those in the control condition. Central analyses focused on between-groups differences in pain indexes. To examine condition effects on pain intensity, we conducted Condition · Period mixed design ANOVAs, followed by a between-groups linear contrast. To examine condition effects on ratings of angerspecific, general distress and sensory components of pain, we conducted separate independent samples t-tests. The next group of analyses conducted focused on suppression versus control condition differences in anger accessibility across the modified dot-probe as well as selfreported affect scales. We first examined between-groups differences in attentional biases towards anger- versus anxiety-related word stimuli using a 2 Condition (suppress, control) · Bias Score (anger-related bias, anxiety-related bias) mixed model ANOVA with Condition as the between-groups factor and Bias Score as the repeated measure. Regarding self-reported affect, we first computed simple change scores for angry, anxious, and positive affect composites for baseline-to-dot-probe and baselineto-cold-pressor. We then conducted independent 2 Condition (suppress, control) ANOVAs on these change scores. Lastly, to examine whether condition effects on pain intensity ratings were accounted for by (a) perceptions of the anger-specific dimension of pain and/or (b) selfreported increases in anger, we examined zero-order correlations among these variables and self-reported pain intensity ratings. Where appropriate preconditions for mediation were met (Shrout and Bolger 2002), we conducted a Condition · Period ANCOVA with pain intensity ratings adjusted for ratings on, for instance, the angerspecific composite. Evidence for mediation was considered present if controlling for anger-specific composite ratings and/or self-reported increases in anger rendered previously significant Condition effects on self-reported pain intensity non-significant.

J Behav Med (2007) 30:455–469

463

suppress thoughts and feelings related to an angerprovoking event appeared to enhance the intensity of pain induced by the cold pressor.

Results Manipulation Checks A series of 2 Condition (suppression, control) · 2 Period (baseline, speech task) mixed design ANOVAs were conducted for the affect ratings. Condition · Period interactions were non-significant for all items (F’s < 2.0). Significant Period effects emerged for anger [F(1, 95) = 12.02, p < .001, g2 = .19; MBaseline = 1.26, SDBaseline = 1.70; MSpeech = 2.09, SDSpeech = 2.09], anxiety [F(1, 95) = 11.53, p < .001, g2 = .11; MBaseline = 2.67, SDBaseline = 1.87; MSpeech = 3.30, SDSpeech = 2.14], and positive affect [F(1, 95) = 50.58, p < .001, g2 = .35; MBaseline = 3.21, SDBaseline = 2.22; MSpeech = 1.95, SDSpeech = 1.80]. As in Study 1, the effect size of the speech task on self-reported anger was notably stronger than for self-reported anxiety, and so appeared to affect changes in anger most robustly. Participants in the suppression condition (M = 5.45, SD = 2.29) reported significantly greater attempts to suppress speech-related thoughts and feelings during the computer task than participants in the control condition [M = 3.77, SD = 2.42; t(94) = 4.46, p < .001, d = .70].

Pain Adjective Rating Form

Condition Effects on Pain Indexes

Five participants’ data were lost due to greater than half of trials having extreme reaction times (i.e., <200 ms or >1,000 ms) and/or response errors. Hence, reaction time data were available for 91 participants. Among these participants, response errors occurred on less than 5% of trials. The overall mean reaction time was 459.11 ms (SD = 93.52). To examine relative differences in accessibility between conditions of anger-related versus anxiety-related stimuli as planned comparisons (Rosenthal and Rosnow 1985), we conducted analyses on single indexes of anger and anxiety attentional bias. As outlined above, anger bias index and anxiety bias index scores were computed using the formula provided by MacLeod and Mathews (1988). These indexes were submitted to a 2 Condition (suppression, control) · 2 Bias (anxiety bias index, anger bias index) mixed design ANOVA. Mean (SD) bias index scores for anger- and anxiety-related stimuli are presented by condition in Table 5.

Self-reported Pain Intensity Effects of condition on pain intensity were examined with a 2 Condition (suppression, control) · 4 Period (0-s, 40-s, 80-s, 120-s) mixed design ANOVA. Mean (SD) pain intensity values by condition and period appear in Table 3. A significant Condition · Period effect was found [F(3, 282) = 3.83, p < .05, g2 = .04]. A between-groups linear contrast was significant [F(1, 94) = 6.49, p < .025, g2 = .07], suggesting that increases in pain intensity during pain-induction were significantly greater among suppression than control condition participants. To illustrate this effect, we conducted a 2 Condition (suppression, control) between-groups ANOVA on 40-s and 120 s pain ratings. The Condition effect was non-significant at 40-s [F(1, 95) = 1.15, p > .25, g2 = .01], but was significant at 120-s [F(1, 95) = 7.99, p < .01, g2 = .08]. In sum, efforts to

Sensory, general distress and anger-specific composite ratings were subjected to independent samples t-tests. Mean (SD) sensory, general distress and anger-specific component ratings by condition are presented in Table 4. Significant condition effects did not emerge for sensory and general distress component ratings [t’s(94) £ 1.25, p’s ‡ .22, d’s £ .26]. Participants in the suppression condition provided greater anger-specific component ratings than those in the control condition, but this effect failed to reach conventional levels of statistical significance [t(94) = 1.74, p = .08, d = .36]. Results suggest that suppression condition participants rated the pain more strongly in terms of anger than participants in the control condition, whereas participants in both conditions perceived and described the pain in terms of sensory and generally distressing qualities to a similar degree. Condition Effects on Attentional Biases

Table 3 Mean (SD) pain intensity ratings by condition and time—Study 2 40-s

80-s

120-s

Control

5.27 (1.65)

6.67 (1.97)

6.40 (2.35)

Suppression

5.69 (2.13)

7.25 (2.10)

7.67 (2.05)

Note. Participants reported no discomfort at 0-s, and so the mean and SD were each zero

Table 4 Mean (SD) sensory, general distress, and anger-specific composite ratings by condition—Study 2 Sensory

General distress

Anger-specific

Control

1.00 (.55)

.49 (.61)

1.23 (.67)

Suppression

1.15 (.59)

.58 (.60)

1.48 (.74)

123

464

J Behav Med (2007) 30:455–469

Table 5 Mean (SD) attentional bias index scores for anger-related and anxiety-related stimuli by condition—Study 2 Anger bias index

Anxiety bias index

Control

1.42 (15.05)

12.56 (10.63)

Suppression

5.94 (14.04)

10.21 (9.32)

Note. Larger bias index values represent greater bias towards affectladen relative to neutral stimuli

A significant Bias effect emerged [F(1, 89) = 16.19, p < .001, g2 = .16] such that, regardless of condition, participants showed a stronger bias towards anxiety than anger stimuli. A trend emerged for the Condition · Bias interaction effect [F(1, 89) = 3.12, p = .07, g2 = .04]. Provided the a priori nature of our hypothesis, we proceeded to conduct hypothesis-driven within-condition simple effect tests of Bias (i.e., anxiety bias index versus anger bias index) at each level of Condition. These tests revealed a significant difference between attentional bias towards anxiety- versus anger-related stimuli within the control condition [t(44), 3.95, p < .001, d = .86], but not within the suppression condition [t(45) = 1.66, p = .10, d = .37]. Two additional comparisons were then conducted to elucidate the source of these between-group differences in relative attentional bias. Specifically, we inspected group differences in (a) anger-related stimuli (irrespective of attentional bias towards anxiety) and (b) anxiety-related stimuli. As can be seen in Table 5, group differences in attentional bias toward anger were evident, although this difference failed to reach statistical significance [t(89) = 1.49, p = .14, d = 32]. Moreover, there was virtually no evidence for group differences in attentional bias towards anxiety [t(89) < 1]. Hence, the group differences that were identified regarding relative attentional bias toward anger- versus anxiety-related stimuli appeared to be accounted for by the greater attentional bias toward anger in the suppression versus the control condition, thus creating a smaller relative attentional bias to anger- versus anxiety-related stimuli for participants in the suppression (Manxiety–anger = 4.27) versus those in the control (Manxiety–anger = 11.14) condition. In sum, these data hint at an overall greater attentional bias to anger-related stimuli for participants in the suppression versus control condition.

To examine whether resources were simply depleted as a result of suppression (i.e., a limited-resource account), we conducted a planned single-df contrast on mean reaction times collapsed across all trials by Condition. This analyses yielded a nonsignificant Condition effect [t(89) < 1, d = .14; suppression M = 460.58, SD = 97.23; control M = 447.63, SD = 89.80], suggesting that suppression did not merely exhaust participant’s reserve capacity for selfregulation more than the control condition. Condition Effects on Self-Reported Affect Simple change scores were computed for baseline-to-dotprobe task and baseline to cold-pressor changes by subtracting baseline from respective task values for each self-reported affect item (i.e., mad, nervous, and joyful). These change scores were submitted to a series of ANOVAs. Mean (SD) values for these change scores are provided in Table 6. It is important to note that the results reported were not significantly different than those achieved with ANCOVAs with baseline values entered as a covariate, and so only results of ANOVAs with simple change scores as DVs are presented. Condition effects were non-significant for baseline-tocomputer task change in anger and anxiety [F’s(1, 94) < 1]. Baseline-to-computer task changes in positive affect significantly differed between conditions [F(1, 94) = 4.86, p = .03, g2 = .05], such that greater decreases in positive affect were reported among those in the suppression condition compared to the control condition. The condition effect was non-significant for baseline-to-coldpressor anxiety change [F(1, 94) < 1] as well as for baseline-to-cold-pressor positive affect change [F(1, 94) = 1.29, p = .30, g2 = .01]. Finally, the condition effect for baseline-to-cold-pressor changes in anger was significant [F(1, 94) = 6.92, p = .01, g2 = .07], such that participants in the suppression condition reported greater increases in anger during the cold pressor than participants in the control condition. To sum, results indicate that participants in the suppression condition reported significantly greater increases in anger during pain-induction than those in the control condition. No between-groups differences were observed, however, across changes in positive affect and anxiety

Table 6 Mean (SD) self-reported anger, anxiety, and positive affect change scores by Condition and Experimental Task—Study 2 Baseline-to-computer task changes Anger

Anxiety

Baseline-to-cold-pressor changes Positive affect

Anger

Anxiety

Positive affect

Control

.21 (1.17)

–.61 (1.82)

–.65 (1.86)

.17 (2.10)

–.67 (2.22)

–1.37 (2.05)

Suppression

.34 (1.66)

–.70 (1.20)

–1.43 (1.57)

1.29 (2.09)

–.29 (2.03)

–1.77 (1.64)

Note. Positive values indicate baseline-to-task increases; negative values indicate baseline-to-task decreases

123

J Behav Med (2007) 30:455–469

during pain-induction. Further, during efforts to suppress thoughts and feelings regarding the speech task, those in the suppression condition reported greater decreases in positive affect than those in the control condition. Accounting for Condition Effects on Pain Intensity We tested mediation effects that were predicted by our ironic process model. First, we examined the role of the anger-specific element of pain. Correlations between anger-specific composite ratings and pain intensity were significant at the 80- and 120-s ratings (r’s = .39 and .35, respectively, p’s < .001). Hence, we conducted a 2 Condition (suppression, control) · 4 Period (0-s, 40-s, 80-s, 120-s) mixed model ANCOVA on pain intensity ratings with anger-specific component ratings entered as a covariate. This analysis rendered the formerly significant Condition · Period interaction nonsignificant [F(3, 279) = 2.94, p = .05, g2 = .03], suggesting that the greater reports of pain intensity of suppression condition participants compared to control participants were partly accounted by between-groups differences in perception of anger-specific qualities of the pain. Although significant group differences did not emerge across general negative affective and sensory composite ratings, we examined the patterns of correlations among these measures and pain intensity ratings at 80- and 120-s intervals. Significant correlations were observed for both general negative affect (mean r = .46) and sensory (mean r = .46) composite ratings. Next, we examined the potential mediating role of relative attentional bias towards anger-related stimuli. To do so, we first computed a ‘‘relative anger bias index’’ by subtracting the anger bias index from the anxiety bias index, wherein smaller relative anger bias index values represent greater relative bias towards anger. Zero-order correlations were then generated among relative anger bias index scores and pain intensity reports. These relationships were not significant (r’s £ .10). Hence, relative attention bias towards anger-related stimuli could not be considered a potential mediator of the condition–pain intensity relationship. We subsequently examined whether baseline-to-coldpressor changes in anger report mediated condition effects on pain intensity. Significant correlations emerged among these factors (r’s ‡ .21, p’s < .05), suggesting that effects of suppression on pain intensity may have accounted for effects of condition observed on pain intensity. Hence, we conducted a 2 Condition (suppression, control) · 4 Period (0-s, 40-s, 80-s, 120-s) ANCOVA on pain intensity ratings with baseline-to-cold-pressor anger change scores entered as a covariate. This analysis rendered the formerly significant Condition · Period interaction non-significant [F(3, 279) = 3.03, p = .07, g2 = .03], suggesting that greater

465

reports of pain intensity of suppression condition participants compared to control participants were partly accounted by the former group’s greater feelings of anger during pain. Recall that those in the suppression condition reported greater decreases in positive affect during the computer task. However, zero-order correlations were non-significant between positive affect baseline-to-computer task change scores and pain intensity ratings (r’s £ .13). Hence, the pattern of correlations among positive affect changes during efforts to suppress and self-reported pain intensity suggest that decreases in positive affect did not appreciably overlap with the effect of suppression on pain intensity. Exploratory Analyses Participants in the suppression condition revealed a marginally more robust attentional bias towards anger-related stimuli relative to anxiety-related stimuli and reported greater increases in self-reported anger in response to coldpressor pain. Not inconsistent with our ironic process model, the case could be made that enhanced attentional biases toward anger-relevant stimuli among suppression participants would predict greater increases in anger inspired by the cold pressor. Indeed, within the suppression condition, we found that ‘‘relative anger bias index’’ values (i.e., anger bias index–anxiety bias index scores) were inversely related to baseline-to-cold-pressor anger change values (r = –.33, p = .02), indicating that larger attentional biases towards anger-relative to anxiety-related stimuli (i.e., smaller relative anger bias index values) predicted increases in anger during subsequent pain. Moreover, correlations between ‘‘relative anger bias index’’ values were unrelated to changes in anxious and positive affect changes inspired by the cold pressor (r’s £ .10). Taken together, these data appear to suggest that attempts to suppress anger at one point may significantly influence the magnitude of angry affect elicited by a later noxious event, perhaps through exaggerated attentional bias to suppression target-relevant stimuli (i.e., anger).

Discussion Results of Study 2 largely replicated those of Study 1. Specifically, participants in the suppression condition reported greater increases in pain intensity during paininduction than those in the control condition. Moreover, those in the suppression condition tended to endorse stronger perceptions of the anger-specific element of pain than those in the control condition. Perceptual differences were much less evident between suppression and control conditions for the sensory and general distress components

123

466

of pain. As in Study 1, the effects of suppression on pain were in part accounted for by perception of the angerspecific aspect of pain. Thus, it appears that efforts to suppress anger may enhance accessibility of anger-related thoughts and feelings, thereby contaminating perceptions of subsequent pain in an anger-congruent fashion. In order to further evaluate our ironic process model, we also expanded our assessment of anger accessibility. First, we examined the effects of suppression on attentional biases towards anger- versus anxiety-related stimuli using a modified dot-probe paradigm. Results suggested somewhat stronger attentional biases towards anger-related stimuli relative to anxiety-related stimuli among suppression compared to control participants. However, these relative attentional biases were not related significantly to reports of subsequent pain intensity, nor were they related significantly to perception of any distinct dimension of pain. Interestingly, the greater relative attentional bias towards anger-related stimuli observed among suppression participants did predict changes in anger (but not anxiety or positive affect) during cold-pressor pain. Thus, although not related significantly to pain indexes, this relative bias index was related to emotional responses to pain, and anger in particular. Lastly, there was little evidence that a limited-resource model better accounts for the data than an ironic process model. We also assessed fluctuations in self-reported anxious, angry and positive affect across each experimental epoch. Findings here revealed that those in the suppression condition reported significantly greater increases in anger during pain-induction. Moreover, changes in anger during pain largely accounted for the effect of suppression on reports of pain intensity. Of note, those in the suppression condition also reported greater decreases in positive affect during the computer task than those in the control condition. However, fluctuations in positive affect were largely unrelated to reports of pain intensity as well as pain descriptors. It seems that efforts to suppress anger following anger-provocation may enhance accessibility of anger, which apparently can affect responses to a subsequent painful encounter. Indeed, the experience of pain following efforts to suppress anger appears to be colored heavily by angry affect.

General Discussion Although clinicians and researchers advocate a causal connection between suppression of strong negative emotions and heightened pain experience, the study of these links has largely relied on cross-sectional, correlational methodologies. To examine whether the case could be made that anger suppression actually influences subsequent

123

J Behav Med (2007) 30:455–469

levels of pain intensity, as well as to explore a potential mechanism that may underlie this relationship, we examined hypotheses derived from an ironic process model of anger suppression and pain (Quartana and Burns 2007). We argue that efforts to suppress anger paradoxically increase the cognitive accessibility of anger-related thoughts and feelings, which in turn contaminate (cf., Cioffi and Holloway 1993) perceptions of later pain. Results of the studies presented herein lend further support for this model, and suggest that there might indeed exist a causal connection between anger suppression at one point in time and magnification of pain experience at a subsequent point in time, particularly with respect to those aspects of pain tied tightly to anger and irriation. Moreover, findings implicate the paradoxical increase in anger accessibility following suppression as a potential cognitive–affective mechanism of this link. It is important to note and address some key limitations and shortcomings of these studies, as well as to suggest important future research directions. Foremost among them, we did not find unequivocal evidence for suppression-induced increases in the cognitive accessibility of anger. We did find evidence that suppression resulted in increased anger during pain as well as greater perceptions of pain that appeared to be confined to pain’s inherent anger-specific quality (Berkowitz 1990, 1993; Berkowitz and Thome 1987; Berkowitz and Harmon-Jones 2004; Quartana and Burns 2007). Although data stemming from the dot-probe task were in the direction predicted by our model, findings were only marginally significant for increased accessibility of anger tied to suppression through this pathway. We can provide at least three reasons why the effect of suppression on the cognitive accessibility of anger revealed through the dot-probe task may have gone undetected. On one level, meta-analyses suggest that the ironic effects of suppression are most robust in the moments after suppression efforts have been suspended (Abramowitz et al. 2001). Recall that we employed the dot-probe task during the suppression manipulation. This procedure may have severely hampered our ability to detect heightened levels of anger accessibility among suppression condition participants. Moreover, we were able to detect effects of suppression on angry affect during the cold pressor; that is, following efforts to suppress. It will be important for future studies to examine levels of the cognitive accessibility of anger between suppression and control condition both during and following suppression. On a second level, the dot-probe task in the present investigation involved 500 ms stimulus duration times. Thus, stimuli were open to conscious perception and recognition (LeBerge 1995; MacLeod et al. 1986; Mogg et al. 1993). As a result, it is quite possible that the stimuli underwent some level of conscious, effortful cognitive

J Behav Med (2007) 30:455–469

manipulation. In terms of an ironic process model, a 500 ms stimulus duration may have left a sufficient amount of resources available for the operator to scan for distractors (i.e., thoughts and feelings unrelated to the to-besuppressed stimulus). If so, attentional resources dedicated to the anger-related stimuli would be decreased, and would likely diminish between-groups differences in anger accessibility. Indeed, Page et al. (2005) found evidence for the ironic effects of suppression under conditions in which cognitive resources available for use by the operator were few. We are presently pursuing this possibility through systematic examination of suppression-induced anger accessibility using emotional information processing paradigms in which we vary the stimulus presentation time between short (i.e., 10 ms) and long (i.e., 500 ms) durations. On a final level, it is important to keep in mind that our modified dot-probe paradigm presented words related to anger and/or anxiety paired simultaneously with an affectively neutral counterpart within each trial. As a result, we may have provided not only the resources (i.e., time) for the operator to search for and find distractors from the tobe-suppressed speech-related thoughts and feelings, but also a viable distractor—the neutral stimulus itself. However, we did not find any evidence to suggest that attentional bias was shifted away from affect-laden stimuli in the direction of the neutral stimulus pairs. Still, it may be more appropriate and methodologically robust to examine suppression-induced changes in cognitive accessibility of to-be-suppressed content (e.g., anger) via alternative paradigms such as an emotional Stroop (e.g., Mogg et al. 1993), which presents only one stimulus at a time. Indeed, studies have shown ironic, suppression-induced increases in cognitive accessibility of to-be-suppressed content using a modified Stroop paradigm (Wegner and Erber 1992; Wegner et al. 1993). Another notable shortcoming may reside in our angerinduction methods. On the one hand, we provoked anger in an interpersonal context, which likely represents a relatively high degree of ecological validity. On the other hand, our method was comprised of a heavy social evaluation component. Indeed, our anger-induction task led to increases in anxiety as well as anger. However, the magnitude of the effect of the manipulation on anger exceeded the effect of the manipulation on anxiety, and the effects of suppression on pain and self-reported affect during pain were confined to anger-specific descriptors of pain. Thus, it may be the case that the social-evaluative nature of the task limited our ability to detect evidence of heightened anger accessibility between suppression and control conditions, particularly on an indirect measure of accessibility such as a modified dot-probe task. Inspection of anger versus anxiety bias index means revealed that the attentional

467

bias toward anxiety-related stimuli was stronger, irrespective of condition, than the attentional bias toward angerrelated stimuli. However, suppression appeared to increase attentional bias toward anger-related stimuli, but had a null effect on attentional bias toward anxiety-related stimuli. To bolster the claim that the effects of the suppression of anger in particular on pain are more detrimental than, say, the suppression of sadness, will require future studies to cross suppression manipulations with conditions in which distinct emotions are provoked to the greatest extent possible (e.g., anger versus sadness recall interviews; Burns 2006a). As noted at the outset, there has been much speculation regarding the distinct relationship between the inhibition of anger and clinical pain. Indeed, some have suggested that the tendency to inhibit or suppress anger may be part and parcel of a pain-prone personality (Engel 1959). We have argued, however, that only well-founded theoretical conceptualizations will help demystify how exactly anger suppression affects increases in acute and, by extension, chronic pain. To illustrate, we have shown that chronic low back pain patients who scored high on both anger-in and hostility (Burns 1997; Burns et al. 2006) showed the greatest increases in lower paraspinal muscle tension—that is, muscles of the lower back—during anger than other groups. Such anger-induced lower paraspinal muscle tension has also been found to correlate with everyday chronic pain levels (Burns et al. 1997; Burns 2006a). Such results guided by well-defined theory will shed light on specific causal pathways that underlie effects of suppressed anger on pain—such as intrusive anger-related thoughts that remain unexpressed and excessive low back muscle tension—that may then constitute empirically supported targets for accurate therapeutic interventions. Findings based on an ironic process model may ultimately provide a new impetus to pursue therapeutic techniques to decrease anger suppression not only among chronic low back pain patients but among patients suffering from a variety of painful medical conditions. In sum, many studies suggest that the regulation of negative emotions (e.g., anger) affects physical health, but study designs have limited our ability to draw strong causal inferences. Regarding anger suppression and pain, our ironic process model of anger suppression and pain is offered to provide a theoretical and methodological framework from which to move toward statements about causal effects (Quartana and Burns 2007). Here, our findings point to a potentially unique connection between initial attempts to suppress anger during or following provocation and subsequent intensity of acute pain, and that this link may be driven, at least in part, by contamination effects brought about by highly accessible anger content. Much research on anger suppression variables, such as trait anger inhibition, and pain is focused on

123

468

patients suffering from chronically painful conditions. It will be crucial to extend this model to those with chronic pain to examine whether anger suppression causally influences clinically relevant variables. Finally, research indicates that deliberate attempts to disinhibit negative emotions tied to troubling events or chronic stressors appear to confer detectable benefits on patients with chronic pain conditions (e.g., Broderick et al. 2004; Greenberg and Stone 1992; Kelley et al. 1997; Smyth et al. 1999). Thus, results of studies that pursue and expand an ironic process model of emotion suppression on physical pain, and which uncover active mechanisms, should provide an empirical boost to the development of therapeutic techniques to decrease patients’ habitual suppression of negative emotion, perhaps—especially—anger. Acknowledgements This research was partly supported by a Ruth L. Kirschstein National Research Service Award (1 F31 NS05120001A1) from the National Neurological Disorders and Stroke awarded to Phillip J. Quartana, and Grant NS37164 from the National Institutes of Neurological Disorders and Stroke awarded to John W. Burns.

References Abramowitz, J. S., Tolin, D. F., & Street, G. P. (2001). Paradoxical effects of thought suppression: A meta-analysis of controlled studies. Clinical Psychological Review, 21, 683–703. Bargh, J. A., & Pietromonaco, P. (1982). Automatic information processing and social perception: The influence of trait information presented outside of conscious awareness on impression formation. Journal of Personality and Social Psychology, 43, 437–449. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Berkowitz, L. (1990). On the formation and regulation of anger and aggression: A cognitive-neoassociationistic analysis. American Psychologist, 45, 494–503. Berkowitz, L. (1993). Pain and aggression: Some findings and implications. Motivation and Emotion, 17, 277–293. Berkowitz, L., & Thome, P. R. (1987). Pain expectation, negative affect, and angry aggression. Motivation and Emotion, 11, 183–193. Berkowitz, L., & Harmon-Jones, E. (2004). Toward and understanding of the determinants of anger. Emotion, 4, 107–130. Bonica, J. J. (1990). Definitions and taxonomy of pain. In J. J. Bonica (Ed.), The management of pain. Philadelphia, PA: Lea & Febiger. Broderick, J. E., Stone, A. A., Smyth, J. M., & Kaell, A. T. (2004). The feasibility and effectiveness of an expressive writing intervention for rheumatoid arthritis via home-based videotaped instructions. Annals of Behavioral Medicine, 27, 50–59. Bruehl, S., Burns, J. W., Chung, O. Y., Ward, P., & Johnson, B. (2002). Anger and pain sensitivity in chronic low back pain patients and pain-free controls: The role of endogenous opioid blockade. Pain, 99, 923–933. Bruehl, S., Chung, O. Y., & Burns, J. W. (2003a). Differential effects of expressive anger regulation on chronic pain intensity in CRPS and non-CRPS limb pain patients. Pain, 104, 647–654.

123

J Behav Med (2007) 30:455–469 Bruehl, S., Chung, O. Y., Burns, J. W., & Biridepalli, S. (2003b). The association between anger expression and chronic pain intensity: Evidence for partial mediation by endogenous opioid dysfunction. Pain, 106, 317–324. Burns, J. W. (1997). Anger management style and hostility: Predicting symptom-specific physiological reactivity among chronic low back pain patients. Journal of Behavioral Medicine, 20, 505–522. Burns, J. W. (2006a). Arousal of negative emotions and symptomspecific reactivity in chronic low back pain patients. Emotion, 6, 309–319. Burns, J. W. (2006b). The role of attentional strategies in moderating links between acute pain-induction and subsequent emotional stress: Evidence for symptom-specific reactivity among chronic pain patients versus healthy nonpatients. Emotion, 6, 180–192. Burns, J. W., Bruehl, S., & Caceres, C. (2004). Anger management style, blood pressure reactivity and acute pain sensitivity: Evidence for a trait · situation model. Annals of Behavioral Medicine, 27, 195–204. Burns, J. W., Bruehl, S., & Quartana, P. J. (2006). Anger management style and hostility among chronic pain patients: Effects on symptom-specific physiological reactivityduring anger- and sadness-inductions. Psychosomatic Medicine, 68, 786–793. Burns, J. W., Johnson, B. J., Devine, J., Mahoney, N., & Pawl, R. (1998). Anger management style and the prediction of treatment outcome among male and female chronic pain patients. Behaviour Research and Therapy, 36, 1051–1062. Burns, J. W., Quartana, P. J., & Bruehl, S. (in press). Anger management style moderates effects of emotion suppression during stress on pain and cardiovascular responses during paininduction. Annals of Behavioral Medicine. Burns, J. W., Wiegner, S., Derleth, M., Kiselica, K., & Pawl, R. (1997). Linking symptom-specific physiological reactivity to pain severity in chronic low back pain patients: A test of mediation and moderation models. Health Psychology, 16, 319–326. Cioffi, D., & Holloway, J. (1993). Delayed costs of suppressed pain. Journal of Personality and Social Psychology, 64, 274–282. Clore, G. L., Ortony, A., & Foss, M. A. (1987). The psychological foundations of the affective lexicon. Journal of Personality and Social Psychology, 53, 751–766. DeSteno, D., Petty, R. E., Rucker, D. D., Wegnener, D. T., & Braverman, J. (2004). Discrete emotions and persuasion: The role of emotion-induced expectancies. Journal of Personality and Social Psychology, 86, 43–56. Engel, G. (1959). ‘‘Psychogenic’’ pain and the pain prone patient. American Journal of Medicine, 26, 899–918. Faunce, G. J., Mapledoram, P. K., & Job, R. F. S. (2004). Type A behaviour pattern and attentional bias in relation to anger/ hostility, achievement, and failure. Personality and Individual Differences, 36, 1975–1988. Fernandez, E., & Turk, D. C. (1995). The scope and significance of anger in the experience of chronic pain. Pain, 61, 165–175. Forgas, J. P. (1995). Mood and judgment: The affect infusion model (AIM). Psychological Bulletin, 117, 39–66. Freud, S. (1915/1957). Repression. In J. Strachey (Ed. and Trans.), The standard editions of the complete psychological works of Sigmund Freud (Vol. 14). London: Hogarth Press. Gelkopf, M. (1997). Laboratory pain and styles of coping with anger. The Journal of Psychology, 13, 121–123. Greenberg, M. A., & Stone, A. A. (1992). Emotional disclosure about traumas and its relation to health: Effects of previous trauma and trauma severity. Journal of Personality and Social Psychology, 63, 75–84. Greenwood, K. A., Thurston, R., Rumble, M., Waters, S. J., & Keefe, F. J. (2003). Anger and persistent pain: Current status and future directions. Pain, 103, 1–5.

J Behav Med (2007) 30:455–469 Gross, J. J. (2002). Emotion regulation: Affective, cognitive, and social consequences. Psychophysiology, 39, 281–291. Janssen, S. J., Spinhoven, P., & Brosschot, J. F. (2001). Experimentally induced anger, cardiovascular reactivity, and pain sensitivity. Journal of Psychosomatic Research, 51, 479–485. Keefe, F. J., Rumble, M. E., Scipio, C. D., Giordano, L. A., & Perri, L. M. (2004). Psychological aspects of persistent pain: Current state of the science. Journal of Pain, 54, 195–211. Kelly, J. E., Lumley, M. A., & Leisen, J. C. C. (1997). Health effects of written emotional disclosure among rheumatoid arthritis patients. Health Psychology, 16, 57–67. Keogh, E., Ellery, D., Hunt, C., & Hannent, I. (2001). Selective attentional bias for pain-related stimuli amongst pain fearful individuals. Pain, 91, 91–100. Kerns, R. D., Rosenberg, R., & Jacob, M. C. (1994). Anger expression and chronic pain. Journal of Behavioral Medicine, 17, 57–67. LeBerge, D. (1995). Attentional processing. Cambridge, MA: Harvard University Press. Lees, A., Mogg, K., & Bradley, B. P. (2005). Health anxiety, anxiety sensitivity, and attentional bias for pictorial and linguistic healththreat cues. Cognition and Emotion, 19, 453–462. MacLeod, C., & Mathews, A. (1988). Anxiety and the allocation of attention to threat. Quarterly Journal of Experimental Psychology, 40A, 653–670. MacLeod, C., Mathews, A., & Tata, P. (1986). Attentional bias in emotional disorders. Journal of Abnormal Psychology, 95, 15–20. Martin, R., Wan, C. K., David, J. P., Wegner, E. L., Olson, B. D., & Watson, D. (1999). Style of anger expression: Relation to expressivity, personality, and health. Personality and Social Psychology Bulletin, 25, 1196–1207. Mayer, B., & Merckelbach, H. (1999). Unconscious processes, subliminal stimulation, and anxiety. Clinical Psychology Review, 19, 571–590. Melzack, R. (1975). The McGill Pain Questionnaire: Major properties and scoring methods. Pain, 1, 277–299. Melzack, R. (1987). The short-form of the McGill Pain Questionnaire. Pain, 30, 191–197. Mogg, K., Bradley, B. P., Williams, R., & Mathews, A. (1993). Subliminal processing of emotional information in anxiety and depression. Journal of Abnormal Psychology, 102, 304–311. Muraven, M., & Baumeister, R. F. (2000). Self-regulation and depletion of limited resources: Does self-control resemble of muscle? Psychological Bulletin, 126, 247–259. Okifuji, A., Turk, D. C., & Curran, S. L. (1999). Anger in chronic pain: Investigations of anger targets and intensity. Journal of Psychosomatic Research, 47, 1–12.

469 Page, A. C., Locke, V., & Trio, M. (2005). An online measure of thought suppression. Journal of Personality and Social Psychology, 88, 421–431. Pilowsky, I., & Spence, N. D., (1976). Pain, anger, and illness behavior. Journal of Psychosomatic Research, 20, 411–416. Quartana, P. J., & Burns, J. W. (2007). Painful consequences of anger suppression. Emotion, 7, 400–414. Rosenthal, R., & Rosnow, R. L. (1985). Contrast analysis: Focused comparisons in the analysis of variance. New York: Cambridge University Press. Schwarz, N. (1990). Feelings as information: Informational and motivational functions of affective states. In R. M. Sorrentino & E. T. Higgins (Eds.), Handbook of motivation and cognition: Foundations of social behavior (Vol. 2, pp. 527–561). New York: Guilford Press. Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimental studies: New procedures and recommendations. Psychological Methods, 7, 422–445. Smyth, J. M., Stone, A. A., Hurewitz, A., & Kaell, A. (1999). Effects of writing about stressful experiences on symptom reduction in patients with asthma or rheumatoid arthritis: A randomized trial. JAMA: Journal of the American Medical Association, 281, 1304–1309. Spielberger, C. D., Johnson, E. H., Russell, S. F., Crane, R. J., Jacobs, G. A., & Worden, T. J. (1985). The experience and expression of anger: Construction and validation of an Anger Expression Scale. In M. A. Chesney & R. H. Rosenman (Eds.), Anger and hostility in cardiovascular and behavioral disorders (pp. 5–30). Washington D.C.: Hemisphere. Sullivan, M. J. L., Rouse, D., Bishop, S., & Johnston, S. (1997). Thought suppression, catastrophization, and pain. Cognitive Therapy and Research, 21, 555–568. Tabacknik, B. G., & Fidell, L. S. (1996). Using multivariate statistics (3rd Ed.). New York: Harper Collins. Wegner, D. M. (1994). Ironic processes of mental control. Psychological Review, 101, 34–52. Wegner, D. M., & Erber, R. (1992). The hyperaccessibilty of suppressed thoughts. Journal of Personality and Social Psychology, 63, 903–912. Wegner, D. M., Erber, R., & Zanakos, S. (1993). Ironic processes in the mental control of mood and mood-related thought. Journal of Personality and Social Psychology, 65, 1093–1104. Wenzlaff, R. M., & Wegner, D. M. (2000). Thought suppression. Annual Review of Psychology, 51, 59–91. Zajonc, R. B. (2000). Thinking and feeling. In J. P. Forgas (Ed.), Feeling and thinking (pp. 31–58). New York: Cambridge University Press.

123