The induction of social pessimism reduces pain responsiveness

Introduction

Pain has been shown to be affected by expectations of future events. These expectations may take the form of very specific expectations that refer, for example, to the outcome of an impending surgery, or they may take the form of very general beliefs about one’s personal future [1], [2], [3]. General optimism is the most prominent and widely studied example of the latter form of expectation, by which the future appears to be associated with well-being, vigor, and positive emotions. Not surprisingly, optimists defocus pain-related stimuli and appear to be less responsive to pain than non-optimists [4, 5]. It is tempting to assume that pessimists should show the opposite pattern, with a stronger responsiveness and higher vulnerability to pain.

However, a series of experiments by DeWall and Baumeister [6] found the opposite: Receiving an ostensibly diagnostic forecast of a socially lonely future with no fulfilling relationships reduced participants’ responsiveness to pain, as indicated by both higher pain thresholds and tolerance thresholds. The intriguing finding that social pessimism^[1] did the same as optimism should be replicated.

In the present study, we aimed to both replicate and elaborate on this finding. For the latter purpose, we added psychological covariates that are likely to make individuals more or less vulnerable to pain, which might augment or counteract the effects of induced social pessimism, i.e., pain catastrophizing, dispositional optimism and self-esteem. Sullivan’s conceptualization of pain catastrophizing [9] has focused on the interpersonal nature of this construct and describes it as a way of communicating pain to others by engaging in pain behaviors. Dispositional optimism can buffer the negative influence of catastrophizing on pain response and may act as counterbalance [10]. A concept very similar to optimism is self-esteem, which is defined as a generalized positive–negative attitude, thinking and feeling, toward himself or herself [11]. Thus, we selected respective questionnaires as covariates, assessing attitudes, which may enhance or dampen the influence of social pessimism.

Because the presence or absence of others modulates pain processing [12], [13], [14] and an absence of others in the future is predicted in social pessimism, we assumed it critical who is the agent in pain stimulation. Therefore, we implemented two conditions where the pain was experimentally induced by a human or a machine as social prototypes for being together with others or alone.

We hypothesized that (i) expectations of a lonely future would promote pain unresponsiveness as in the previous study by DeWall and Baumeister [6], (ii) individuals with high trait optimism and self-esteem would be less susceptible to this intervention, whereas those with high trait pain catastrophizing would be more susceptible because their tendency to cognitively amplify their pain would increase the effect of their negative social expectations, and (iii) the effect of social expectations for the future on present pain sensitivity would also depend on whether the experimental pain was applied by a human or a machine, with the assumption that a human experimenter causing pain would make social aspects more salient and thus lead to stronger effects of the expectation of social exclusion in the future.

Methods

Material and protocol

Protocol

The procedure of this study was adapted from DeWall and Baumeister [6], but we modified it slightly in two ways. First, we were interested in the moderating role of social interaction and therefore introduced a second type of pain stimulation without direct contact with the experimenter by using a computer-controlled algometer. Second, we aimed to enhance the credibility of the experimental manipulation and added an implicit test instead of basing the bogus feedback on one questionnaire only. For this purpose, a dot-probe task with emotional faces as stimuli was administered in addition to a personality questionnaire (EPQ-R), and feedback on both tests was provided in a more detailed manner in order to make the “social future forecast” more convincing.

Participants took part individually in this study, which was advertised as a study for investigating the effect of personality factors on pain sensitivity. The advertised study purpose also formed the common cover story for all three groups. The sessions took place between 8 a.m. and 1 p.m. and lasted for approximately 1.5 h. The female experimenter wore a lab coat during the sessions in order to minimize experimenter gender effects [17], enhance credibility, and ensure professional distance from the participant. Also for guaranteeing professional distance, participants were addressed in the formal German language.

Figure 1 provides an illustration of the experimental paradigm, which will be explained in more detail in the following paragraphs.

Figure 1:

Illustration of the experimental protocol.

Materials

Pain measurements

For the pain measurements, two different pressure algometers were used: a hand-held pressure algometer with a probe area of 1 cm² (Somedic Sales AB, Algometer type II, Sweden), which was also used in the original study by DeWall and Baumeister [6], and additionally a PC-controlled pressure algometer (Mermaid Instruments, Aalborg University, Denmark). The PC-controlled algometer consisted of a rounded aluminum footplate with a 1 cm² padded probe, which was fixed to the tip of a piston and moved by an electric motor. Pressure stimulation was controlled via a built-in force transducer. This PC-controlled pain stimulator was introduced to implement a condition with low social impact. During use, the participant was not able to see the experimenter. The use of two stimulators was announced to the subjects as measure to increase reliability.

Pain threshold and tolerance measurements were first taken by using the handheld algometer and then using the computer-controlled algometer (see Figure 1). The fixed order was not optimal but necessary for ethical reasons because we did not want those participants, who just learnt to be alone in the future, to be left alone immediately thereafter. The hands were placed with the volar side facing up on a hard surface. For each stimulator, two practice trials were conducted by applying pressure to the tips of both ring fingers followed by six test trials to measure pain threshold and six test trials to measure pain tolerance (three on each side of the body). These test trials were conducted by applying pressure to the tips of the left and right index fingers (handheld algometer) and to the tips of the left and right middle fingers (PC-controlled algometer). The sequence of body sides was balanced across participants. In each trial, the pressure was increased from 0 kPa at a rate of change of 50 kPa/s either by the experimenter or by the electric motor until the participant pressed a button. The handheld algometer by Somedic allowed the experimenter to see and control the rate of change by watching a display. Participants were instructed to press the button either when they felt the first pain sensation (pain threshold) or when the pain reached an intolerable level (pain tolerance threshold). The score recorded at the time of the button press was the measure. The PC-controlled algometer had a built-in safety stop at an intensity of 761 kPa so that a higher pain tolerance could not be recorded exactly (see Section Data reduction and analysis - Pain tolerance threshold). There was an inter-stimulus interval (ISI) of 20 s between the single trials for each of the two thresholds as well as a longer ISI of 90 s between the pain threshold and pain tolerance measurements in order to prevent sensitization.

This whole procedure lasted for about 15 min and was conducted twice, once to obtain a baseline pre-measurement and once to obtain a post-measurement after the experimental intervention in which various levels of social pessimism were induced (see Figure 1).

Questionnaires

Participants completed a set of self-report questionnaires for assessing several psychologically relevant covariates: German versions of the Pain Catastrophizing Scale (PCS) [18, 19]; the Life-Orientation-Test Revised (LOT-R) [20, 21] and the Single-Item Self-Esteem Scale (SISE) [22].

The PCS [18, 19] was developed as a measure of catastrophizing related to pain. It contains 13 items that can be divided into three subscales, namely, rumination, magnification, and helplessness. The items (e.g., “I worry all the time about whether the pain will end”) are rated on a 5-point scale. As recommended in previous studies [18, 19], the total PCS score (range 0–52) was used. The German version showed good test-retest reliability (rtt=0.80) as well as high internal consistency (Cronbach’s α=0.92) for the total scale [19].

The LOT-R is designed to measure dispositional optimism. It consists of 10 items: three positively framed, three negatively framed, and four filler items. The items are rated on a 5-point scale. The total score for the LOT-R can be interpreted as a person’s generalized positive outcome expectancy. The total score for the German version showed a Cronbach’s α of 0.59 and a satisfactory test-retest reliability (rtt=0.75) [21].

The SISE [20] is a global measure of self-esteem. The single item “I have high self-esteem” was translated into German and rated on a 7-point scale. The item has been shown to have high convergent validity with the Rosenberg Self-Esteem Scale (RSE) [11], which is the most widely used self-esteem scale. Thus, the SISE can be used as a proxy for the RSE [22].

Bogus forecast of participants’ social future

To induce the impression that personality variables that are related to social behavior were assessed, we used two instruments, which were chosen due to their relative impenetrability: The Eysenck Personality Questionnaire – Revised (EPQ-R) [23] and a pictorial dot-probe task with emotional faces as a picture set [24]. The EPQ-R and the dot-probe were administered immediately after the other questionnaires (see Section Materials and protocol -Questionnaires and Figure 1). The “social future forecast” that was given later to the experimental groups was based on the bogus feedback from the alleged individual results of these tests (see Section Material and protocol -Bogus forecast of particpants’ social future).

The German Version of the EPQ-R [25] contains 107 items, which can be divided into four subscales: Psychoticism (32 items), Extraversion (23 items), Neuroticism (25 items) and Social Desirability (22 items). It contains various inscrutable questions, which make it difficult for the participants to guess what the instrument is supposed to measure. Participants were provided only with the information that specific personality factors were assessed and that their individual scores would be combined to form a global coefficient, which would then be compared with the population norm. This procedure was used to make it hard for participants to uncover the fact that the presented results were actually not related to their individual responses.

The version of the dot-probe task used in the present study was developed in a previous study for assessing attentional bias to pain faces [24]. It consists of monochromatic photographs of faces displaying different emotional expressions (anger, joy, pain and neutral) in 70 trials (shortened version) for the present study. In each trial, two pictures of faces (either one emotional face paired with a neutral face or two neutral faces) were presented on the left and right sides of the screen for up to 500 ms. Immediately thereafter, a dot appeared in the same position as one of the two pictures. Participants were instructed to indicate the location of the dot as fast as possible by using a response panel. The test yields attentional bias indices, which are calculated on the basis of the reaction times and can be interpreted as vigilance to or avoidance of emotional in comparison with neutral faces [24, 26, 27]. This dot-probe task was chosen for two reasons: First, due to its implicit nature, participants are not aware of their results so the task is ideally suited for providing bogus feedback. Second, the use of emotional faces as stimulus material supports the impression that variables that are related to social perception were actually measured.

After participants completed the two tests, the experimenter told the participants in the experimental groups to wait in the hallway for a few minutes so that she could assess the test results and provide the participants with feedback. The participants in the control group did not receive any feedback and were told only that the purpose of the short intermission was simply to maintain attention.

In the experimental groups, the experimenter introduced the feedback by stating that previous research had shown that both the personality questionnaire (EPQ-R) and the dot-probe task had proven to be reliable predictors of future social behavior and attachment. The experimenter then explained the subscales of the EPQ-R and the principle of the dot-probe task as well as the interpretation of the resulting bias indices. After this, participants were provided with the randomly assigned bogus feedback on their test results and received a “social future forecast,” which was allegedly based on their personal scores. In order to enhance the credibility of the bogus feedback, we used a “results profile” sheet that contained two tables displaying different values that were related to the EPQ-R and the dot-probe task.

Utilizing the EPQ-R results, we presented high values for psychoticism and neuroticism and low values for extraversion and social desirability for the negative forecast group, whereas this pattern was reversed for the positive forecast group. In addition, a normal distribution was displayed, on which the alleged position of the participant was indicated. The corresponding percentile rank was around 2% for the negative forecast group and around 98% for the positive forecast group, thus suggesting that the participant’s test results were either extremely below average (negative forecast) or extremely above average (positive forecast).

For the dot-probe task, the attentional bias indices for happy and angry faces were presented. For the negative forecast group, these indices indicated a vigilance-avoidance pattern (i.e., initial fixation followed by avoidance) for happy faces and a consistent avoidance of angry faces. Vice versa, values inserted for the positive forecast group indicated a consistent fixation on happy faces and a vigilant-avoidant pattern for angry faces.

Participants in the negative forecast group were told that their EPQ-R scores indicated a tendency toward insecurity, mood swings, and a lack of interest in other people’s opinions. Furthermore, the experimenter told them that their attentional bias pattern, which was derived from the dot-probe task, showed correlations with an insecure attachment style, which is characterized by having difficulties dealing with conflict situations. A brief remark was added to increase the plausibility of the feedback, saying that these things do not have to be obvious to the participant but are rather general personality propensities that often become obvious later in life. The social future forecast for this group was the following (adapted from [28]):

Based on these results, you are the personality type who is likely to be alone later in life. You may have friends and relationships now, but with time they will start to drift away. You may even have one or several romantic relationships, but these are likely to be short-lived. And, when you are past the age where people are constantly forming new relationships, the odds are you will probably end up being alone more and more.

Participants in the positive forecast group were told that their EPQ-R scores indicated that they are interested in the well-being of other people and that they exhibit very friendly behavior toward others. Furthermore, the experimenter told them that their attentional bias pattern derived from the dot-probe task showed correlations with extraversion and social competence (fixation on happy faces) as well as a tendency to avoid anger and a preference for harmonious togetherness (vigilance-avoidance for angry faces).

The social future forecast for this group was the following (adapted from [28]):

Based on these results, you are the personality type who will have rewarding relationships throughout life. You are likely to have a long and stable romantic relationship and friendships that will last into your later years. The odds are that you will always have friends and people who care about you.

Termination and debriefing

At the end of the experiment, participants were asked whether they had trusted in the feedback. Cases, in which participants had seriously questioned the feedback would have been recorded, which did however not happen. Thereafter, the participants were debriefed. The experimenter ensured that all participants were finally convinced that the feedback given by the experimenter had nothing to do with their test results and was based only on random assignment to the experimental groups. Moreover, the experimenter explained the necessity of deceiving participants about the true purpose of the experiment and sincerely apologized.

Results

Effects of the experimental manipulation on pain threshold (see Figure 2)

The ANOVA yielded a significant main effect of “time point” (F(1,72)=7.481, p=0.008, η²=0.094), with overall lower pain thresholds at T2 than at T1. However, this effect was more pronounced for the handheld algometer, indicated by a significant “time point” x “algometer” interaction (F(1,72)=10.595, p=0.002, η²=0.128). We also detected a significant main effect of “algometer” (F(1,72)=21.932, p<0.001, η²=0.233), which was driven by higher values for pain threshold obtained with the handheld algometer. Most interestingly, there was a trend toward a significant “group” x “time point” interaction (F(2,72)=2.927, p=0.060, η²=0.075): The pain threshold changes from T1 to T2 in the negative forecast group appeared to be different from the two other groups. The interaction between “group” and “algometer” (F(2,72)=0.327, p=0.722, η²=0.009) as well as the three-way interaction “group” x “time point” x “algometer” (F(2,72)=0.543, p=0.583, η²=0.015) clearly failed to reach significance.

Due to the findings suggesting differences between the algometers, we separately computed two repeated-measures ANOVAs with the within-subject factor “time point” (T1, T2) and the between-subjects factor “group” (positive forecast, negative forecast, control) for the two algometers.

For the handheld algometer, the ANOVA yielded a significant main effect of “time point” (F(1,72)=12.792, p=0.001, η²=0.151): The pain threshold generally decreased from T1 to T2 (see Figure 2). There was no significant “group” x “time point” interaction (F(2,72)=0.822, p=0.444, η²=0.022). However, post hoc tests showed that the pre-post decrease was significant in the control group (t(24)=3.008, p=0.006, d=0.36) and in the positive forecast group (t(24)=2.570, p=0.017, d=0.29) but not in the negative forecast group (t(24)=0.910, p=0.372, d=0.16); due to Bonferroni correction alpha=0.017.

Figure 2:

Pain threshold (M, SD) (in kPa) in the T1 and T2 measurement for the three experimental groups and the two pressure algometers.

For the PC-controlled algometer, we observed no main effect of “time point” (F(1,72)=0.043, p=0.836, η²=0.001) but a significant “group” x “time point” interaction (F(2,72)=4.547, p=0.014, η²=0.112). This interaction was further explored by separately computing dependent t-tests between the T1 and T2 measurements for each of the three groups. These post hoc tests indicated that pain threshold increased significantly from T1 to T2 only in the negative forecast group (t(24)=2.697, p=0.013, d=0.29), whereas there were no significant changes in the positive forecast group (t(24)=1.687, p=0.105, d=0.16) or in the control group (t(24)=0.557, p=0.582, d=0.06) (see Figure 2); due to Bonferroni correction alpha=0.017.

Effects of the experimental manipulation on pain tolerance

For the handheld algometer (see Figure 3), the ANOVA yielded a significant main effect of “time point” (F(1,72)=9.504, p=0.003, η²=0.117): Pain tolerance generally decreased from T1 (M=633.3, SD=152.5) to T2 (M=605.7, SD=135.5). There was no significant “group” x “time point” interaction (F(2,72)=2.025, p=0.139, η²=0.53). However, post hoc tests showed that the pre-post decrease was significant only in the control group (t(24)=2.943, p=0.007, d=0.36) but neither in the positive forecast group (t(24)=1.970, p=0.060, d=0.17) nor in the negative forecast group (t(24)=0.268, p=0.791, d=0.04); due to Bonferroni correction alpha=0.017.

Figure 3:

Pain tolerance threshold (M, SD) (in kPa) for the handheld algometer in the T1 and T2 measurement for the three experimental groups.

For the PC-controlled algometer, there was a descriptive increase in pain tolerance from T1 to T2 in the negative forecast group (T1: Mdn=537.3, T2: Mdn=549.7); however, this effect just failed to reach significance (T=7, p=0.073, r=−0.25). In the control group, there was a small descriptive pre-post decrease (T1: Mdn=664.17; T2: Mdn=650.0) which, however, clearly failed to reach significance (T=8, p=0.570, r=−0.08). In the positive forecast group, we observed a very small pre-post change (T1: Mdn=549.8, T2: Mdn=544.5) which was confirmed by a non-significant test statistic (T=9, p=0.444, r=−0.11).

Discussion

In the present study, we attempted to replicate DeWall and Baumeister’s [6] finding, which suggested that social pessimism induced by a bogus forecast of a participant’s personal future lowers pain. The replication was largely successful, that is, receiving an ostensibly diagnostic forecast of a lonely future life reduced responsiveness to experimental pain. More specifically, our pessimism intervention either rendered individuals pain unresponsive or tentatively prevented them from becoming more pain responsive in a situation that usually results in higher pain responsiveness. However, the effects of experimental manipulation were weak and could be substantiated only in post-hoc tests and not for all pain parameters. Thus, we could not give strong evidence that the pain-dampening effects of social pessimism shown for the first time by DeWall and Baumeister’s [6] are robust against contextual variations, which we introduced by a second type of pain stimulator and a methodological modification of the basis for the bogus forecast of a lonely future. This problem for perfect replication occurred although we more than doubled the number of subjects tested compared to the original study.

Contrary to our expectations, neither high trait-optimism nor high self-esteem protected individuals against these effects of social pessimism. Likewise, we failed to find that pain catastrophizing augmented it. Thus, these personality variables did not appear to be involved in modulating pain unresponsiveness or preventing pain responsiveness due to the forecast of a lonely future.

The question that therefore arises is which other factors not included in our hypotheses might have mediated the observed effects of social pessimism on pain responsiveness? A good guess is that the concerns elicited by the bogus forecast of a lonely future might have been enormous, thus causing a higher cognitive load in comparison with the other two conditions (positive forecast and control). It is well-known that pain can be reduced by producing competition for a limited amount of cognitive resources that are available for pain processing by adding another highly absorbing cognitive task [32], [33], [34]. Considering this explanation, a negative social forecast would be one form of potent cognitive distraction from pain with its known consequences. This cognitive hypothesis might also explain why in one case (pain tolerance assessment by the hand-held algometer) the positive forecast had a similar effect as the negative one, which both differed from the effect of the control condition. Concurrent positive expectations for the social future may also serve as still potent distractor from cognitive pain processing. Future studies might want to systematically examine whether acute social concern causes people to become distracted so that they fail to notice pain.

Another possibility is that emotional changes could have caused the effect. Past work has found that positive emotions reduce pain and negative emotions increase pain [35], [36], [37]. However, there are exceptions to this rule with one of them being especially informative in the present context. Depression has appeared to numb the sense of pain when people are tested with experimental approaches that are similar to the ones used in the present study, especially when applying brief pain stimuli [38], [39], [40]. Social pessimism is without a doubt an integral part of depressive syndromes and may therefore act in a similar way. For sure, there are always conceptual and empirical distances between chronic conditions such as depression and acute states such as experimentally induced social pessimism. Nevertheless, it may be interesting to investigate the similarities between social pessimism and depression in their effects on acute pain in future studies. There are also arguments that speak altogether against an explanation through emotions. DeWall and Baumeister [6] measured emotional responses in parts of their series of experiments and found that the negative forecast manipulation produced a kind of acute emotional numbness, that is, an absence of either positive or negative emotions. Furthermore, an explanation of the observed effects by an emotional impact on pain processing is also made unlikely because that at least in one case (pain tolerance assessed by the hand-held algometer) negative and positive expectations do not differ.

As expected, the social context mattered by modulating the effects of social pessimism on pain, but not in the hypothesized way. When the pain stimulator was driven by a computer and the experimenter remained in the background, the present findings were very similar to the original ones from DeWall and Baumeister [6]. Only the negative social forecast led to a significant change, which was an increase in the pain threshold after the intervention. However, when the experimenter came close to the participants, used a hand-held algometer, and personally caused the pain, participants appeared to be more pain responsive after no intervention; in contrast, if they received any forecast and, more reliably, if they received the negative forecast was the increase in pain responsiveness prevented.

Thus, the social context did not change the principal effect of the forecasted social exclusion on pain responsiveness but rather modified the manner of its appearance. When the machine appeared to cause the pain, the negative forecast led to true pain unresponsiveness; when the human appeared to cause the pain, the negative forecast prevented the normally occurring increase in responsiveness to pain after repeated stimulation. It is well-known that the presence or absence of others can be a major influence on pain sensitivity and pain responsiveness [12], [13], [14]. We know that, for example, the mere presence of familiar individuals such as significant others, the social reinforcement history (e.g., rewarding of pain behavior) shared with a present partner and the gender of the experimenter can affect the results of a pain test to give only a few examples [13, 17, 41]. Substantial differences occurred primarily when pain responsiveness was compared between situations with a stranger or no person present on the one hand and a familiar person present on the other hand. Compared with this, our social variation with no person (machine) and a stranger (experimenter) present was potentially too small to lead to very clear effects.

Contrary to our hypotheses, none of the investigated personality features such as trait-optimism, trait-pain catastrophizing or self-esteem had any moderating influences on the effects of the negative forecast for the future on pain responsiveness. It is rather unlikely that we erroneously missed the hypothesized influences because the largest amount of variance that was shared between changes in pain responsiveness and our personality measures did not exceed 6%, which would not reach significance even in much larger samples. The most obvious explanation is that social pessimism acts by taking another route as we expected. We already speculated earlier in this discussion about alternative forms of influence such as attentional distraction or lowered mood. Alternatively, it may well be the case that the conversion of dispositional optimism or pain catastrophizing into the situationally active state forms was not as straightforward as it was in other studies. For future studies on the present topic, it may therefore be advisable to assess both trait and state measures.

A limitation of the study may be that all participants encountered the same female experimenter. We tried to minimize potential gender effects of the female experimenter by asking her to wear a medical lab coat. Furthermore, we avoided any between-experimenter variance by keeping the experimenter constant across all participants. Nevertheless, a better gender-balanced design may have offered advantages in this respect. In our attempt to replicate the study by DeWall and Baumeister [6], we kept changes to the design to a minimum. According to the present interpretations, the assessment of mood changes and attentional absorption due to the forecasts of participants’ social future might be worthwhile additions for future investigations. Finally, the assessment of pain tolerance was compromised because too many subjects reached the safety limit of our PC-controlled algometer, which led to a shut-off, and therefore did not produce exact data so that the direct comparison between the two algometers was possible only for pain threshold, i.e., for the lower pain range. By contrast, our method of inducing social pessimism as administered stood the test and can be recommended for further use.

For ease of understanding the similarities and differences of our replication study compared to the original study by DeWall and Baumeister [6], we again highlight the critical items. We added a second algometer, which was PC-controlled, to the handheld algometer as used alone in the original study. We hoped to clarify the role of the pain activator (human – handheld vs. machine – PC-controlled). It was just this new device, which perfectly replicated the pain threshold findings of the original study. The pain tolerance findings were only tentatively successful replications. Furthermore, we added an intransparent implicit attention test to the personality questionnaire as second mean to increase the credibility of the bogus feedback but did not see major improvement compared to the original study.

Conclusions

Because we were able to largely replicate DeWall and Baumeister’s [6] finding that the forecast of a negative social future reduces pain responsiveness in a different laboratory using a slightly different design and slightly different methods, this seems to be a robust finding. Thus, optimism (even though we did not find that the positive social future manipulation led to increased pain thresholds in the present study in contrast to earlier studies) and pessimism must both be theoretically and empirically considered to reduce acute pain. It is very likely that the mechanisms of action are different in these diametrically opposed conditions, even though they are equally related to general and not pain-specific personal expectations of the future. Our observation that the degree of social contact between the experimenter and participant slightly influenced the pattern of results stresses the importance of also carefully considering the social context in which pain is examined in future studies on this topic.