Paradoxical Effects of Worrisome Thoughts Suppression: the Influence of Depressive Mood

Thought suppression increases the persistence of unwanted idiosyncratic worrisome thoughts when individuals try to suppress them. The failure of suppression may contribute to the development and maintenance of emotional disorders. Depressive people seem particularly prone to engage in unsuccessful mental control strategies such as thought suppression. Worry has been reported to be elevated in depressed individuals and a dysphoric mood may also contribute to the failure of suppression. No studies examine, however, the suppression of worrisome thoughts in individuals with depressive symptoms. To investigate the suppression effects of worrisome thoughts, 46 participants were selected according to the cut-off score of a depressive symptomatology scale, and they were divided in two groups (subclinical and nonclinical group). All of the individuals took part in an experimental paradigm of thought suppression. The results of the mixed factorial analysis of variance revealed an increased frequency of worrisome thoughts during the suppression phase, depending on the depressive symptoms. These findings confirm that a depressive mood can reduce the success of suppression.


Introduction
Thought suppression is defined as the effort not to think about a particular thought (Wegner et al. 1987). This is one of the most-used mental control strategies to control unwanted negative thoughts (Iijima and Tanno 2012). Paradoxically, however, suppression can increase the intrusiveness of the thoughts intended to be suppressed (Muris et al. 1996). Wegner et al. (1987) found that the suppression of unwanted white bear thoughts 1 led to an intensification of the suppressed thoughts in a post-suppression period (rebound effect). Lavy and van den Hout (1990) also reported an immediate enhancement effect during the thought suppression period when individuals were not allowed to use distraction.
In order to explain the paradoxical effects of thought suppression, Wegner (1994) developed the ironic processes theory of mental control. This theory proposes the existence of two simultaneous mechanisms in thought suppression: (1) an intentional operating process (IOP) that performs a voluntary and conscious search for alternative thoughts (with the purpose of replacing suppression target thoughts and distracting unwanted thoughts); and (2) an ironic monitoring process (IMP), automatic and unconscious, that constantly examines possible 1 In the classical study of The White Bear experiment, participants who were instructed to suppress thoughts about white bear reported more white bear thoughts than participants who were required to not suppress. failures in the IOP. The IOP is responsible for trying to produce efforts to achieve the desired mental state. For this purpose, the operating process needs intensive resources. In cases of adjusted cognitive resources, thought suppression may not only fail but may also prompt the opposite effect: the rebound effect. A dysphoric mood could also hamper the suppression process, resulting in higher ironic effects (Wegner et al. 1993).
According some authors (e.g., Wenzlaff et al. 1988;Wenzlaff 2005), depressed individuals are more likely than non-depressed individuals to suffer the ironic effects of suppression due to three main reasons: (a) depressed individuals depend more often on distractors emotionally associated with unwanted thoughts; (b) a depressed mood may consume resources of the operating process; and (c) depressed individuals often experience high levels of subjective stress that can minimize the ability to use distraction effectively. Wenzlaff et al. (1988) conducted a study with the objective to examine the skills and strategies used by depressed individuals to suppress thoughts. The results of this study were that: (a) depressed individuals use distractors with essentially negative content; (b) they display increased difficulty in thought suppression due to the use of memories related to their unwanted thoughts as a distraction strategy; and (c) they seem to reveal a deficit in the ability to suppress unwanted negative thoughts, experiencing a resurgence of suppressed thoughts as a result of suppression attempts (Wenzlaff et al. 1988).
On the other hand, worry is a transdiagnostic process across depressive and generalized anxiety disorders (Kircanski et al. 2015). The suppression of worrisome thoughts also seems to give rise to paradoxical effects (Becker et al. 1998), and, at the same time, the perceived failure of thought suppression effort produces an increase of worry associated with the failure of suppression (Beevers et al. 1999). Iijima and Tanno (2012) investigated the moderating effect of trait worry on the success of thought suppression related to everyday worries. They found that by suppressing their daily worries, individuals with a high trait worry reported higher rebound effect levels than individuals with low trait worry.
In summary, the tendency for excessive worry seems to be a common transdiagnostic construct to anxiety and mood disorders. Thought suppression is a strategy used by individuals to cope with their unpleasant thoughts (including worry); paradoxically, however, suppression may contribute to the development and maintenance of such disorders. Suppressing thoughts seems to be a difficult task that can be even more difficult if individuals are under dysphoric emotional states. The strategies employed to suppress unpleasant thoughts seem to contribute to the suppression success (i.e., reduction/elimination of the thought to suppress) or failure (i.e., increased frequency of the thought to suppress both during and after suppression).
Replicability in psychology is necessary to increase the reliability of findings and to extract valid conclusions (Asendorpf et al. 2013;Yong 2012). In the field of thought suppression, replicability is especially important because the results are frequently inconsistent. While some authors found ironic effects (e.g., Clark et al. 1991;Lavy and van den Hout 1990;Wegner et al. 1987), other authors failed to find it (e.g., Behar et al. 2005;Merckelbach et al. 1991;Rosenthal and Follette 2007). These mixed findings may be influenced by negative mood-related variables. In their meta-analysis, Abramowitz et al. (2001) suggested that the variability of results of the enhancement effects found in the different studies may be due to a negative mood; for this reason, we included mood variables in this study.
Another source of the results' variability may be related to instructions for suppression. Classical instructions for suppression based in the white bear paradigm included: BTry not to think about x …^. However, we don't know what the individuals do when they have to suppress a thought, and Salkovskis and Campbell (1994) found that the enhancement effect did not appear if participants were instructed to focus on a specific distraction task. We are interested in the strategies used by individuals when they have to suppress a thought, and if these strategies may be different depending on mood. We also want to know if differences exist between the perception of successful, the effort made and difficultly of suppression depending on real performance and mood.
Given this rationale, we aimed to replicate and extend the paradigm developed by Iijima and Tanno (2012) in depressed and non-depressed participants. We expect: (a) to replicate the results obtained from Ilma and Tanno finding a rebound paradoxical effect derived from worrisome thoughts suppression; (b) to find the occurrence of an enhancement effect depending on the subclinical symptoms of depression; (c) if nondepressed participants have more success in suppressing worrisome thoughts than the subclinical, then non-depressed participants will also show a wider range of strategies to suppress worries than the negative mood participants; (d) during the suppression phase, subclinical individuals, according to their negative mood, will evoke more neutral or negative thoughts (Wenzlaff et al. 1988) not related to the most worrisome target thoughts than non-depressive participants; and (e) according to their worse results in suppressing worrisome thoughts, individuals with depressive mood will show less perception of success, more effort, and more thought suppression difficulty, than non-depressed individuals.
Participants were recruited from the general population through the snowball sampling method. The study protocol was approved by the Scientific Committee of the Faculty of Social and Human Sciences of our university.

Measures
Penn State Worry Questionnaire (PSWQ; Meyer et al. 1990;Portuguese Version Jiménez-Ros 2011) The PSWQ measures the general tendency to excessive worry. Each item consists of five levels of response, in which 1 represents BNot typical at all^and 5 BVery typical.^In the Portuguese version of the PSWQ, 16 items used in this study showed an internal consistency of α = .88, suggesting a good reliability (Jiménez-Ros 2011). The internal consistency obtained from the sample in this study was acceptable (α = .74).

Hospital Anxiety and Depression Scale (HADS; Zigmond and Snaith 1983; Portuguese Version Pais-Ribeiro et al. 2007)
HADS is a 14-item scale with four response levels ranging from 0 to 3. HADS is divided into two subscales, each of them with seven items: The anxiety subscale (HADS-A) and the depression subscale (HADS-D). In the present study, we used the HADS-D. Scores between 0 and 7 were considered Bnormal,^between 8 and 10 Baverage,^between 11 and 14 Bmoderate,^and between 15 and 21 Bsevere.^The Portuguese version of the HADS showed a good internal consistency for the depression scale (α = .81) (Pais-Ribeiro et al. 2007). The internal consistency obtained for this study was acceptable (α = .73).

Procedure
The participants signed informed consent, completed a sociodemographic questionnaire and the PSWQ and the HADS-D. Participants were informed that there would be three phases of thought verbalization, but no mention was made that the experience would be related to thought or worry suppression.
We used the HADS-D to screen participants: those who scored above or equal to a cut-off score of eight were considered for the subclinical group (SG; n = 24), and the participants who scored below eight were considered for the nonclinical group (NG; n = 22).
The final sample of 46 was invited to participate in a laboratory experiment. Participants completed the different phases of the study in a quiet room, seated in a comfortable chair and facing the computer screen.
Similarly to Iijima and Tanno's (2012), in our study, participants trained for a three minute period before the experiment. We asked them to write down a word or phrase describing their thoughts when the computer randomly emitted a sound three times. We programed the Psychopy 2 (Peirce 2007) to produce the sounds.
Once the training period concluded, the experiment was run in three phases (initial expression, suppression and final expression). During the initial expression phase (IEP), we asked the participants to write their thoughts whenever they heard the sound (exactly as they did during the training period). The computer randomly made a sound 10 times over a period of 10 min, and the time gap between two consecutive sounds was 40 to 80 s.
After the first phase, the investigator asked the participants to describe their most distressing worrisome thought during the last five days and rate it regarding three aspects: (1) BIt's a difficult subject to handle^; (2) BI feel depressed when I think of it^; and (3) BI think regularly on this matter^, according to a 7-point scale ranging from 1 (Completely disagree) to 7 (Completely agree).
The investigator then asked, participants to rate the similarity between the content of their more most distressing worrisome thought and the content of the spontaneous thoughts described in the first experimental phase, according to a scale with the four categories used by Iijima and Tanno (2012). (1) BThe content of this thought is the same or very close to the most distressing worrisome thought^; (2) BThe content of this thought is about another matter but mainly related to worries or negative thoughts^; (3) BThe content of this thought is about another matter but mainly related to neutral or positive themes^; and (4) Ba blank mind.Ô nce these two tasks were completed, in the suppression phase (SP), we asked the participants not to think about their most distressing worrisome thought and to write down, again, a word or phrase that reflected what they were thinking when they heard the computer sound. Again, the computer randomly made a sound 10 times over a period of 10 min. We requested the classification of their thoughts in accordance with the previous rating scale.
During the final expression phase (FEP), the participants could think about whatever they like. We just required the participants to write down a word or phrase reflecting what they were thinking every time they heard the computer sound, which was again 10 times over a period of 10 min. After that, the participants rated their thought content according to the scale used in the previous phases.
After the experience, we assessed on an seven point scale (1 = nothing to 7 = extremely) the participants' subjective perception about the success, difficulty, and effort spent to suppress their most distressing worrisome thought; and in two open questions, the strategies they used to suppress the thought. The experience finished with a debriefing about the study objectives.

Statistical Analysis
The data were processed with the IBM SPSS software (version 22.0).
The relative frequency of the worrisome target thought was calculated in each phase (i.e., number of thoughts ranked with 1/10). If the suppression was successful, the relative frequency of 1 was low. That means, participants' spontaneous thoughts were not related to their most distressing worrisome thought. If the suppression was not successful, the relative frequency was higher, meaning that the participant thought about their most distressing worrisome thought during the whole 10 min of the suppression phase.
To verify if the paradoxical effects of the suppression of idiosyncratic recent daily worries depended on depressive symptoms, a mixed ANOVA (3 × 2) was performed. The between groups' factor was the depressive symptoms (SG vs. NG), and the within factor was the three experimental phases (IEP, SP, FEP).
For the SP, we also calculated the relative frequency of negative thoughts, but not related to the target worrisome thoughts, and the relative frequency of positive and neutral thoughts (i.e., number of thoughts ranked with 2/10 and 3/10). Low levels of these frequencies means that, during this phase, the participants did not evoke other negative/positive or neutral thoughts to suppress their most distressing worrisome thoughts. In contrast, high values (near 1), mean that the participants thought about negative/positive and neutral thoughts' issues to suppress their most worrisome thoughts. To verify whether if during the SP, the use of positive, neutral or negative thoughts, but not related to the most worrisome target thoughts, depends on the depressive state, we used the relative frequency of negative, positive, and neutral thoughts as a factor within an ANOVA (2 × 2); and the factor between groups was the depressive symptoms.
The partial eta squared (ƞ p 2 ) was used to estimate the effect size, with values lower than .010 being small effect magnitudes, between .058 and .13 average magnitudes, and higher than .13 large magnitudes (Cohen 1973).
To compare the significance of the differences between two average values for independent samples a t-test was used and was considered the value of Cohen's d for the effect size estimation. The values of d equal to .2 were considered small effect magnitudes, the values of .5 were considered average magnitudes, and the ones higher than .7 were considered large magnitudes (Cohen 1988).
To compare the significance of the differences between two average values for dependent samples a t-test was also used.
The magnitude of the difference was calculated using the following formula: r = ffiffiffiffiffiffiffiffi ffi t2 t2þdf q (Field 2009). The results of the magnitude of the effect were interpreted according to the criteria suggested by the author, where .10 is a weak association, .30 is an average association, and .50 is a strong association.
To analyze the predictive value of variables, we performed a linear regression analysis.
We classified the content of the recent daily worries reported by the participants into three themes (academic and labor issues, interpersonal relationships, health and welfare). The performed chi-square test indicated that these themes were not associated with the depressive symptoms groups (χ 2 (2) = .773, p = .680).

The Rebound Effect
To test our first hypothesis (the occurrence of a rebound effect derived from worrisome thoughts suppression), we replicated the procedure and analysis conducted for Iijima and Tanno (2012) (regression analysis with all independent variables centered and entered in one step). The result showed that a single predictor variable (relative frequency of the worrisome thought target suppression phase) contributed to the variability of the outcome variable (relative frequency of the target in the final expression phase) (Table 1). That means, the failure of worrisome thoughts during the suppression phase explained the rebound effect.

The Enhancement Effect of Depending on the Depressive Symptoms
To study our hypothesis about an enhancement effect depending on the subclinical symptoms of depression, we conducted a mixed ANOVA (3 × 2). The results suggest that this paradoxical effect of the suppression depends on the depressive symptom levels (F [1.690, 74.363] = 3.134, p = .058; ƞ p 2 = .066). Although, there were no differences in the frequency of the target thoughts due to depressive symptoms (F [1, 44] = 1.043; p = .313; η p 2 = .023), the occurrence of the most distressing worrisome thought depends on the experimental phase (F [1.690, 74.363] = 3.529; p = .042; η p 2 = .074). The visual inspection of Fig. 1 suggests that the evolution of the occurrence of the most distressing worrisome thought along the three experimental phases depends on depressive symptoms.
When both groups were analyzed separately, there was a significant increase in the relative frequency of the worrisome target thought in the subclinical group (t (23) = −3.757, p = .001; α ≤ .016 with Bonferroni correction; r = .62) from the IEP (M = .16, SD = .20) to the SP (M = .31, SD = .22). The observed decrease of frequency of worries from the SP to the FEP (M = .24, SD = .23) was not significant (t (23) = 1.690, p = .105, r = .33), neither was the observed increased frequency of worries from the IEP to the FEP (t (23) = −1.284, p = .212, r = .25).
In the nonclinical group, the occurrence of the most worrisome thought was very similar in the three experimental phases. There were no differences (t (21) = −.137, p = .892, r = .003) in the occurrence of the worrisome target thought from the IEP (M = .17, SD = .05) to the SP (M = .173, SD = .06) or from the IEP to FEP (M = .17, SD = .06).
These results suggest that the efforts of suppression increase the immediate occurrence of the worrisome target thought during the SP just in the subclinical group. Suppression does not seem to affect the occurrence of these thoughts in the nonclinical group. That means that depression levels moderated the immediate enhancement suppression effects of worrisome thoughts.

Strategies to Deal with the most Distressing Worrisome Thoughts in the Suppression Phase (SP)
According to previous results, it seems that SG was less successful in suppressing worrisome thoughts than in the NG group. We analyzed if there were any differences in the strategies used for the groups to suppress the thought target during the suppression phase. There was an association between the type of strategies used in the SP to cope w i t h t h e w o r r i s o m e t h o u g h t s a n d t h e g r o u p s (χ 2 (2) = 11.127, p = .004). During the suppression phase, participants of the SG used more cognitive distraction strategies to deal with the most distressing worrisome thought (91.7 %), while the non-clinical participants employed more behavioral distraction strategies (33.3 %) or did nothing (19.0 %).

Affective Valence of Thoughts (Negative Vs. Positive) during Suppression Phase (SP)
We further examined the affective valence (negative vs. neutral/positive) of non target related thoughts in the SP. After each phase, participants also rated their negative, but not related with the target worrisome thoughts, and the positive and neutral. We found that the content of the thoughts (negative, positive or neutral) of the participants during the SP depends on depressive symptoms (F [1, 1] = 3.099; p = .085; η p 2 = .066). The visual inspection of the graph (Fig. 2) and the analysis of multiple comparisons (α ≤ .025, after Bonferroni correction) suggested that, although, there are more positive thoughts in both groups, in the SG, the suppression of a worrisome thought did not lead to the occurrence of significant differences between negative thoughts and positive thoughts (t (23) = − 2.003, p = .054, r = .39). However, the NG showed significantly (t (21) = − 3.545, p = .002, r = .61) more thoughts with positive than negative affective valence.

Perception of Difficulty, Effort and Success in the Suppression Phase (SP)
To answer our final hypothesis, we analyzed the differences between groups in perceived difficulty, effort to suppress, and appraisal of suppression success. Groups did not differ neither in the perception of difficulty (t (44) = −.680, p = .500, d = −.21) nor in the perception of effort expended to suppress the recent daily worries (t (44) = −.200, p = .842, d = −.06). Nonetheless, the appraisal of the success was significantly lower (M = 3.50, SD = .29) for the SG (t (44) = −3.141, p = .003, d = − .95) than for the NG (M = 5.10, SD = .42).

Discussion
Thought suppression is a common strategy used by individuals to cope with their unpleasant thoughts (including worry). The effectiveness of this strategy to reduce unwanted thought frequency is still unclear. To achieve this success seems to be more difficult if individuals are under dysphoric emotional states. Our main objective was to analyze if the success of idiosyncratic worrisome thought suppression may be more difficult under depressive mood. We are also interested in replicating the results obtained by Iijima and Tanno (2012) and knowing the most efficient strategies used by the participants to reduce the frequency of unwanted worrisome thoughts.
In the current study, we used an experiment for suppression considering idiosyncratic worries, and examined the effects of suppressing worrisome thoughts in individuals with and without depressive symptoms. As expected, we found an association between the presence of trait worry and depression (e.g., McEvoy et al. 2013;Starcevic et al. 2007). The SG showed a higher tendency for excessive worry than the NG.
Although there were no differences in the content of the most distressing worrisome thoughts experienced by the participants during the days preceding the experience (according to the level of depressive symptoms), the SG rated their primary worry as more frequent, harder to bear, and more responsible for dysphoria (when compared to the nonclinical). This result can be explained by metacognitive models that consider the differences between normal and pathological worry as the result of metacognitive beliefs about it (evaluation of thought, ability to monitor and regulate), rather than the differences in the content of thought (Wells and Matthews 1996). It was not associated, however, with the trait of excessive worry, greater in participants of the SG. This result provides evidence for Wegner's (1994) perspective, which considers that the suppression affects the occurrence of worrisome thoughts but not the worrisome process itself.
Contrary to our predictions, just the frequency of target worrisome thoughts predicted the frequency of the final expression phase. That means that the rebound effect was not predicted by the interaction between the frequency of thought during the suppression phase and pathological worry. This result allows us to partially replicate the results obtained by Iijima and Tanno (2012) and highlights the importance of replicating research to explain the variability of results. Negative mood may be one of the variables involved in the variability of results.
Our results also showed a paradoxical immediate enhancement effect associated with depressive symptoms, and the participants with depressive symptoms presented a higher frequency of thoughts related with their idiosyncratic worry during suppression. This may mean that high levels of an active negative mood in individuals can deplete the cognitive resources required for suppression (Purdon 1999). Similar results were obtained by Wegner et al. (1987) and Ashton and Boschen (2011). These authors found that the suppression of unwanted thoughts leads to an immediate enhancement effect on the occurrence of such thought; they further demonstrated that the presence of depressive symptoms has an effect on the failure to suppress worrisome thoughts during the SP. Our findings are also consistent with the results of the study conducted by Conway et al. (1991), who found that, when instructed to suppress thoughts, dysphoric individuals had a higher rate of unwanted thoughts than non-dysphoric individuals. Depressed mood seems to threaten the suppression, allowing unwanted thoughts to return to mind immediately.
Participants with depressive symptoms realistically perceived their lower performance to suppress their idiosyncratic worrisome thoughts. However, they did not differ from the non-depressed individuals in the effort perceived or in the difficulty experienced in suppressing the worry.
As Murray et al. (2015) concluded in two experimental studies (comparison memory suppression between young and elderly adults), the failure of the participants' memory suppression appears to be associated more with the failure to select inhibitory strategies than with the incapacity to implement such strategies. In our study, the failure of thought suppression in individuals with depressive symptoms may be related to the type of strategies used to suppress the thoughts. In fact, the nonclinical group used a more diverse range of strategies to suppress thoughts (cognitive distraction, behavioral and doing nothing), while participants with depressive symptoms used mainly the cognitive distraction. Furthermore, the contents of the selected thoughts to implement the distraction during the SP were also different. The participants with depressed mood more often used thoughts with negative content than the participants without depressive symptoms. These results are consistent with those found by Wenzlaff et al. (1988), and may help to explain some results obtained by these authors, who found, in a longitudinal study (Wenzlaff and Bates 1998), an association between suppression and high levels of rumination. In a later investigation, they further found that the suppression efforts may, ironically, lead to depressive rumination (Wenzlaff and Luxton 2003). To confirm these results, future studies are required to manipulate the repertoires of strategies and the content of distractors during the worrisome thought suppression in individuals with or without depression.
Notwithstanding the relevance of our results, we faced some limitations during the development of this study that did not allow us to draw more general inferences from the data. Some aspects to highlight are the following: (1) the small size of the sample; (2) the heterogeneous group with clinical symptoms; and (3) the use of self-report measures may have biased the research results. Specifically, we cannot be assured that the participants have registered all occurrences of thought that should be suppressed. Nor can we be sure that due to social desirability, participants were able to correctly classify all of their main worrisome related thoughts; and (4) finally, although we did not find an association between the tendency for excessive worry and the individuals' evaluation about their most distressing worrisome thought, this variable may have influenced the results.
Future research may replicate this study using more homogeneous samples and with diagnosed depression, and a superior and representative sample size of the population would allow greater external validity. We also suggested controlling the role of the variable trait worry in the suppression of worrisome thoughts in individuals with depressive symptoms.
Despite these limitations, this study is believed to provide valuable insights about a field of increased importance and with clinical and scientific interest. These insights are referred in particular to the paradoxical effect of immediate enhancement during thought suppression and the importance of negative mood in the suppression of unwanted worrisome thoughts. The participants with depressive symptoms are aware of the failure of thought suppression; yet, they neither assign this task a greater difficulty nor consider making more effort than those without depressive symptoms. The lack of diverse strategies and the negative content of the distractors used by the individuals with depressive symptoms may have contributed to the suppression failure. The use of suppression and the type of strategies used to suppress worrisome thoughts can contribute to an increase of rumination and negative mood in these individuals. These results may inform clinicians about the importance of assessing and increasing (in therapeutic interventions with depressed individuals or with dysphoric mood) the repertoire of coping strategies and changing the contents of distractor thoughts to address the worry. Future research may modify the number and kind of strategies associated with suppression as well as the content of the distractors' thoughts in clinically depressed individuals to know the best strategies to suppress and cope with worrisome thoughts.

Compliance with Ethical Standards
Conflict of Interest The authors declare that they have no conflict of interest.
Funding This study was not funded.
Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Informed Consent Informed consent was obtained from all individual participants included in the study.