J. Behav. Ther. & Exp. Psychiat. 48 (2015) 133e139
Contents lists available at ScienceDirect
Journal of Behavior Therapy and Experimental Psychiatry journal homepage: www.elsevier.com/locate/jbtep
Is trait resilience characterized by specific patterns of attentional bias to emotional stimuli and attentional control? €fer a, *, Hans-Ulrich Wittchen b, 1, Michael Ho €fler b, 2, Anke Heinrich b, 3, Judith Scha € nfeld a, 6 Peter Zimmermann c, 4, Stefan Siegel c, 5, Sabine Scho €t Dresden, Institute of Clinical Psychology and Psychotherapy, Chemnitzer Str. 46, D-01187 Dresden, Germany Technische Universita €t Dresden, Institute of Clinical Psychology and Psychotherapy, Center of Epidemiology and Longitudinal Studies (CELOS), Chemnitzer Technische Universita Str. 46, D-01187 Dresden, Germany c German Armed Forces Center of Military Mental Health, Scharnhorststraße 13, D-10115 Berlin, Germany a
b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 19 December 2014 Received in revised form 4 March 2015 Accepted 20 March 2015 Available online 28 March 2015
Background and objectives: Attentional processes have been suggested to play a crucial role in resilience defined as positive adaptation facing adversity. However, research is lacking on associations between attentional biases to positive and threat-related stimuli, attentional control and trait resilience. Methods: Data stem from the follow-up assessment of a longitudinal study investigating mental health and related factors among German soldiers. Trait resilience was assessed with the Connor-Davidson Resilience Scale and attentional control with the Attentional Control Scale. A subset of n ¼ 198 soldiers also completed a dot probe task with happy, neutral and threatening faces. Results: Attentional control was positively related to trait resilience. Results revealed no associations between both attentional biases and trait resilience. However, there was a negative association between attentional bias to threat and trait resilience when attentional control was low and a positive association between attentional bias to threat and trait resilience when attentional control was high. No such associations were found for attentional bias to positive stimuli. Limitations: Generalizability to other populations may be limited since we exclusively focused on male soldiers. Also, the cross-sectional design does not allow for causal conclusions. Conclusions: Findings suggest that attentional processing may promote trait resilience. Future research on preventive interventions should consider these findings. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Attentional control Attentional bias Resiliency/resilience Soldiers
1. Introduction Trait resilience is defined as a stress coping ability, which enables individuals to successfully adapt facing adversity (Connor & Davidson, 2003). Empirical evidence has shown that lower levels
* Corresponding author. Tel.: þ49 351 463 42233; fax: þ49 351 463 36984. E-mail addresses: [email protected] (J. Sch€ afer), hans-ulrich. [email protected] (H.-U. Wittchen), hoefl[email protected] € fler), [email protected] (A. Heinrich), peter1zimmermann@ (M. Ho bundeswehr.org (P. Zimmermann), [email protected] (S. Siegel), schoenfeld@ €nfeld). psychologie.tu-dresden.de (S. Scho 1 Tel.: þ49 351 463 36983; fax: þ49 351 463 36984. 2 Tel.: þ49 351 463 36921; fax: þ49 351 463 36984. 3 Tel.: þ49 351 463 42232; fax: þ49 351 463 36984. 4 Tel.: þ49 30 28411600; fax: þ49 30 28411603. 5 Tel.: þ49 160 92542971; fax: þ49 30 28411603. 6 Tel.: þ49 351 463 36958; fax: þ49 351 463 36984. http://dx.doi.org/10.1016/j.jbtep.2015.03.010 0005-7916/© 2015 Elsevier Ltd. All rights reserved.
of trait resilience are associated with an increased risk of developing mental disorders after stressful life events, e.g. PTSD (Lee, Ahn, Jeong, Chae, & Choi, 2014) as well as other anxiety (e.g. Scali et al., 2012), depressive (e.g. Edward, 2005; Kukihara, Yamawaki, Uchiyama, Arai, & Horikawa, 2014), and substance use disorders (Wingo, Ressler, & Bradley, 2014). Furthermore, trait resilience has been shown to predict treatment response in subjects with depression and PTSD (Davidson et al., 2012; Min, Lee, Lee, Lee, & Chae, 2012). Even though previous evidence suggests that trait resilience might protect from maladaptive outcomes and might help to recover from stressful life events little is known about cognitive characteristics and underlying mechanisms of trait resilience. This may be of pivotal importance regarding the development of empirical based interventions for promoting resilience. Theoretical accounts postulate that attentional processing may play a crucial role in trait resilience. Schwager and Rothermund (2013) proposed that attention is the core of cognition and affect,
134
€fer et al. / J. Behav. Ther. & Exp. Psychiat. 48 (2015) 133e139 J. Scha
which is responsible for adaptation in stressful situations. Theories of attention postulate two systems (e.g. Corbetta & Shulman, 2002) which can be related to the concepts of attentional control and attentional bias. Accordingly, attentional bias can be seen as a bottom-up, stimulus-driven attentional process, responsible for the detection and attentional holding of relevant stimuli. Attentional control is described as a top-down process, supposed to be responsible for preparation, regulation and application of goaldirected selective attention. Even though theoretical accounts suggest that attentional biases to positive and negative stimuli may encourage trait resilience (e.g. Schwager & Rothermund, 2013), little effort has been made to directly investigate these associations. However, indirect evidence comes from research investigating associations with variations in the serotonin transporter gene. Serotonin is an important neurotransmitter involved in different psychological processes. The 5HTTLPR polymorphism of serotonin transporter has been found to be associated with different mental disorders (e.g. Karg, Burmeister, Shedden, & Sen, 2011; Kenna et al., 2012). Therefore, one might assume that it is also involved in trait resilience. Accordingly, Stein, Campbell-Sills, and Gelernter (2009) found a negative association between the number of s-alleles of 5-HTTLPR and trait resilience. Additionally, Perez-Edgar et al. (2010) and Fox, Ridgewell, and Ashwin (2009) found that attentional bias for angry faces was positively associated with the number of long alleles of 5-HTTLPR and the reverse pattern was evident for attentional bias to happy faces. These findings suggest that trait resilience might be positively associated with attentional bias toward positive stimuli and negatively associated with attentional bias toward negative stimuli. However, to our best knowledge no study so far directly examined associations of trait resilience with attentional biases. Moreover, a better ability to control attention may enable individuals to decide which internal and external stimuli they attend to and thus promote adaptive emotion regulation (Troy & Mauss, 2011). This may support coping with adverse situations. Consistent with this proposition Eisenberg et al. (2004) found that effortful control, a superordinate construct including AC, predicted trait resilience in a longitudinal study in children. Furthermore, Bardeen, Fergus, and Orcutt (2014) found that higher attentional control predicted lower symptoms of PTSD in traumatized individuals compared to non-traumatized individuals. However, to our knowledge research is lacking on examining the relations between attentional control and trait resilience directly. According to theories about attentional processing attentional control and attentional biases are distinct systems but supposed to interact with each other (e.g. Corbetta & Shulman, 2002; Petersen & Posner, 2012). In line with this, Verwoerd, Wessel, de Jong and Nieuwenhuis (2009) found in a laboratory study that attentional bias and attentional control were related in the prediction of intrusions after watching a trauma film. Furthermore, Bardeen and Orcutt (2011) and Schoorl, Putman, Van Der Werff, and Van Der Does (2014) found that the interaction of attentional control and symptoms of PTSD was associated with attentional bias to threat. Results of the study of Bardeen and Orcutt (2011) using general threat stimuli (whereas Schoorl et al. (2014) used trauma-related stimuli) indicated that participants with strong AC and strong symptoms of PTSD showed attentional avoidance of threat whereas participants with poor AC and strong symptoms of PTSD showed an attentional bias towards threat. In summary, it can be proposed that attentional control may be associated with a differential association between attentional biases to threat and positive stimuli and trait resilience. Attentional control may allow individuals to cope with stress and regulate negative emotions by attending to positive stimuli and disengaging
the attention from threat-related, negative information (e.g. Gross, 2002; Troy & Mauss, 2011). 1.1. Present study This study aims at investigating the basic cognitive mechanisms of trait resilience, concentrating on attentional processing. We examined whether resilient individuals are characterized by attentional biases to emotional stimuli, i.e. avoiding threat and turning to positive stimuli relative to neutral stimuli, indicating that resilient individuals use emotional stimuli for adaptive emotional responses (e.g. Gross, 2002; Troy & Mauss, 2011). This may be related to the ability to voluntarily control attention. Therefore, we examined the relationships between attentional biases, attentional control and trait resilience in German soldiers e a sample at increased risk of experiencing stressful life events. We expect a) a positive association between attentional bias to positive stimuli and trait resilience and a negative association between attentional bias to threat and trait resilience, b) a positive association between attentional control and trait resilience and c) that a higher attentional control is associated with heightened attentional bias to positive stimuli and more trait resilience and d) that a higher attentional control is associated with diminished attentional bias to threat and more trait resilience. Since attentional bias to threat, attentional control and trait resilience are associated with symptoms of anxiety disorders, respectively (Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg, & van IJzendoorn, 2007; Reinholdt-Dunne, Mogg, & Bradley, 2013; Scali et al., 2012), we tested these associations also by adjusting for these symptoms. Accordingly, since attentional bias to positive stimuli, attentional control and trait resilience are associated with symptoms of depression, respectively, (Armstrong & Olatunji, 2012; Edward, 2005; Reinholdt-Dunne et al., 2013) we tested these associations also by adjusting for these symptoms. Thereby, we tested whether the associations are only due to these specific symptoms but not to trait resilience. 2. Method Data were collected during the follow-up measurement of the longitudinal component of the study “Prevalence, incidence and determinants of PTSD and other mental disorders” (PID-PTSDþ3). A detailed description of the study's methods, design and findings has been published previously (Trautmann et al., 2014; Wittchen et al., 2012). 2.1. Participants Participants were recruited from the follow-up sample (n ¼ 383). N ¼ 198 participants provided complete data sets with all of the measurements used for this study purpose (see below). Six participants had to be excluded because they had answered less than 80% of the trials in the dot probe task correctly. Additionally, we excluded the only female soldier because of empirical evidence suggesting gender differences in attentional biases (Tran, Lamplmayr, Pintzinger, & Pfabigan, 2013) and lack of power. This resulted in a final sample of n ¼ 191 participants. Demographic and clinical characteristics of the sample are displayed in Table 1. 2.2. Self-reported measures Number of combat-related experiences and traumatic events. Potentially traumatic events according to DSM-IV A1-criterion (American Psychological Association, 2000) were assessed using a list from the military version of the fully standardized diagnostic interview of DIA-X/M-CIDI (Wittchen & Pfister, 1997; Wittchen,
€fer et al. / J. Behav. Ther. & Exp. Psychiat. 48 (2015) 133e139 J. Scha Table 1 Descriptive statistics on demographic data, deployment, symptoms of anxiety and depression, traumatic events, attentional control, trait resilience and attentional biases indices (n ¼ 191). Age (M, SD) Number of deployments (M, SD) Number of different combat-related events (M, SD) Number of traumatic events (M, SD) Educational level (n, %) low middle high Marital status (n, %) never been married married widowed/divorced HADS anxiety (M, SD) HADS depression (M, SD) ACS (M, SD) CD-RISC (M, SD) Threat bias (M, SD) Positive bias (M, SD)
28.0 1.6 6.2 1.1
(5.4) (1.4) (6.2) (1.9)
31 (16.2) 121 (63.4) 39 (20.4) 141 40 10 3.1 1.8 55.0 31.3 0.1 1.3
(73.8) (20.9) (5.2) (3.2) (2.4) (5.5) (4.8) (31.6) (34.4)
Note. Attentional bias indices are scaled in milliseconds. ACS ¼ Attentional Control Scale, CD-RISC ¼ Connor-Davidson Resilience Scale. HADS¼ Hospital Anxiety and Depression Scale. M ¼ Mean, SD¼ Standard deviation.
€ nfeld, Recknagel, & Siegert, 2009). Participants were instructed Scho to mark distressing events with regard to all areas of life they had experienced. Additionally, participants were asked which combatrelated events they had experienced (MHAT IV mod., based on Hoge, Auchterlonie, & Milliken, 2006; German: Wittchen et al., 2009). Participants who indicated the presence of traumatic events according to DSM-IV A1-criterion were subsequently asked whether they had reacted with anxiety or feelings of helplessness and horror (DSM-IV criterion A2) (American Psychological Association, 2000). Attentional control. Attentional control was assessed with the German version of the Attentional Control Scale (ACS, Derryberry & Reed, 2002, German: Busch & Hofer, 2012). This questionnaire consisted of 20 items asking how often participants are able to shift and focus the attention in different situations using a 4-point Likert scale (“almost never” to “always”, range 20e80). Psychometric qualities of this questionnaire have been demonstrated in several studies (Derryberry & Reed, 2002; Fajkowska & Derryberry, 2010). Internal consistency in the present sample was high (a ¼ .83). Trait Resilience. For the assessment of trait resilience we used the 10-item German version of the Connor-Davidson Resilience Scale (CD-RISC, Connor & Davidson, 2003). A 5-point Likert scale ranging from “not true at all” to “true nearly all of the time” were provided to answer the questions (range 0e40). Several studies demonstrated psychometric qualities of this questionnaire in a variety of samples (e.g. Campbell-Sills, Cohan, & Stein, 2006; Karairmak, 2010). Internal consistency in the current sample was high (a ¼ .85). Symptoms of anxiety and depression. Symptoms of anxiety and depression were assessed using the Hospital Anxiety and Depression Scale (HADS, Zigmond & Snaith, 1983). Seven questions for each domain were asked. A 4-point Likert scale indicating the frequency of each symptom was provided for answering the questions. Psychometric properties have been proven in different studies in regard to different populations (Bjelland, Dahl, Haug, & Neckelmann, 2002). Internal consistency in this sample was good and acceptable (anxiety: a ¼ .80, depression: a ¼ .77). 2.3. Dot probe task For the measurement of attentional biases to threat and positive stimuli we used an adapted version of the dot probe task described
135
in Sipos, Bar-Haim, Abend, Adler, and Bliese (2013). 36 different pictures of faces of twelve actors (half male) from the NimStim stimulus set (Tottenham et al., 2009) were presented in the task. Facial expressions were happy, threatening and neutral. The pictures (each picture: 3.6 * 4.8 cm) were placed one above the other with a distance of 1.4 cm between them. They were centered on a white background (9.8 * 6.1 cm) in the middle of an otherwise black display. The paradigm consisted of 120 trials. Each trial (see Fig. 1) started with a fixation cross (black, 500 ms) presented in the center of the white background. Then, two pictures of faces of the same actor were presented for 500 ms. Afterwards the probe (letter in black ink, E or F) appeared at the position of one of the pictures. Participants' task was to identify the probe as quickly as possible without sacrificing accuracy by pressing a mouse key in accordance with the instructions given before the task started. The trial finished with a blank screen (500 ms) followed by the next trial. In 96 trials pairs of faces with an emotional and a neutral expression (half happy, half threat) were presented. In half of them the probe appeared at the position of the picture with the emotional facial expression (congruent trials) and in the other half after the picture with the neutral facial expression (incongruent trials). In the remaining 24 trials two pictures with neutral facial expressions were presented. The probe's type (E or F), sequence of pictures and the location of the emotional picture and the probe (up/down) were randomized. Participants were placed in a distance of 50 cm to the computer screen. The dot probe task was conducted with EPrime software package (Psychology Software Tools, Pittsburgh) using PA Dell Vostro 1015 laptops (Dell, Texas, USA) with a 15.600 display. We used facial expressions as stimuli because they are assumed to be superior to emotional words in measuring attentional biases. Using facial expressions as signs of danger (threat) or safety (happy) might be an evolutionary developed adaptive process (BarHaim et al., 2007). Accordingly, it is supposed that they have a higher ecological validity (e.g. Bradley, Mogg, Falla, & Hamilton, 1998; Staugaard, 2009). Additionally, Bradley et al. (Bradley et al. 1997, 1998) argued that the use of words involves the risk of confounding effects due to increased usage of subjective more relevant words compared to subjective less relevant words. 2.4. Procedure Participants were interviewed individually in buildings of the German Federal Armed Forces or at home by trained staff of the €t Dresden. All participants were asked to Technische Universita provide informed consent according to the ethic protocol (approved by the university's Ethics Board, EK 72022010 and in line with ICH-Good Clinical Practice Guidelines). Participation was voluntary and no compensation was provided. Participants could withdraw from the study at any time without giving reasons. Other soldiers and seniors were blind regarding participation to avoid any dissimulation. The interview started with the DIA-X/M-CIDI (Wittchen & Pfister, 1997; Wittchen et al., 2009) with supplemental questionnaires used for this study purpose. Afterwards participants could give hair strands for other study purposes. Then, four cognitive affective tests were conducted including the dot probe task. 2.5. Data preparation First, data of the dot probe task were analyzed for substantive plausibility. When participants answered more than 90% of the task incorrectly responses were recoded because we assumed that the participant mixed up the keys indicating the correct answer (n ¼ 14). For the calculation of means of reaction times (RT) for
136
€fer et al. / J. Behav. Ther. & Exp. Psychiat. 48 (2015) 133e139 J. Scha
Fig. 1. Sequence of events in incongruent and congruent trials of the dot probe task.
incongruent and congruent trials only data were included if the task was answered correctly, the RT was greater than 150 ms and less than 2000 ms, the RT deviated not more than 2.5 SDs of the participants own RT mean and not more than 2.5 SDs from the RT mean of the same trial type of the whole follow-up sample. This procedure is in accordance with data preparation of Abend et al. (2013). In sum, 7.0% of all trials were excluded. In a second step, attentional biases to positive and threatening stimuli were calculated by subtracting the mean of RTs of congruent trials from the mean of RTs of incongruent trials. Accordingly, a positive value indicates an attentional bias towards emotional stimuli and a negative value indicates attentional avoidance of emotional stimuli.
confounders. Additionally, since trait resilience is associated with trauma history (Scali et al., 2012) we adjusted also for number of traumatic experiences. b) We added to a) in case of attentional bias to threat symptoms of anxiety and in case of attentional bias to positive stimuli symptoms of depression. This was done in order to take symptoms of the domain associated with the respective attentional bias into account, whereby attentional control was assumed to be related to both domains and both variables were adjusted for. Statistical significance was evaluated at the 5% level (two-sided).
2.6. Data analysis
Descriptive statistics of the analyzed variables are displayed in Table 1.
Our data analysis was based on ordinary linear regression models. Additionally we fitted two robust alternatives for each model to avoid misleading results due to potentially violated model assumptions: a) linear regression with standard errors using the robust sandwich method and b) robust linear regression which weights the observations in order to avoid that single observations dominate the results. Since the results were similar and lead to the same conclusions, we report only the results of ordinary linear regressions. Analyses were conducted with the “regress” procedure in Stata 12.1 (Stata Corp., 2012). We started our analysis with 1) testing the overall association between trait resilience and attentional biases by regressing trait resilience on each attentional bias, 2) We then tested the association between attentional control and trait resilience by regressing trait resilience on attentional control. 3) In order to test associations between attentional biases, attentional control and trait resilience we calculated main effect models by regressing trait resilience on each attentional bias in separate models with attentional control as additional covariate and regressed the interaction of attentional control and trait resilience on each attentional bias, respectively. For 1) through 3) we fitted two models each: a) Age, educational level and marital status were added as potential
3. Results
3.1. Associations between trait resilience and attentional biases Results of regressions on the association between trait resilience and attentional bias to threat-related stimuli revealed no significant results in both models (a: b ¼ 0.34, CI: 0.63, 1.31, p ¼ 0.490; b: b ¼ 0.21, CI: 0.83, 1.25, p ¼ 0.693). Also results on the association between trait resilience and attentional bias to positive stimuli revealed no significant overall associations in both models (a: b ¼ 0.38, CI: 1.44, 0.68 p ¼ 0.479; b: b ¼ 0.25, CI: 1.41, 0.91, p ¼ 0.676).
3.2. Association between resilience and attentional control Trait resilience and attentional control were positively associated in both models (a: b ¼ 0.58, CI: 0.44, 0.73, p < 0.001; b: b ¼ 0.58, CI: 0.42, 0.75, p < 0.001). This means that an increase of one attentional control unit was associated with an increase of 0.58 units of trait resilience, which is a relatively strong positive association.
€fer et al. / J. Behav. Ther. & Exp. Psychiat. 48 (2015) 133e139 J. Scha
137
3.3. Associations between resilience, attentional control and attentional biases Table 2 shows the associations for the models on the interactions between attentional control and trait resilience on attentional bias to threat and positive stimuli, respectively. We found only evidence for attentional control predicting a differential association between trait resilience and attentional bias to threat. Fig. 2 illustrates this finding: For low attentional control (values 54) a negative association between attentional bias to threatrelated stimuli and trait resilience is predicted. With attentional control values higher than 54 the association between attentional bias to threat and trait resilience becomes increasingly positive. 4. Discussion The aim of this study was to investigate whether trait resilience is characterized by specific patterns of attentional biases to emotional stimuli and attentional control abilities. Based on theoretical accounts of Schwager and Rothermund (2013) and related evidence (Fox et al., 2009; Perez-Edgar et al., 2010; Stein et al., 2009) we expected attentional biases to threat and positive stimuli to be associated with trait resilience. However, interestingly, regarding threat this was only the case when we considered attentional control. Specifically, when attentional control was high results indicate that attentional allocation towards threat is associated with heightened trait resilience and attentional avoiding of threat is associated with less trait resilience. When attentional control was low and individuals showed attentional avoidance of threat they showed heightened trait resilience whereas attentional allocation towards threat stimuli was associated with less trait resilience. These results are in line with findings of Bardeen and Orcutt (2011) and Derryberry and Reed (2002). They found the inverse pattern of results in anxious individuals (Derryberry & Reed, 2002) and in individuals with symptoms of PTSD (Bardeen & Orcutt, 2011). However, since our results remain when we statistically accounted for symptoms of anxiety we were able to show that attentional bias to threat and attentional control are related to trait resilience and the associations were not due to symptoms of anxiety. The missing association between attentional bias to threat and trait resilience is not in line with findings of studies showing that symptoms of PTSD and depression are related to attentional avoiding of threat in soldiers from the Israeli Defense Force and United States (US) army. (Sipos et al., 2013; Wald, Shechner, et al., 2011; Wald et al., 2013). Associations of these samples may reflect our association found when attentional control was high.
Table 2 Results of regressions on the interactions between attentional control and trait resilience on attentional biases. Model a
b Threat bias ME CD-RISC CD-RISC x ACS Positive bias ME CD-RISC CD-RISC x ACS
Model b
b
95% CI
p
0.56 0.26
[0.56, 1.68] [ 0.12, 0.41]
0.327
Contents lists available at ScienceDirect
Journal of Behavior Therapy and Experimental Psychiatry journal homepage: www.elsevier.com/locate/jbtep
Is trait resilience characterized by specific patterns of attentional bias to emotional stimuli and attentional control? €fer a, *, Hans-Ulrich Wittchen b, 1, Michael Ho €fler b, 2, Anke Heinrich b, 3, Judith Scha € nfeld a, 6 Peter Zimmermann c, 4, Stefan Siegel c, 5, Sabine Scho €t Dresden, Institute of Clinical Psychology and Psychotherapy, Chemnitzer Str. 46, D-01187 Dresden, Germany Technische Universita €t Dresden, Institute of Clinical Psychology and Psychotherapy, Center of Epidemiology and Longitudinal Studies (CELOS), Chemnitzer Technische Universita Str. 46, D-01187 Dresden, Germany c German Armed Forces Center of Military Mental Health, Scharnhorststraße 13, D-10115 Berlin, Germany a
b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 19 December 2014 Received in revised form 4 March 2015 Accepted 20 March 2015 Available online 28 March 2015
Background and objectives: Attentional processes have been suggested to play a crucial role in resilience defined as positive adaptation facing adversity. However, research is lacking on associations between attentional biases to positive and threat-related stimuli, attentional control and trait resilience. Methods: Data stem from the follow-up assessment of a longitudinal study investigating mental health and related factors among German soldiers. Trait resilience was assessed with the Connor-Davidson Resilience Scale and attentional control with the Attentional Control Scale. A subset of n ¼ 198 soldiers also completed a dot probe task with happy, neutral and threatening faces. Results: Attentional control was positively related to trait resilience. Results revealed no associations between both attentional biases and trait resilience. However, there was a negative association between attentional bias to threat and trait resilience when attentional control was low and a positive association between attentional bias to threat and trait resilience when attentional control was high. No such associations were found for attentional bias to positive stimuli. Limitations: Generalizability to other populations may be limited since we exclusively focused on male soldiers. Also, the cross-sectional design does not allow for causal conclusions. Conclusions: Findings suggest that attentional processing may promote trait resilience. Future research on preventive interventions should consider these findings. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Attentional control Attentional bias Resiliency/resilience Soldiers
1. Introduction Trait resilience is defined as a stress coping ability, which enables individuals to successfully adapt facing adversity (Connor & Davidson, 2003). Empirical evidence has shown that lower levels
* Corresponding author. Tel.: þ49 351 463 42233; fax: þ49 351 463 36984. E-mail addresses: [email protected] (J. Sch€ afer), hans-ulrich. [email protected] (H.-U. Wittchen), hoefl[email protected] € fler), [email protected] (A. Heinrich), peter1zimmermann@ (M. Ho bundeswehr.org (P. Zimmermann), [email protected] (S. Siegel), schoenfeld@ €nfeld). psychologie.tu-dresden.de (S. Scho 1 Tel.: þ49 351 463 36983; fax: þ49 351 463 36984. 2 Tel.: þ49 351 463 36921; fax: þ49 351 463 36984. 3 Tel.: þ49 351 463 42232; fax: þ49 351 463 36984. 4 Tel.: þ49 30 28411600; fax: þ49 30 28411603. 5 Tel.: þ49 160 92542971; fax: þ49 30 28411603. 6 Tel.: þ49 351 463 36958; fax: þ49 351 463 36984. http://dx.doi.org/10.1016/j.jbtep.2015.03.010 0005-7916/© 2015 Elsevier Ltd. All rights reserved.
of trait resilience are associated with an increased risk of developing mental disorders after stressful life events, e.g. PTSD (Lee, Ahn, Jeong, Chae, & Choi, 2014) as well as other anxiety (e.g. Scali et al., 2012), depressive (e.g. Edward, 2005; Kukihara, Yamawaki, Uchiyama, Arai, & Horikawa, 2014), and substance use disorders (Wingo, Ressler, & Bradley, 2014). Furthermore, trait resilience has been shown to predict treatment response in subjects with depression and PTSD (Davidson et al., 2012; Min, Lee, Lee, Lee, & Chae, 2012). Even though previous evidence suggests that trait resilience might protect from maladaptive outcomes and might help to recover from stressful life events little is known about cognitive characteristics and underlying mechanisms of trait resilience. This may be of pivotal importance regarding the development of empirical based interventions for promoting resilience. Theoretical accounts postulate that attentional processing may play a crucial role in trait resilience. Schwager and Rothermund (2013) proposed that attention is the core of cognition and affect,
134
€fer et al. / J. Behav. Ther. & Exp. Psychiat. 48 (2015) 133e139 J. Scha
which is responsible for adaptation in stressful situations. Theories of attention postulate two systems (e.g. Corbetta & Shulman, 2002) which can be related to the concepts of attentional control and attentional bias. Accordingly, attentional bias can be seen as a bottom-up, stimulus-driven attentional process, responsible for the detection and attentional holding of relevant stimuli. Attentional control is described as a top-down process, supposed to be responsible for preparation, regulation and application of goaldirected selective attention. Even though theoretical accounts suggest that attentional biases to positive and negative stimuli may encourage trait resilience (e.g. Schwager & Rothermund, 2013), little effort has been made to directly investigate these associations. However, indirect evidence comes from research investigating associations with variations in the serotonin transporter gene. Serotonin is an important neurotransmitter involved in different psychological processes. The 5HTTLPR polymorphism of serotonin transporter has been found to be associated with different mental disorders (e.g. Karg, Burmeister, Shedden, & Sen, 2011; Kenna et al., 2012). Therefore, one might assume that it is also involved in trait resilience. Accordingly, Stein, Campbell-Sills, and Gelernter (2009) found a negative association between the number of s-alleles of 5-HTTLPR and trait resilience. Additionally, Perez-Edgar et al. (2010) and Fox, Ridgewell, and Ashwin (2009) found that attentional bias for angry faces was positively associated with the number of long alleles of 5-HTTLPR and the reverse pattern was evident for attentional bias to happy faces. These findings suggest that trait resilience might be positively associated with attentional bias toward positive stimuli and negatively associated with attentional bias toward negative stimuli. However, to our best knowledge no study so far directly examined associations of trait resilience with attentional biases. Moreover, a better ability to control attention may enable individuals to decide which internal and external stimuli they attend to and thus promote adaptive emotion regulation (Troy & Mauss, 2011). This may support coping with adverse situations. Consistent with this proposition Eisenberg et al. (2004) found that effortful control, a superordinate construct including AC, predicted trait resilience in a longitudinal study in children. Furthermore, Bardeen, Fergus, and Orcutt (2014) found that higher attentional control predicted lower symptoms of PTSD in traumatized individuals compared to non-traumatized individuals. However, to our knowledge research is lacking on examining the relations between attentional control and trait resilience directly. According to theories about attentional processing attentional control and attentional biases are distinct systems but supposed to interact with each other (e.g. Corbetta & Shulman, 2002; Petersen & Posner, 2012). In line with this, Verwoerd, Wessel, de Jong and Nieuwenhuis (2009) found in a laboratory study that attentional bias and attentional control were related in the prediction of intrusions after watching a trauma film. Furthermore, Bardeen and Orcutt (2011) and Schoorl, Putman, Van Der Werff, and Van Der Does (2014) found that the interaction of attentional control and symptoms of PTSD was associated with attentional bias to threat. Results of the study of Bardeen and Orcutt (2011) using general threat stimuli (whereas Schoorl et al. (2014) used trauma-related stimuli) indicated that participants with strong AC and strong symptoms of PTSD showed attentional avoidance of threat whereas participants with poor AC and strong symptoms of PTSD showed an attentional bias towards threat. In summary, it can be proposed that attentional control may be associated with a differential association between attentional biases to threat and positive stimuli and trait resilience. Attentional control may allow individuals to cope with stress and regulate negative emotions by attending to positive stimuli and disengaging
the attention from threat-related, negative information (e.g. Gross, 2002; Troy & Mauss, 2011). 1.1. Present study This study aims at investigating the basic cognitive mechanisms of trait resilience, concentrating on attentional processing. We examined whether resilient individuals are characterized by attentional biases to emotional stimuli, i.e. avoiding threat and turning to positive stimuli relative to neutral stimuli, indicating that resilient individuals use emotional stimuli for adaptive emotional responses (e.g. Gross, 2002; Troy & Mauss, 2011). This may be related to the ability to voluntarily control attention. Therefore, we examined the relationships between attentional biases, attentional control and trait resilience in German soldiers e a sample at increased risk of experiencing stressful life events. We expect a) a positive association between attentional bias to positive stimuli and trait resilience and a negative association between attentional bias to threat and trait resilience, b) a positive association between attentional control and trait resilience and c) that a higher attentional control is associated with heightened attentional bias to positive stimuli and more trait resilience and d) that a higher attentional control is associated with diminished attentional bias to threat and more trait resilience. Since attentional bias to threat, attentional control and trait resilience are associated with symptoms of anxiety disorders, respectively (Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg, & van IJzendoorn, 2007; Reinholdt-Dunne, Mogg, & Bradley, 2013; Scali et al., 2012), we tested these associations also by adjusting for these symptoms. Accordingly, since attentional bias to positive stimuli, attentional control and trait resilience are associated with symptoms of depression, respectively, (Armstrong & Olatunji, 2012; Edward, 2005; Reinholdt-Dunne et al., 2013) we tested these associations also by adjusting for these symptoms. Thereby, we tested whether the associations are only due to these specific symptoms but not to trait resilience. 2. Method Data were collected during the follow-up measurement of the longitudinal component of the study “Prevalence, incidence and determinants of PTSD and other mental disorders” (PID-PTSDþ3). A detailed description of the study's methods, design and findings has been published previously (Trautmann et al., 2014; Wittchen et al., 2012). 2.1. Participants Participants were recruited from the follow-up sample (n ¼ 383). N ¼ 198 participants provided complete data sets with all of the measurements used for this study purpose (see below). Six participants had to be excluded because they had answered less than 80% of the trials in the dot probe task correctly. Additionally, we excluded the only female soldier because of empirical evidence suggesting gender differences in attentional biases (Tran, Lamplmayr, Pintzinger, & Pfabigan, 2013) and lack of power. This resulted in a final sample of n ¼ 191 participants. Demographic and clinical characteristics of the sample are displayed in Table 1. 2.2. Self-reported measures Number of combat-related experiences and traumatic events. Potentially traumatic events according to DSM-IV A1-criterion (American Psychological Association, 2000) were assessed using a list from the military version of the fully standardized diagnostic interview of DIA-X/M-CIDI (Wittchen & Pfister, 1997; Wittchen,
€fer et al. / J. Behav. Ther. & Exp. Psychiat. 48 (2015) 133e139 J. Scha Table 1 Descriptive statistics on demographic data, deployment, symptoms of anxiety and depression, traumatic events, attentional control, trait resilience and attentional biases indices (n ¼ 191). Age (M, SD) Number of deployments (M, SD) Number of different combat-related events (M, SD) Number of traumatic events (M, SD) Educational level (n, %) low middle high Marital status (n, %) never been married married widowed/divorced HADS anxiety (M, SD) HADS depression (M, SD) ACS (M, SD) CD-RISC (M, SD) Threat bias (M, SD) Positive bias (M, SD)
28.0 1.6 6.2 1.1
(5.4) (1.4) (6.2) (1.9)
31 (16.2) 121 (63.4) 39 (20.4) 141 40 10 3.1 1.8 55.0 31.3 0.1 1.3
(73.8) (20.9) (5.2) (3.2) (2.4) (5.5) (4.8) (31.6) (34.4)
Note. Attentional bias indices are scaled in milliseconds. ACS ¼ Attentional Control Scale, CD-RISC ¼ Connor-Davidson Resilience Scale. HADS¼ Hospital Anxiety and Depression Scale. M ¼ Mean, SD¼ Standard deviation.
€ nfeld, Recknagel, & Siegert, 2009). Participants were instructed Scho to mark distressing events with regard to all areas of life they had experienced. Additionally, participants were asked which combatrelated events they had experienced (MHAT IV mod., based on Hoge, Auchterlonie, & Milliken, 2006; German: Wittchen et al., 2009). Participants who indicated the presence of traumatic events according to DSM-IV A1-criterion were subsequently asked whether they had reacted with anxiety or feelings of helplessness and horror (DSM-IV criterion A2) (American Psychological Association, 2000). Attentional control. Attentional control was assessed with the German version of the Attentional Control Scale (ACS, Derryberry & Reed, 2002, German: Busch & Hofer, 2012). This questionnaire consisted of 20 items asking how often participants are able to shift and focus the attention in different situations using a 4-point Likert scale (“almost never” to “always”, range 20e80). Psychometric qualities of this questionnaire have been demonstrated in several studies (Derryberry & Reed, 2002; Fajkowska & Derryberry, 2010). Internal consistency in the present sample was high (a ¼ .83). Trait Resilience. For the assessment of trait resilience we used the 10-item German version of the Connor-Davidson Resilience Scale (CD-RISC, Connor & Davidson, 2003). A 5-point Likert scale ranging from “not true at all” to “true nearly all of the time” were provided to answer the questions (range 0e40). Several studies demonstrated psychometric qualities of this questionnaire in a variety of samples (e.g. Campbell-Sills, Cohan, & Stein, 2006; Karairmak, 2010). Internal consistency in the current sample was high (a ¼ .85). Symptoms of anxiety and depression. Symptoms of anxiety and depression were assessed using the Hospital Anxiety and Depression Scale (HADS, Zigmond & Snaith, 1983). Seven questions for each domain were asked. A 4-point Likert scale indicating the frequency of each symptom was provided for answering the questions. Psychometric properties have been proven in different studies in regard to different populations (Bjelland, Dahl, Haug, & Neckelmann, 2002). Internal consistency in this sample was good and acceptable (anxiety: a ¼ .80, depression: a ¼ .77). 2.3. Dot probe task For the measurement of attentional biases to threat and positive stimuli we used an adapted version of the dot probe task described
135
in Sipos, Bar-Haim, Abend, Adler, and Bliese (2013). 36 different pictures of faces of twelve actors (half male) from the NimStim stimulus set (Tottenham et al., 2009) were presented in the task. Facial expressions were happy, threatening and neutral. The pictures (each picture: 3.6 * 4.8 cm) were placed one above the other with a distance of 1.4 cm between them. They were centered on a white background (9.8 * 6.1 cm) in the middle of an otherwise black display. The paradigm consisted of 120 trials. Each trial (see Fig. 1) started with a fixation cross (black, 500 ms) presented in the center of the white background. Then, two pictures of faces of the same actor were presented for 500 ms. Afterwards the probe (letter in black ink, E or F) appeared at the position of one of the pictures. Participants' task was to identify the probe as quickly as possible without sacrificing accuracy by pressing a mouse key in accordance with the instructions given before the task started. The trial finished with a blank screen (500 ms) followed by the next trial. In 96 trials pairs of faces with an emotional and a neutral expression (half happy, half threat) were presented. In half of them the probe appeared at the position of the picture with the emotional facial expression (congruent trials) and in the other half after the picture with the neutral facial expression (incongruent trials). In the remaining 24 trials two pictures with neutral facial expressions were presented. The probe's type (E or F), sequence of pictures and the location of the emotional picture and the probe (up/down) were randomized. Participants were placed in a distance of 50 cm to the computer screen. The dot probe task was conducted with EPrime software package (Psychology Software Tools, Pittsburgh) using PA Dell Vostro 1015 laptops (Dell, Texas, USA) with a 15.600 display. We used facial expressions as stimuli because they are assumed to be superior to emotional words in measuring attentional biases. Using facial expressions as signs of danger (threat) or safety (happy) might be an evolutionary developed adaptive process (BarHaim et al., 2007). Accordingly, it is supposed that they have a higher ecological validity (e.g. Bradley, Mogg, Falla, & Hamilton, 1998; Staugaard, 2009). Additionally, Bradley et al. (Bradley et al. 1997, 1998) argued that the use of words involves the risk of confounding effects due to increased usage of subjective more relevant words compared to subjective less relevant words. 2.4. Procedure Participants were interviewed individually in buildings of the German Federal Armed Forces or at home by trained staff of the €t Dresden. All participants were asked to Technische Universita provide informed consent according to the ethic protocol (approved by the university's Ethics Board, EK 72022010 and in line with ICH-Good Clinical Practice Guidelines). Participation was voluntary and no compensation was provided. Participants could withdraw from the study at any time without giving reasons. Other soldiers and seniors were blind regarding participation to avoid any dissimulation. The interview started with the DIA-X/M-CIDI (Wittchen & Pfister, 1997; Wittchen et al., 2009) with supplemental questionnaires used for this study purpose. Afterwards participants could give hair strands for other study purposes. Then, four cognitive affective tests were conducted including the dot probe task. 2.5. Data preparation First, data of the dot probe task were analyzed for substantive plausibility. When participants answered more than 90% of the task incorrectly responses were recoded because we assumed that the participant mixed up the keys indicating the correct answer (n ¼ 14). For the calculation of means of reaction times (RT) for
136
€fer et al. / J. Behav. Ther. & Exp. Psychiat. 48 (2015) 133e139 J. Scha
Fig. 1. Sequence of events in incongruent and congruent trials of the dot probe task.
incongruent and congruent trials only data were included if the task was answered correctly, the RT was greater than 150 ms and less than 2000 ms, the RT deviated not more than 2.5 SDs of the participants own RT mean and not more than 2.5 SDs from the RT mean of the same trial type of the whole follow-up sample. This procedure is in accordance with data preparation of Abend et al. (2013). In sum, 7.0% of all trials were excluded. In a second step, attentional biases to positive and threatening stimuli were calculated by subtracting the mean of RTs of congruent trials from the mean of RTs of incongruent trials. Accordingly, a positive value indicates an attentional bias towards emotional stimuli and a negative value indicates attentional avoidance of emotional stimuli.
confounders. Additionally, since trait resilience is associated with trauma history (Scali et al., 2012) we adjusted also for number of traumatic experiences. b) We added to a) in case of attentional bias to threat symptoms of anxiety and in case of attentional bias to positive stimuli symptoms of depression. This was done in order to take symptoms of the domain associated with the respective attentional bias into account, whereby attentional control was assumed to be related to both domains and both variables were adjusted for. Statistical significance was evaluated at the 5% level (two-sided).
2.6. Data analysis
Descriptive statistics of the analyzed variables are displayed in Table 1.
Our data analysis was based on ordinary linear regression models. Additionally we fitted two robust alternatives for each model to avoid misleading results due to potentially violated model assumptions: a) linear regression with standard errors using the robust sandwich method and b) robust linear regression which weights the observations in order to avoid that single observations dominate the results. Since the results were similar and lead to the same conclusions, we report only the results of ordinary linear regressions. Analyses were conducted with the “regress” procedure in Stata 12.1 (Stata Corp., 2012). We started our analysis with 1) testing the overall association between trait resilience and attentional biases by regressing trait resilience on each attentional bias, 2) We then tested the association between attentional control and trait resilience by regressing trait resilience on attentional control. 3) In order to test associations between attentional biases, attentional control and trait resilience we calculated main effect models by regressing trait resilience on each attentional bias in separate models with attentional control as additional covariate and regressed the interaction of attentional control and trait resilience on each attentional bias, respectively. For 1) through 3) we fitted two models each: a) Age, educational level and marital status were added as potential
3. Results
3.1. Associations between trait resilience and attentional biases Results of regressions on the association between trait resilience and attentional bias to threat-related stimuli revealed no significant results in both models (a: b ¼ 0.34, CI: 0.63, 1.31, p ¼ 0.490; b: b ¼ 0.21, CI: 0.83, 1.25, p ¼ 0.693). Also results on the association between trait resilience and attentional bias to positive stimuli revealed no significant overall associations in both models (a: b ¼ 0.38, CI: 1.44, 0.68 p ¼ 0.479; b: b ¼ 0.25, CI: 1.41, 0.91, p ¼ 0.676).
3.2. Association between resilience and attentional control Trait resilience and attentional control were positively associated in both models (a: b ¼ 0.58, CI: 0.44, 0.73, p < 0.001; b: b ¼ 0.58, CI: 0.42, 0.75, p < 0.001). This means that an increase of one attentional control unit was associated with an increase of 0.58 units of trait resilience, which is a relatively strong positive association.
€fer et al. / J. Behav. Ther. & Exp. Psychiat. 48 (2015) 133e139 J. Scha
137
3.3. Associations between resilience, attentional control and attentional biases Table 2 shows the associations for the models on the interactions between attentional control and trait resilience on attentional bias to threat and positive stimuli, respectively. We found only evidence for attentional control predicting a differential association between trait resilience and attentional bias to threat. Fig. 2 illustrates this finding: For low attentional control (values 54) a negative association between attentional bias to threatrelated stimuli and trait resilience is predicted. With attentional control values higher than 54 the association between attentional bias to threat and trait resilience becomes increasingly positive. 4. Discussion The aim of this study was to investigate whether trait resilience is characterized by specific patterns of attentional biases to emotional stimuli and attentional control abilities. Based on theoretical accounts of Schwager and Rothermund (2013) and related evidence (Fox et al., 2009; Perez-Edgar et al., 2010; Stein et al., 2009) we expected attentional biases to threat and positive stimuli to be associated with trait resilience. However, interestingly, regarding threat this was only the case when we considered attentional control. Specifically, when attentional control was high results indicate that attentional allocation towards threat is associated with heightened trait resilience and attentional avoiding of threat is associated with less trait resilience. When attentional control was low and individuals showed attentional avoidance of threat they showed heightened trait resilience whereas attentional allocation towards threat stimuli was associated with less trait resilience. These results are in line with findings of Bardeen and Orcutt (2011) and Derryberry and Reed (2002). They found the inverse pattern of results in anxious individuals (Derryberry & Reed, 2002) and in individuals with symptoms of PTSD (Bardeen & Orcutt, 2011). However, since our results remain when we statistically accounted for symptoms of anxiety we were able to show that attentional bias to threat and attentional control are related to trait resilience and the associations were not due to symptoms of anxiety. The missing association between attentional bias to threat and trait resilience is not in line with findings of studies showing that symptoms of PTSD and depression are related to attentional avoiding of threat in soldiers from the Israeli Defense Force and United States (US) army. (Sipos et al., 2013; Wald, Shechner, et al., 2011; Wald et al., 2013). Associations of these samples may reflect our association found when attentional control was high.
Table 2 Results of regressions on the interactions between attentional control and trait resilience on attentional biases. Model a
b Threat bias ME CD-RISC CD-RISC x ACS Positive bias ME CD-RISC CD-RISC x ACS
Model b
b
95% CI
p
0.56 0.26
[0.56, 1.68] [ 0.12, 0.41]
0.327
