Working Memory and Emotional Interpretat

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

European Child & Adolescent Psychiatry

https://doi.org/10.1007/s00787-020-01656-8

ORIGINAL CONTRIBUTION

Working memory and emotional interpretation bias in a sample


of Syrian refugee adolescents
Sven C. Mueller1,2 · Cagla Unal3 · Michela Saretta1 · Fadwa Al Mughairbi4 · Joana Gómez‑Odriozola2 ·
Esther Calvete2 · Baris Metin5

Received: 5 June 2020 / Accepted: 25 September 2020


© Springer-Verlag GmbH Germany, part of Springer Nature 2020

Abstract
The number of adolescent refugees around the world has been continuously increasing over the past few years trying to
escape war and terror, among other things. Such experience not only increases the risk for mental health problems including
anxiety, depression, and post-traumatic stress disorder (PTSD), but also may have implications for socio-cognitive devel-
opment. This study tested cognitive-affective processing in refugee adolescents who had escaped armed conflict in Syria
and now resided in Istanbul, Turkey. Adolescents were split into a high trauma (n = 31, 12 girls, mean age = 11.70 years,
SD = 1.15 years) and low trauma (n = 27, 14 girls, mean age = 11.07 years, SD = 1.39 years) symptom group using median
split, and performed a working memory task with emotional distraction to assess cognitive control and a surprise faces task
to assess emotional interpretation bias. The results indicated that high (vs. low) trauma symptom youth were ~ 20% worse
correctly remembering the spatial location of a cue, although both groups performed at very low levels. However, this find-
ing was not modulated by emotion. In addition, although all youths also had a ~ 20% bias toward interpreting ambiguous
(surprise) faces as more negative, the high (vs. low) symptom youth were faster when allocating such a face to the positive
(vs. negative) emotion category. The findings suggest the impact of war-related trauma on cognitive-affective processes
essential to healthy development.

Keywords War · Trauma · Adolescents · Refugee · Emotion · Working memory · interpretation bias

Introduction of global refugees has been estimated at 25.9 million, with


41.3 million being internally displaced due to violence and
Increasing worldwide war, conflict, or political turmoil is conflict [1]. The continuing armed conflict in Syria forced
leading to an ever-greater number of migrants wishing to around 2.7 million people to seek refuge in Turkey, many
escape these desperate situations. In 2018 alone, the number among them minors [2]. Yet, the cognitive-affective conse-
quences of these experiences and displacement from their
homes in youths are still little understood.
* Sven C. Mueller Experience of war and conflict has many consequences
Sven.Mueller@UGent.be
including, but not limited to, scarcity of food, loss of jobs,
1
Dept. of Experimental Clinical and Health Psychology, death, physical, emotional, sexual traumatic experiences
Faculty of Psychology and Educational Sciences, Ghent including life threats, beatings, rape, murder, torture, impris-
University, Ghent, Belgium onment, and disappearance of family [3, 4]. Furthermore,
2
Department of Personality, Psychological Assessment once in safety, the post immigration phase continues to carry
and Treatment, University of Deusto, Bilbao, Spain stressful consequences such as the loss of contact with the
3
Graduate School of Psychology, Özyeğin University, family and social network, culture shock, and an inability to
Istanbul, Turkey communicate (foreign language). Particularly children and
4
Psychology Program, College of Humanities and Social teenagers are at elevated risk of encountering such stressful
Sciences, United Arab Emirates University, Al Ain, life experiences, which may leave long-term mental (and
United Arab Emirates
physical) scars and result in a reduced quality of life and low
5
Department of Psychology, Uskudar University, Istanbul, life attainment goals. Many children stay in refugee camps
Turkey

13
Vol.:(0123456789)
European Child & Adolescent Psychiatry

for several years, facing difficult living conditions and lack that may drive such effects, Gutiérrez-García [22] examined
of education [4–6]. eye movements and documented an interpretation bias that
Youths are double at risk, because on one hand being was linked to the parts of the face individuals focused on.
exposed to such traumatic experiences increases vulner- Whereas high socially anxious undergraduates focused on
ability for psychopathologies, including learning difficul- the eye region, undergraduates low in social anxiety focused
ties, anxiety disorders, depression, or post-traumatic stress more on the mouth region of the presented faces. Taken
disorder (PTSD) [7]. On the other hand, traumatic experi- together, these data point at some possible underlying fac-
ence may also impair cognitive and affective processing and tors that maintain a negative bias.
future psycho-social and cognitive development. Affected Crucially, understanding both how working memory and
cognitive functions include cognitive control functions such interpretation bias changes in traumatized youth may con-
as working memory and inhibitory control, but also atten- tribute to therapeutic approaches. While working memory
tion, emotion processing, or language acquisition [8, 9]. training in healthy children [23] and those with PTSD [24]
Cognitive control, for example, is related to the prefrontal has shown promising results, support for interpretation bias
cortex, which is still maturing throughout the adolescent modification training in youths has been ambivalent [25, 26].
period and, therefore, is especially vulnerable to stress [10]. Yet, undoubtedly, knowledge on areas of need is essential
Furthermore, traumatic experiences could have an impact when trying to establish therapeutic approaches that might
on behavioural and emotional processes, self-regulation, be required when aiming to facilitate integration into their
motivation, aggression, and frustration tolerance [11, 12]. new host countries and educational systems.
Emotion regulation also driven by prefrontal cortical devel- Therefore, to enhance our understanding of the influence
opment includes several domains such as awareness and of war and trauma on working memory and interpretation
understanding emotions, acceptance of emotions, or the bias, we recruited and tested adolescents who had escaped
capacity to control impulsive behaviours, using emotion the war in Syria and who were residing as refugees in Istan-
regulation strategies [13, 14]. bul, Turkey. Youths high in trauma symptoms (HTS) were
Few studies to date have experimentally examined the compared with youths low in trauma symptoms (LTS).
cognitive-affective impact of war-related trauma in youths We had two main hypotheses. (1) With regard to working
[8, 15, 16], while much research has been conducted in memory, we predicted that HTS youths would show worse
(adult) military samples [17]. Recently, Mirabolfathi et al. working memory performance relative to LTS youths, given
[15] examined working memory, which is an essential cog- prior findings in adult [17] and adolescent [8] samples. In
nitive control skill and hypothesized to underlie emotion addition, given previous findings in maltreated women [27]
regulation [18], in Afghani refugee youths. They [15] docu- and youths with PTSD [24], we expected that emotion would
mented poorer working memory capacity in these youths further modulate this performance. (2) With regard to inter-
relative to comparisons. Likewise, earlier, we were able to pretation bias, we predicted that HTS individuals would
document impaired sustained attention during an inhibitory interpret ambiguous faces (surprise) as more negative than
control task (Stroop task) in youths who had experienced LTS individuals, given prior findings of a negative interpre-
parental loss and who were living in a high conflict zone tation bias in soldiers with war-related trauma [21].
[16]. Hecker et al. [8] documented an association between
working memory and PTSD symptoms in adult refugees
from the African continent. Yet, these studies are still few Materials and methods
in number and require replication and extension.
Moreover, to our knowledge, no studies have yet exam- Participants
ined emotional interpretation biases in such youth, although
interpretation bias has been postulated as a maintaining fac- Sixty-two Syrian teenagers (26 girls) who resided in
tor of post-traumatic stress disorder (PTSD) in cognitive Istanbul, Turkey as asylum seekers participated (mean
models [19]. In one classic study, Amir et al. [20] presented age = 11.43 years, age range 10–15 years). We used a median
homographs (words with two meanings, e.g., “arms” mean- split procedure to divide the sample on adolescents reporting
ing either weapons or the part of your body) to people with high symptoms (HTS) vs. low symptoms (LTS) on a trauma
and without PTSD and found delayed response times when scale (Impact of Events Scale Revised) (cf. Table 1). That
individuals with PTSD had to inhibit the negative connota- median value was 36 in the present sample and is close to
tion word (e.g., weapon) to complete the task. Most simi- the clinically suggested cut-off of 33 or more [28]. HTS
lar to the present study, Gebhardt et al. [21] documented (relative to LTS) youths had more symptoms of depression
a negative interpretation when participants had to evaluate (t(57) = 4.06, p < 0.001, d = 1.05), anxiety (t(57) = 2.57,
emotional facial expressions in post-deployment soldiers p < 0.001, d = 0.66), and experience of traumatic events
with trauma (relative to controls). Focusing on the features (IES-R) (t(57) = 10.55, p < 0.001, d = 2.18), but did not

13
European Child & Adolescent Psychiatry

Table 1 Demographic Mean (SD) Low trauma LTS N = 27 High trauma HTS N = 31
information of the study sample
split by group Age, years 11.07 (1.39) 11.70 (1.15)
CDI* 7.34 (4.08) 12.77 (6.09)
STAI-C* 17.57 (6.62) 22.86 (9.11)
IES-R total* 25.48 (7.82) 44.74 (5.96)
IES-R internalizing* 8.41 (3.87) 16.78 (4.17)
IES-R avoidance* 11.02 (4.83) 17.93 (5.08)
IES-R hyperarousal* 5.96 (3.56) 9.99 (3.35)
Resilience 62.02 (18.24) 64.09 (13.08)
War trauma total 10.35 (5.17) 11.85 (5.12)
Demolition of property 2.68 (1.30) 2.59 (1.42)
Violence against self 3.58 (2.16) 4.48 (2.06)
Violence against others 4.10 (2.60) 4.78 (2.76)

CDI child depression inventory, STAI-C state-trait anxiety inventory-child version, IES-R impact of event
scale-revised
*Indicates significant difference between groups at p < 0.05

differ in levels of resilience (t(56) = 0.49, p = 0.62), expe- The Children’s Depression Inventory (CDI) is a paper and
rience of war traumatic events (t(56) = 1.11, p = 0.27), or pencil self-report questionnaire, used to assess childhood
gender distribution (x2(1) = 0.59, p = 0.44). Ethical approval depression. It has 27 items and each of them has three state-
was obtained by the Institutional Review Board of Uskudar ments that are assigned values from 0 (sad sometimes) to 2
University in accordance with the Declaration of Helsinki (sad always), with total score between 0 and 54. Examiner
(1964 and its later amendments). Participants for the current reads the items and the subjects have to choose the sentences
study were recruited through the Hiraeth Refugee Integra- that better describe their feelings in the last two weeks [31].
tion and Support Association. Adolescents consented to par- CDI was validated in Arab population [32]. A cut-off score
ticipate in the study and their parents gave written informed of 19 has been suggested [32, 33], which indicated that none
consent. Children were compensated with small gifts for of the LTS youth but 5 of the HTS youth met this cut-off.
participation. Reliability was Cronbach’s alpha = 0.67.
The Connor–Davidson Resilience Scale (CD-RISC) is a
brief self-rated assessment used to quantify resilience. It has
Measures 25 items with 5 points of responses, from 0 (not true at all)
to 4 (true nearly all the time), with the total score between
Sociodemographic data were collected for each participant 0 to 100 and highest score represents a better resilience [34,
with five questionnaires, to assess the nature of their stress 35]. The scale was validated in Arab population [35]. Reli-
conditions, and each questionnaire had previously been ability was Cronbach’s alpha = 0.87.
translated and validated in Arabic. The State-Trait Anxiety Inventory for children (STAIc) is
The Impact of Events Scale (IES) is a self-report measure a self-report measure used to assess the presence and the
that assesses responses to stressful and traumatic experi- severity of anxiety’s symptoms with 40 items [36]. There are
ences. It has 15 items, of which 7 items evaluate intrusion two subscales: the State Anxiety Scale, which has 20 items
(ideas, images, feelings) and 8 items evaluate avoidance that evaluate the current intensity of feelings with 4 points
(feelings, situations). It analyses how often the subjects from 1 (not at all) to 4 (very much so). The Trait Anxi-
experienced those reactions in the last week. The total dis- ety Scale, instead, has 20 items and it evaluates the stable
tress score is the sum of the 15 items [29]. Developers of the aspects and frequency from 1 (almost never) to 4 (almost
instrument and psychometric investigations have suggested always). The scale was validated in Arab population [37]. A
a cut-off of 33 for a probable diagnosis of PTSD [28]. Given cut-off score of 39 has been suggested [36], which indicated
the present sample’s median split of 36 nicely follows this that none of the LTS youths and only 1 of the HTS youth met
suggested cut-off. Although this cut-off has to be taken with this cut-off. Reliability was Cronbach’s alpha = 0.76.
caution and does not replace a clinical assessment, it was The War-Traumatic Events Checklist [38] has 28 items
useful within the present context. Good discriminant validity and probe exposure to different kinds of war experiences.
for the scale in its Arabic version had been confirmed [30]. It features 3 categories: personal trauma, witnessing trauma
Reliability for the present sample Cronbach’s alpha = 0.73. to others, and seeing demolition of property. Items include,

13
European Child & Adolescent Psychiatry

for example, witnessing the destruction of the own home was shown another time and the subjects had to indicate
by shelling or bulldozing, being beaten or shot by oppo- if it was in the same locations as before [27] (for a sample
nent forces, being used as a human shield, or having had to trial, see Fig. 1a). The participants had to press “ = ” when
watch the opponent forces destroying land, farms, killing or the memory probes were in the start location, and they were
hurting people. Response options include “Yes” and “No”, defined as ‘match’ trials, while the other conditions were
and are added up to get a total score. However, because the the ‘miss’ trials and participants had to press “ ≠ ” when the
original questionnaire by El-Khodary and Samara [38] was probe did not match the cue’s location (with the index and
designed to examine trauma in Palestine youth in Gaza, we middle finger of the right hand). Furthermore, to manipulate
had to alter the wording in items that specifically referred the level of difficulty, there were three kinds of ‘miss’ trials
to the IDF (Israeli Defence Forces) of their questionnaire depending on distance between memory cue and memory
to ‘foreign army’ as IDF was not applicable in the present probe: 15° (miss1), 20° (miss2), and 25° (miss3) away from
context. Cronbach’s alpha = 0.82. the cue [27]. The participants had the goal to respond as
quickly as possible, and reaction times (in ms) and accuracy
Experimental tasks (% correct) were recorded.

Spatial emotional match to sample (SEMTOS)


Surprise faces task
The SEMTOS task has previously been used to assess vis-
The surprise face task was based on the one by Neta et al.
ual working memory in the presence of distracting emo-
[42] and included 18 gender-balanced (nine males and nine
tional stimuli in adult women with childhood abuse history
females) facial expressions, which were again obtained
[27]. Based on the effect sizes of this previous work [27],
from the NimStim standardized facial expression stimulus
a priori power calculations revealed that only 14 partici-
set [39], also using ovals to avoid the display of women’s
pants would be necessary to obtain a statistically significant
hair. These expressions represented three of the six uni-
within-between factor group interaction. The task measures
versal and basic emotions: anger, happiness, and surprise
the influence of incidental emotional stimuli (happy, angry,
[42]. The three facial expressions were randomly shown
and neutral facial expressions) on remembering the accurate
in three blocks of 54 trials, with the same numbers of the
location of a white oval and a black screen. The task has
three emotional faces, for 1000 ms which was followed by a
120 trials divided into four blocks of 30 trials each. Within
blank screen for a variable ITI (intertrial interval) of 1500,
each block, the face of each actor was shown only once.
2000, or 2500 ms. This jittering was done to avoid anticipa-
The emotions are given via faces of 10 individuals perform-
tory responses and enforce task compliance. In this task,
ing 24 expressions (20 female actors and 24 male actors).
participants were asked to recognize and label each facial
At the time of planning the study, we were unaware of any
emotion as a negative or a positive emotion (for a sample
face database that consisted of faces that displayed faces of
trial see Fig. 1b). To this goal, they were asked to press one
a Middle Eastern ethnic background. We therefore selected
of two response buttons on the keyboard, one for positive
faces from the most used and cited databases in the litera-
and one for negative expressions, respectively. The order
ture, namely, from the NimStim face stimulus database [39]
of response assignment was counterbalanced across partici-
for the trials proper and from the Radboud Face Database
pants. A priori power calculations based on the effect size
[40] for the practice trials. Importantly, we used the same
reported by Neta et al. [42] indicated that we would need a
oval faces as in the previous study as here it was also cultur-
minimum of 56 participants across two groups to obtain a
ally relevant; namely, because women in Muslim cultures are
power of 0.95 for a within-between factor interaction on the
not allowed to show their hair, our ovals permitted seeing
emotional interpretation task.
both male and female faces without hair. Moreover, we used
black and white versions of the faces to limit visual salience
and reduce a potential influence of ethnicity (skin bright- Procedure
ness). However, like most research in anxiety, we opted to
present emotional faces of adults rather than children or ado- Participants completed two computer-based tasks assess-
lescents for easier comparison [41]. The task started with ing emotional working memory and affect recognition and
a yellow cross for 1000 ms followed by a memory cue as filled out questionnaires. Tasks and questionnaires were
face with one of the three emotions, in 1 of 24 locations conducted by a trained researcher accompanied by a native
for 400 ms. In the next 1000 ms, a blue or red colour probe Arabic speaker who was trained on the aims and ethical pro-
appeared and participants had to select the colour, “w” or cedures of the study and helped with translation of the tasks
“x” on the keyboard (with the index and middle finger of and study aims as well as instructions. The tasks and forms
the left hand). In the last 1500 ms, the face, memory probe, were administered in a silent room provided by the Hiraeth

13
European Child & Adolescent Psychiatry

Fig. 1 Sample trial for (a) the SEMTOS task, in which a miss1 trial is presented (i.e., a slight change between cue and target). b Sample trial of
the emotional interpretation task. Please note that the ITI in the task was variable and differed from trial to trial to avoid anticipatory responding

Association and lasted approximately 60–90 min for each the different allocations were calculated. Effect sizes are
participant. reported as Cohen’s d or partial eta squared, as appropriate.
Alpha was set at p = 0.05, two-tailed.
Statistical analysis

For the SEMTOS task, a repeated-measures ANOVA was Results


conducted with the within factors of emotion and location
and the between-subjects factor of trauma group (HTS vs. SEMTOS task
LTS) resulting in a 3 (emotion: angry, happy, neutral) by 4
(location: match, miss1, miss2, miss3) by 2 (group: high Accuracy
vs. low trauma symptoms) design. This rmANOVA was run
separately for accuracy (percent correct) and reaction time The analysis of the accuracy data revealed a significant
(only including correct trials). Follow-up tests were con- location-by-group interaction (F(3,165) = 3.99, p = 0.009,
ducted with paired t test (within group) or ANOVA (between ηp2 = 0.07). In addition, the main effect of location was
group) and Bonferroni corrected for multiple comparisons also significant (F(3,165) = 8.93, p < 0.001, ηp2 = 0.14).
(corrected alpha two-tailed p = 0.008). By contrast, the sim- No other effects were significant. Follow-up of the interest-
plicity and straightforward hypothesis of the surprise face ing location-by-group interaction revealed that LTS youths
task allowed us to simply compare the percentage of surprise were statistically significantly better at identifying match
faces allocated to positive vs. negative emotions by means trials relative to HTS youth (F(1,54) = 8.06, p = 0.006,
of a univariate ANOVA with the difference score as depend- d = 0.76), but did not differ from each other on any of the
ent and the group as independent variable, run separately miss trials. When following-up within group, LTS youth
for RT and allocation. For RT, reaction time differences to had significantly better accuracy on match trials vs. miss1

13
European Child & Adolescent Psychiatry

trials (t(29) = 5.12, p < 0.001), match trials vs. miss2 trials
(t(29) = 4.09, p < 0.001), while accuracy was better on miss3
trials vs. miss1 trials (t(29) = − 4.79, p < 0.001) and miss2
trials (t(29) = − 4.14, p < 0.001). Comparisons between
match and miss3 and miss1 vs. miss2 trials did not survive
Bonferroni correction. By contrast, none of the compari-
sons within the HTS group were significantly different from
one another (all p > 0.07) (Fig. 2a). The follow-up test on
the main effect of the location indicated higher accuracy
on match vs. miss1 trials (t(65) = 3.12, p = 0.003), but also
larger accuracy as the miss trials became more dissimilar
from one another, i.e., better performance on miss2 vs.
miss1trials (t(65) = 4.46, p < 0.001), better performance on
miss3 vs. miss1trials (t(65) = 7.34, p < . 001), and, in turn,
better performance on miss3 vs. miss2 trials (t(65) = 4.36,
p < 0.001).

Reaction time

The ANOVA on reaction time only revealed a significant


effect of location (F(3,162) = 2.88, p = 0.038, ηp2 = 0.05)
(Fig. 2b). This effect indicated faster performance on match
trials relative to (a) miss1 trials (t(61) = 3.01, p = 0.003), (b)
miss2 trials (t(60) = 3.67, p = 0.001), and (c) miss3 trials
(t(62) = 3.35, p = 0.001). No other effects were significant.

Surprise faces task

Fig. 2 a The difference in % accuracy (from 0.00–1.00, 1.00 = 100%) Emotion allocation
between the HTS (dotted line) and LTS (dashed line) youths for the
four different locations in the SEMTOS task. b The corresponding
reaction times in ms showing no evidence for any influence of reac- As predicted, adolescents likened surprise faces more to neg-
tion time on the accuracy finding. Error bars denote S.E.M. ative emotions (60.49%) than to positive emotions (39.55%),

Fig. 3 a The percent to which surprise faces were allocated to a positive or a negative category. b The significant group difference in reaction
time, with a lower value indicating faster allocation to the positive category. Error bars denote S.E.M

13
European Child & Adolescent Psychiatry

but this did not differ between groups (Fig. 3a). When these symptoms. This association indicated that the more war trau-
two values were directly compared against one another, matic events youths had experienced, the more they allo-
the difference was statistically significant (t(61) = − 3.78, cated surprise faces to the negative category (r(51) = 0.281,
p < 0.001). Also as expected, happy faces were generally p = 0.046)(Fig. 4).
evaluated as positive (83.70%) and angry faces were pre-
dominantly evaluated as negative (85.20%). Groups did not
differ statistically in their frequency of allocation across the Discussion
three emotions (cf. Table 2).
Because of the tremendous impact of war experience on the
Reaction times cognitive-affective skills in adolescents, this investigation
specifically aimed to examine the influence of emotion on
Although groups did not differ in the frequency of their allo- working memory as well as the interpretation of emotion-
cation, an interesting effect of group emerged when exam- ally ambiguous faces. Two main findings emerged. First,
ining the speed with which this was done. For the RT, the HTS (vs. LTS) were ~ 20% worse at remembering the correct
difference score of how long it took participants to qualify location in a working memory task. This was not qualified
a respective emotion as positive or negative was calculated. by emotion. Second, although both groups also interpreted
A significant effect of group (F(1,51) = 6.53, p = 0.014, surprise faces ~ 20% more negative than positive, a statis-
ηp2 = 0.11) indicated that LTS youth were equally fast when tically significant group difference indicated that HTS vs.
deciding whether surprise faces were positive or negative LTS were faster when allocating surprise faces to positive
(mean = − 4.31 ms, SD = 68.89), but HTS youth were faster (relative to negative) emotions.
allocating a positive judgement to surprise faces relative to With regard to the working memory task, the findings
negative judgements (mean = − 57.42 ms, SD = 82.32 ms) are in-line with the expectations and indicated worse visual
(Fig. 3b). working memory accuracy in HTS relative to LTS youth.

Exploratory analyses of comorbid psychopathology


and dimensional associations of psychopathology
with task performance

To examine to what extent individual differences and


comorbid symptoms may have driven the findings, the
main rmANOVA was repeated including CDI, STAI, and
resilience as covariates (separately). These analyses did
not change the significant interaction between location and
group and also indicated no main effects of these variables
thus suggesting little influence of depression, anxiety, or
resilience on the SEMTOS task. Likewise, re-running the
ANOVA for the RT difference score in allocating surprise
faces to a happy or a negative category did not change the
significant difference when adding resilience or anxiety, but
reduced the significant effect when adding depression to
p = 0.07 thus showing a slight effect. The direct correlation
(Pearson’s Product–Moment correlation) between these two
variables was trending (r(54) = -0.25, p = 0.065). Fig. 4 Scatterplot illustrates the positive association between war
traumatic events (y-axis) and allocation of surprise faces to the nega-
Finally, an interesting positive correlation emerged tive category (x-axis) for the low trauma symptom group (white cir-
between performance on the interpretation task and trauma cles, LTS) and high trauma symptom group (black circles, HTS)

Table 2 Allocation ratios for the three different types of emotions (happy, angry, and neutral faces) and whether they were perceived to be posi-
tive or negative
Mean(SD) Happy—positive Happy—negative Angry—negative Angry—positive Surprise—positive Surprise—negative

Low trauma (LTS) 85.57% (12.11%) 15.54% (11.43) 85.05% (13.29%) 14.96% (3.28%) 40.29% (24.25%) 59.71% (24.25%)
High trauma (HTS) 81.27% (20.88%) 13.92% (13.27%) 85.39% (11.96%) 14.7% (11.73%) 38.58% (24.71%) 61.51% (24.80%)

13
European Child & Adolescent Psychiatry

Interestingly, while this finding was cemented by a strong play a role in maintaining PTSD [19]. In our simple ambi-
effect size and a 20% difference in accuracy between groups guity task, both groups interpreted surprise faces 20% more
on the match trials, no group differences emerged on any of as negative than positive (or 10% more relative to chance).
the miss trials. Moreover, no specific modulation of emo- While this bias confirmed our hypothesis, this finding is a)
tion was found in the previous study of women with expe- smaller than expected and b) did not differ between groups.
rience of interpersonal violence using the same task [27]. Yet, a group difference emerged in the reaction times when
However, direct comparison between the studies is difficult making these judgements. While LTS youth showed no dif-
given fundamental differences across samples (i.e., youths ference, HTS youth responded faster once they had decided
vs. adults, war-related trauma vs. sexual abuse trauma, and that a surprise face appeared positive (vs. negative) to them.
Western sample vs. Middle Eastern sample). Whereas the Generally speaking, the main effect is consistent with the
general response pattern of the task was replicated, indicat- study by Gebhardt et al. [21], who found a negative bias
ing increasing accuracy with increasing spatial distance from in soldiers post-deployment when evaluating ambigu-
the target stimulus, the overall performance of both groups ous faces. However, the differential reaction time effect
was surprisingly poor. Of note, although the data indicated deserves further scrutiny and any conclusions are specula-
that, within the present sample, there did not appear to have tive at this stage. One possible explanation may be linked
been a speed-accuracy trade-off (cf. Fig. 2a vs. Fig. 2b), to how trauma affects basic low-level visual processes. For
interestingly, relative to the previous study [27], youths in example, in a sample of healthy participants, Neta et al. [42]
the present study were on average 100–150 ms faster than had documented a differential influence of spatial frequency
the traumatized adult women from that earlier work, but on the perception of surprise faces depending on whether
also showed much worse accuracy. Three sets of findings people had a positive or negative interpretation bias. In addi-
suggest that even though there may not have been a speed- tion, during an fMRI study, persons with PTSD had lower
accuracy trade-off within the present sample, there may have neuronal activity in the ventral visual stream when compared
been one between the two studies. This line of reasoning is to controls [47]. Taken together, these findings might sup-
based on (1) the fact that typically developing youths are port the notion of altered visual processes after trauma that
normally slower than adults on reaction time tasks [43], (2) may contribute to the interpretation bias and a gearing of the
adolescents relative to children and adults have been shown attention system toward threat.
to exhibit altered speed-accuracy trade-off behaviour in the In terms of clinical relevance, understanding what
presence of emotional stimuli [44] and (3) developmental cognitive-affective factors may contribute to maintaining
evidence of protracted lateral prefrontal cortical develop- mental health problems including anxiety, depression, and
ment during adolescence [45], which appears essential for post-traumatic stress symptoms will aid in improving ther-
accurate performance within speed-accuracy trade-off set- apeutic approaches. Within the present context, research-
tings [46]. Despite this low performance, the presence of ers have advocated the use of working memory training as
a statistically significant interaction between the groups well as interpretation bias training in people with affective
and memory condition suggests that a floor effect may not problems [24, 48]. Whereas working memory training in
play an important role. In addition, a finding of severely healthy children [23] as well as those with PTSD [24] has
decreased visual working memory accuracy fits perfectly shown some promising results, support for interpretation
with a recent meta-analysis of cognitive performance in trau- bias modification training in youths is limited [25, 26, 49].
matized adults relying mostly on military and interpersonal The current data illustrate the severe impact of war trauma
violence samples. In these adults (mean age ~ 44 years), an on both of these abilities in adolescents and underscore the
average effect size of d = 0.5 was documented for poorer importance of further research into this area. One inter-
performance of the traumatized groups [17]. The main find- esting mechanistic question is why studies of interpreta-
ing also fits with the few existing studies in similar cohorts. tion bias training do not appear to exert the desired effect
Ainamani et al. [8] have reported an association between despite clear presence of negative interpretation biases
PTSD symptoms severity and working memory capacity in the present study (among many others). The current
using a classic neuropsychological task in African refu- finding of speeded response times (in absence of differ-
gees (mean age ~ 31 years of age), while a recent study in ences in overall allocation rates) in the HTS group vs. the
Afghani refugee adolescents with a slightly higher mean age LTS group may suggest a more downstream effect of early
(~ 15 years) also indicated poor working memory capacity in visual processing. In other words, HTS youths interpreted
youth with high PTSD symptoms relative to a control group ambiguous faces more quickly. This would be consist-
[15]. These findings further cement the notion of impaired ent with one reported finding that attention bias training
working memory after war-related trauma. improves interpretation bias but not vice versa [50] and
With regard to the interpretation bias, previous research- the importance of early threat perception in such biases
ers have postulated how negative interpretation bias may [41]. Unfortunately, the potential influence of depressive

13
European Child & Adolescent Psychiatry

symptoms on the interpretation bias finding has to be Compliance with ethical standards
interpreted with caution, because (1) it did not reliably
influence the outcome nor did it convincingly not show any Conflict of interest None of the authors have a conflict of interest to
influence and (2) the reliability for this questionnaire in the declare.
present sample was low. However, because interpretation Ethical approval Ethical approval granted by the Institutional Review
biases are hypothesized to be more associated with depres- Board of Uskudar University, Istanbul, Turkey.
sive symptoms and later stages of cognitive processing, the
trending effects appear consistent with the literature [51].
This also highlights some notable limitations. While
desirable, one problem was that no other comparison group
References
was collected. The main problem was that Turkish and Syr-
ian cultures have different alphabets and speak different 1. United Nations High Commissioner for Refugees (UNHCR)
languages, which would have required to compare ques- (2019) Global trends: forced displacement in 2018. In: UNHCR,
tionnaires across different languages (with likely different Geneva, Switzerland
2. Baban F, Ilcan S, Rygiel K (2016) Syrian refugees in Turkey: path-
validations etc.). Second, relatedly, although both tasks
ways to precarity, differential inclusion, and negotiated citizenship
were validated and have been used in prior work [27, 42], rights. J Ethn Migr Stud 43:41–57
direct comparison of performance to prior work is difficult 3. Alpak G, Unal A, Bulbul F, Sagaltici E, Bez Y, Altindag A,
because of differences in language, culture, and age although Dalkilic A, Savas HA (2015) Post-traumatic stress disorder among
Syrian refugees in Turkey: a cross-sectional study. Int J Psychiatry
direct comparison with another group would have been ben-
Clin Pract 19:45–50
eficial in light of the overall very poor working memory 4. Foster RP (2001) When immigration is trauma: guidelines for the
performance. Third, the selection of questionnaires was individual and family clinician. Am J Orthopsychiatry 71:153–170
partly driven by the availability of validated versions in the 5. Berman H (2001) Children and war: current understandings and
future directions. Public Health Nurs 18:243–252
Arabic language. Given the presently rather limited avail-
6. Soykoek S, Mall V, Nehring I, Henningsen P, Aberl S (2017) Post-
ability of validated mental health questionnaires in Arabic, traumatic stress disorder in Syrian children of a German refugee
future development and cross-cultural validation of scales camp. Lancet 389:903–904
are desired. Yet, an important strength of the present study 7. Fazel M, Reed RV, Panter-Brick C, Stein A (2012) Mental health
of displaced and refugee children resettled in high-income coun-
relative to earlier work [8, 15] was that we assessed both
tries: risk and protective factors. Lancet 379:266–282
skills thus facilitating cross-task comparison, which allows 8. Ainamani HE, Elbert T, Olema DK, Hecker T (2017) PTSD
for a more comprehensive clinical picture. Fourth, unfortu- symptom severity relates to cognitive and psycho-social dys-
nately, information about IQ and general cognitive abilities functioning—a study with Congolese refugees in Uganda. Eur J
Psychotraumatol 8:1283086
and school attendance prior to becoming refugees was not
9. Kaplan I, Stolk Y, Valibhoy M, Tucker A, Baker J (2016) Cogni-
available. However, we would like to point out that unlike tive assessment of refugee children: effects of trauma and new
in the traditional clinical studies where the two groups are language acquisition. Transcult Psychiatry 53:81–109
recruited via separate pathways (e.g., via outpatient clinic 10. McEwen BS, Morrison JH (2013) The brain on stress: vulner-
ability and plasticity of the prefrontal cortex over the life course.
and school) in the present study, we collected as many youth
Neuron 79:16–29
from the same organization and then applied the median split 11. Hecker T, Fetz S, Ainamani H, Elbert T (2015) The cycle of vio-
afterwards thus somewhat mitigating this problem of ‘two lence: associations between exposure to violence, trauma-related
different groups’. symptoms and aggression-findings from Congolese refugees in
Uganda. J Trauma Stress 28:448–455
In summary, the present data highlight problems in work-
12. Endo T, Sugiyama T, Someya T (2006) Attention-deficit/hyperac-
ing memory and negative interpretation bias in youths with tivity disorder and dissociative disorder among abused children.
high trauma symptoms. Any attempt at integrating refugee Psychiatry Clin Neurosci 60:434–438
youths in host countries may need to take into consideration 13. Nickerson A, Bryant RA, Schnyder U, Schick M, Mueller J,
N M, (2014) Emotion dysregulation mediates the relationship
not only youths’ traumatic experiences, but also account for
between trauma exposure, post-migration living difficulties and
possible cognitive-affective perturbations that may require psychological outcomes in traumatized refugees. J Affect Disord
extra attention. However, given the continued scarcity in the 173:185–192
field, more definitive knowledge is required. 14. Tull MT, Barrett HM, McMillan ES, Roemer L (2007) A pre-
liminary investigation of the relationship between emotion regula-
tion difficulties and posttraumatic stress symptoms. Behav Ther
Funding This work was supported by a 2–4 year grant (01J05415)
38:303–313
from the Special Research Fund (Bijzonder Onderzoeksfonds, BOF)
15. Mirabolfathi V, Schweizer S, Moradi A, Jobson L (2020) Affec-
at Ghent University to SCM.
tive working memory capacity in refugee adolescents. Psychol
Trauma. https://doi.org/10.1037/tra0000552
Data availability Both tasks are available from the authors. Due to the 16. Mueller SC, Baudoncq R, De Schryver M (2015) The effect of
sensitive nature of the data, aggregated data were only available upon parental loss on cognitive and affective interference in adolescent
reasonable request. boys from a post-conflict region. J Adolesc 42:11–19

13
European Child & Adolescent Psychiatry

17. Scott JC, Matt GE, Wrocklage KM, Crnich C, Jordan J, Southwick 35. Alqudah AF (2013) Resiliency levels among Iraqi refugees in Jor-
SM, Krystal JH, Schweinsburg BC (2015) A quantitative meta- dan and its relation to some demographic variables. Int J Psychol
analysis of neurocognitive functioning in posttraumatic stress Stud 5:50–59
disorder. Psychol Bull 141:105–140 36. Spielberger CD (1973) Manual for the State-Trait Anxiety Inven-
18. Pruessner L, Barnow S, Holt DV, Joormann J, Schulze K (2020) A tory for Children. Consulting Psychologists Press, Palo Alto,
cognitive control framework for understanding emotion regulation California
flexibility. Emotion 20:21–29 37. Abdel-Khalek AM (1989) The development and validation
19. Ehlers A, Clark DM (2000) A cognitive model of posttraumatic of an Arabic form of STAI: Egyptian results. Pers Individ Dif
stress disorder. Behav Res Ther 38:319–345 10:277–285
20. Amir N, Coles ME, Foa EB (2002) Automatic and strategic activa- 38. El-Khodary B, Samara M (2020) The relationship between multi-
tion and inhibition of threat-relevant information in posttraumatic ple exposures to violence and war trauma, and mental health and
stress disorder. Cognit Ther Res 26:645–655 behavioral problems among Palestinian children and adolescents.
21. Gebhardt C, Alliger-Horn C, Mitte K, Glaesmer H (2017) All- Eur Child Adolesc Psychiatry 29:719–731
or-nothing thinking: the processing of emotional expressions in 39. Tottenham N, Tanaka JW, Leon AC, McCarry T, Nurse M, Hare
traumatized post-deployment soldiers. J Anxiety Disord 47:69–74 TA et al (2009) NimStim set of facial expressions: judgments from
22. Gutiérrez-García A, Calvo MG, Eysenck MW (2018) Social anxi- untrained research participants. Psychiatry Res 168:242–249
ety and detection of facial untrustworthiness: spatio-temporal ocu- 40. Langner O, Dotsch R, Bijlstra G, Wigboldus DHJ, Hawk ST, van
lomotor profiles. Psychiatry Res 262:55–62 Knippenberg A (2010) Presentation and validation of the Radboud
23. Astle DE, Barnes JJ, Baker K, Colclough GL, Woolrich MW Faces database. Cogn Emot 24:1377–1388
(2015) Cognitive training enhances intrinsic brain connectivity 41. Bar-Haim Y, Lamy D, Pergamin L, Bakermans-Kranenburg MJ,
in childhood. J Neurosci 35:6277–6283 van IJzendoorn MH (2007) Threat-related attentional bias in anx-
24. Schweizer S, Samimi Z, Hasani J, Moradi A, Mirdoraghi F, ious and nonanxious individuals: a meta-analytic study. Psychol
Khaleghi M (2017) Improving cognitive control in adolescents Bull 133:1–24
with post-traumatic stress disorder (PTSD). Behav Res Ther 42. Neta M, Norris CJ, Whalen PJ (2009) Corrugator muscle
93:88–94 responses are associated with individual differences in positivity-
25. Chan SW, Lau JY, Reynolds SA (2015) Is cognitive bias modifi- negativity bias. Emotion 9:640–648
cation training truly beneficial for adolescents? J Child Psychol 43. Kail R (1991) Developmental change in speed of processing dur-
Psychiatry 56:1239–1248 ing childhood and adolescence. Psychol Bull 109:490–501
26. Fu X, Du Y, Au S, Lau JY (2013) Reducing negative interpreta- 44. Tottenham N, Hare TA, Casey BJ (2011) Behavioral assessment of
tions in adolescents with anxiety disorders: a preliminary study emotion discrimination, emotion regulation, and cognitive control
investigating the effects of a single session of cognitive bias modi- in childhood, adolescence, and adulthood. Front Psychol. https://
fication training. Dev Cogn Neurosci 4:29–37 doi.org/10.3389/fpsyg.2011.00039
27. Cromheeke S, Herpoel L-A, Mueller SC (2014) Childhood abuse 45. Dahl RE (2004) Adolescent brain development: a period of vul-
is related to working memory impairment for positive emotion in nerabilities and opportunities. Ann N Y Acad Sci 1021:1–22
female university students. Child Maltreat 19:38–48 46. Wenzlaff H, Bauer M, Maess B, Heekeren HR (2011) Neural
28. Creamer M, Bell R, Failla S (2003) Psychometric properties of the characterization of the speed-accuracy tradeoff in a perceptual
impact of event scale—revised. Behav Res Ther 41:1489–1496 decision-making task. J Neurosci 31:1254–1266
29. Weiss DS, Marmar CR (1997) The Impact of Event Scale-Revised. 47. Mueller-Pfeiffer C, Schick M, Schulte-Vels T, O’Gorman R,
In: Wilson JP, Keane TM (eds) Assessing Psychological Trauma Michels L, Martin-Soelch C, Blair JR, Rufer M, Schnyder U,
and PTSD: A Practitioner’s Handbook. Guilford Press, New York, Zeffiro T, Hasler G (2013) Atypical visual processing in post-
USA, pp 399–411 traumatic stress disorder. Neuroimage Clin 3:531–538
30. Davey CG, Heard R, Lennings C (2014) Development of the Ara- 48. Krebs G, Pile V, Grant S, Degli Esposti M, Montgomery P, Lau
bic versions of the impact of events scale-revised and the post- JYF (2018) Research review: cognitive bias modification of inter-
traumatic growth inventory to assess trauma and growth in Middle pretations in youth and its effect on anxiety: a meta-analysis. J
Eastern refugees in Australia. Clin Psychol 19:131–139 Child Psychol Psychiatry 59:831–844
31. Smucker MR, Craighead WE, Craighead LW, Green BJ (1986) 49. LeMoult J, Colich N, Joormann J, Singh MK, Eggleston C, Gotlib
Normative and reliability data for the children’s depression inven- IH (2018) Interpretation bias training in depressed adolescents:
tory. J Abnorm Child Psychol 14:25–39 near- and far-transfer effects. J Abnorm Child Psychol 46:159–167
32. Ghareeb AG, Beshai JA (1989) Arabic version of the children’s 50. Bowler JO, Hoppitt L, Illingworth J, Dalgleish T, Ononaiye M,
depression inventory: reliability and validity. J Clin Child Psychol Perez-Olivas G, Mackintosh B (2017) Asymmetrical transfer
18:323–326 effects of cognitive bias modification: modifying attention to
33. Kovacs M (1992) Children’s Depression Inventory. Multi-Health threat influences interpretation of emotional ambiguity, but not
Systems, New York vice versa. J Behav Ther Exp Psychiatry 54:239–246
34. Connor KM, Davidson JRT (2003) Development of a new resil- 51. Joormann J (2004) Attentional bias in dysphoria: the role of inhib-
ience scale: the Connor-Davidson Resilience Scale (CD-RISC). itory processes. Cogn Emot 18:125–147
Depress Anxiety 18:76–82

13

You might also like