Introduction
Depression is one of the most common psychiatric disorders in childhood and adolescence (Costello et al.
2003; Lewinsohn et al.
1993) with up to 20% of young people having experienced at least one episode of major depression (MD) by the end of adolescence (Thapar et al.
2012). Early-onset MD is associated with adverse outcomes later in life such as educational underachievement (Fergusson and Woodward
2002), impairments in psychosocial functioning (Hammen et al.
2008), and reduced life satisfaction (Lewinsohn et al.
2003). In addition, early-onset MD often follows a recurrent course (e.g., Lewinsohn et al.
1999; Weissman et al.
1999), which further contributes to the negative consequences of the disorder (Wilson et al.
2015; Hammen et al.
2008).
Cognitive theories of depression propose that cognitive vulnerabilities such as cognitive biases play a crucial role in the development and maintenance of depressive disorders (e.g., A. T. Beck and Haigh
2014; Disner et al.
2011). Negative cognitive biases are tendencies to preferentially process negative compared to positive or neutral information and can be found on various levels of information processing, including attention, interpretation, and memory (Everaert et al.
2012; LeMoult and Gotlib
2019). Negative interpretation biases, in particular, refer to tendencies to create more negative and fewer positive meanings to explain ambiguous emotional information (Everaert et al.
2017). For example, a situation in which one is giving a speech in front of a group and people are laughing could be interpreted negatively in terms of people laughing at one or positively in terms of people appreciating one’s jokes. In adults, the association between negative interpretation biases and depression has received particularly substantial empirical support (see Everaert et al.
2017, for a comprehensive meta-analysis).
However, results obtained from studies on adults with MD cannot be directly transferred onto depressed youth (Lakdawalla et al.
2007), as major cognitive and affective development is ongoing during childhood and adolescence (Blakemore and Choudhury
2006; Steinberg
2005). Therefore, cognitive vulnerabilities might either play a smaller role in youth than adult depression as cognitive patterns might not have evolved into stable, trait-like “cognitive styles” yet at this younger age (e.g., Lakdawalla et al.
2007). Alternatively, young people might be particulartly susceptible to negative cues in ambiguous emotional information due to brain maturation and hormonal changes associated with an enhanced emotional sensitivity (see e.g., Paus et al.
2008), resulting in more pronounced negative cognitive biases. Considering the particularly detrimental consequences of early-onset MD, understanding the mechanisms that are involved in the development and maintenance of the disorder at this early age is crucial in order to improve prevention and early intervention (Loechner et al.
2018; Weisz et al.
2006).
Still, research on the association of interpretation biases and depression in children and adolescents is rather scarce (Platt et al.
2017). Some studies have reported correlations between interpretation bias scores and depressive symptoms in unselected adolescent samples (e.g., Klein et al.
2018; Orchard et al.
2016a; Smith et al.
2018) as well as samples with elevated symptoms of depression (de Voogd et al.
2017), but only two studies have compared interpretation biases in clinically depressed versus healthy youth. As part of a validity check in their study of an intervention for clinically depressed adolescents and young adults (14–21 years old), Micco et al. (
2014) compared the depressed group’s baseline interpretation bias (assessed with the experimental Ambiguous Scenarios Task, AST; Mathews and Mackintosh
2000) with that of a healthy control group and found the depressed adolescents and young adults to show a more negative interpretation bias. However, as the comparison of depressed and non-depressed groups was not the main aim of the study, this result is presented only briefly and its importance is not discussed. Orchard et al. (
2016b) on the other hand, used the Ambiguous Scenarios Test for Depression in Adolescents, a questionnaire measure they had previously adapted and validated (Orchard et al.
2016a), to investigate interpretation biases in 12–18-year-old adolescents. They found a more negative interpretation bias in adolescents with a diagnosis of MD not only compared to healthy adolescents from the community but also to clinically-referred non-depressed youth and adolescents from the community with elevated depressive symptoms.
To date, no study has focused on comparing interpretation biases in depressed and non-depressed youth using experimental tasks. These do not rely on participants’ awareness of their depressive cognitions and are less prone to distortions due to demand characteristics (i.e., participants matching their responses to the experimenter’s presumed expectation), response biases (i.e., participants endorsing negative responses irrespective of the content corresponding to their interpretation or not), and deliberate response strategies (i.e., participants generating their responses based on a voluntary strategy instead of their immediate reaction to the ambiguous information) that are typical for self-report measures (e.g., Gotlib and Joormann
2010; Hirsch et al.
2016). Thus, experimental tasks enable a more objective assessment of cognitive processes and allow more automatic and unconscious processes that operate outside a person’s awareness to be captured. Therefore, the first aim of the present study was to investigate interpretation biases in youth depression using age-adapted experimental approaches to assess interpretation biases in children and adolescents with MD.
We administered the AST (Mathews and Mackintosh
2000) in which participants read several self-referent ambiguous scenarios and are then presented with different interpretations of each scenario. Interpretation bias is indexed by the difference between the endorsement of negative and positive interpretations (de Voogd et al.
2017). In addition, the Scrambled Sentences Task (SST; Wenzlaff and Bates
1998), which was specifically developed to assess interpretation biases in depressive disorders, was applied. In this task, participants form sentences out of arrays of words which can be either positive or negative. The proportion of negatively resolved sentences indicates the interpretation bias. Applying two experimental measures of interpretation bias allows the examination of different aspects of interpretation, with the AST presumably measuring a more conscious and explicit aspect and the SST capturing a more automatic and implicit aspect (Sfärlea et al.
2019). Both tasks have already been used in adolescent samples (e.g., de Voogd et al.
2017; Burnett Heyes et al.
2017) where they demonstrated at least acceptable reliability (Micco et al.
2014; Sfärlea et al.
2019).
Children and adolescents with MD were compared to two groups of non-depressed children and adolescents that varied in their risk for depression: children of parents with a history of depression, who are known to have an increased risk for MD themselves (e.g., Weissman et al.
2006) and children of parents with no history of depression or any other mental disorder, who have a low risk for depressive disorders. This allowed us to pursue the second aim of our study: to determine the extent to which interpretation biases are more pronounced in currently depressed youth compared to at-risk youth (that have been found to be characterized by more negative interpretation biases than youth at low risk for depression; Dearing and Gotlib
2009; Sfärlea et al.
2019). While negative interpretation biases in children and adolescents at high risk for depression indicate that these biases might be cognitive vulnerabilities or risk factors contributing to the development of depression (as suggested by theoretical models, e.g., Disner et al.
2011), even more pronounced interpretation biases in currently depressed children and adolescents indicate that these biases might be exacerbated as a consequence of depressive symptomatology. No study to date has directly compared interpretation biases in depressed, high-, and low-risk youth. One study that investigated memory biases in children and adolescents with MD, children and adolescents whose mothers were affected by MD, and children and adolescents without familial history of MD (Fattahi Asl et al.
2015) found negative memory biases in both depressed as well as at-risk youth compared to low-risk youth. However, the negative memory biases were more pronounced in currently depressed children and adolescents than in the at-risk group.
In order to be able to compare currently depressed youth to at-risk youth we focused on children and adolescents aged 9–14 years. Children younger than 9 years were not included due to concerns about their ability to understand and perform the tasks. Adolescents older than 14 years were not included since the incidence of depression in children of parents with a history of depression increases substantially after that age (e.g., Weissman et al.
2006), and investigating older children of depressed parents that had not yet suffered from an episode of MD might result in examining a particularly resilient and therefore non-representative high-risk sample.
With respect to the first aim of the study, we expected to find more negative interpretation biases in children and adolescents with MD in comparison to healthy children and adolescents (both high- and low-risk youth), based on theoretical predictions (e.g., Disner et al.
2011) and previous findings (Orchard et al.
2016b; Micco et al.
2014). Regarding the second aim, we expected negative interpretation biases to be to some extent present in youth at high risk for depression compared to youth at low risk for depression (corresponding to our previous results, Sfärlea et al.
2019; as well as Dearing and Gotlib
2009; Goodman and Gotlib
1999), but to be more pronounced in depressed versus high-risk youth (as found for memory biases; Fattahi Asl et al.
2015).
Discussion
The present study investigated the role of interpretation biases in youth depression. Two experimental tasks capturing different aspects of interpretation were used to assess interpretation biases in three groups of children and adolescents: currently depressed children and adolescents (MD group), children and adolescents at high risk for depression due to having a parent with a history of depression (HR group), and children and adolescents with a low risk for depression (LR group). Both tasks revealed a more negative interpretation bias in children and adolescents with MD compared to both groups of healthy youth and strong correlations between bias scores and depression and anxiety symptoms (collapsed across groups), while only one task (SST) revealed a more negative interpretation bias in youth at risk for depression compared to low-risk youth (see also Sfärlea et al.
2019).
The first aim of the present study was to test the assumption that children and adolescents with MD show more negative interpretation biases compared to healthy youth. As expected, we found the MD group to draw more negative interpretations of ambiguous scenarios (AST) as well as sentences (SST), i.e., to show more negative interpretation biases, than the two groups of healthy children and adolescents. The effect sizes of the group differences were large, especially for the SST, and comparable to those found with questionnaire measures of interpretation bias (Orchard et al.
2016b). Of note, as we calculated relative bias scores, our results do not elucidate if the more negative interpretation biases in depressed children and adolescents were due to a lack of positive interpretations or an excess of negative interpretations. However, an additional analysis of the AST data with absolute positive and negative scores instead of a relative bias score indicated that group differences in the AST were mainly driven by the MD group being more likely to endorse
negative interpretations compared to HR and LR groups while no differences were found for positive interpretations (results of this analysis are presented in Supplement
3). It also has to be acknowledged that the foil ratio of the AST was also more negative in the MD group than in the HR and LR groups (although with smaller effect sizes:
d = 0.6–0.8 vs.
d = 0.8–2.9). As the foil ratio represents the tendency to endorse non-specific negative statements this suggests that the more negative interpretation bias in the MD group may partly be explained by a more general negative response bias. Our study is the first to focus on comparing interpretation biases in depressed versus non-depressed youth using multiple experimental measures. The results extend those of prior studies that have investigated interpretation biases in depressed adolescents (aged 12–18; Orchard et al.
2016b; and 14–21 years; Micco et al.
2014) to a younger age group. The presence of negative interpretation biases in depressed children and adolescents corroborates the assumption that negative interpretation biases are a characteristic of individuals with depression not only in adults and adolescents but also in 9–14 year old youth and provides empirical support that cognitive theories of depression (e.g., Disner et al.
2011) apply to this group as well. However, as it remains unclear how interpretation biases emerge across childhood and adolescence, future studies may compare interpretation biases between different age groups, e.g., children vs. adolescents, or investigate interpretation biases longitudinally across childhood and adolescence.
The bias score was strongly positively related to depressive symptoms in the full sample, replicating previous results in youth with depression (Micco et al.
2014) or elevated symptoms of depression (de Voogd et al.
2017) as well as unselected samples of adolescents (e.g., Klein et al.
2018; Orchard et al.
2016a). However, when correlations were calculated separately within each group, consistent correlations with depressive symptoms were found only for interpretation bias as assessed with the SST, while the interpretation bias assessed with the AST only correlated with depressive symptoms within the MD group, probably due to lower values and/or less variance of depression, anxiety, and IB
AST scores in the HR and LR groups. Similar relationships were found for anxiety symptoms, which is not surprising considering the well-established association of anxiety and interpretation biases in children and adolescents (Stuijfzand et al.
2018). However, a comparison of the correlation coefficients indicated that for the interpretation bias score as assessed with the SST, the association with depressive symptoms was significantly stronger than the association with anxiety symptoms, suggesting at least partial specificity. For the interpretation bias score as assessed with the AST, on the other hand, correlations with symptom scores did not differ.
The second aim of the study was to determine the extent to which interpretation biases are more pronounced in currently depressed youth compared to at-risk youth. In line with our expectations and previous studies (Dearing and Gotlib
2009), children and adolescents at high risk for depression showed a more negative interpretation bias compared to children and adolescents at low risk for depression (see also Sfärlea et al.
2019). However, only the interpretation bias as assessed with the SST (not the AST) was more negative in the HR group than in the LR group and it was much less pronounced than in the MD group. This is the first time interpretation biases are compared between currently depressed children and adolescents and children and adolescents with a high or low risk for depression. The results indicate that while being to some extent already present in at-risk populations,
8 negative interpretation biases are strongly exacerbated in currently depressed children and adolescents.
The two tasks assessing interpretation biases yielded divergent results: the AST differentiated only between depressed and non-depressed children and adolescents and was related to depressive symptoms only within the MD group, while the SST also differentiated between high- and low-risk youth and was associated to depressive symptoms within all groups. Moreover, interpretation bias scores from the two tasks were only related within the MD group. Based also on our previous results (Sfärlea et al.
2019), we suppose that the AST and the SST capture different aspects of interpretation (an issue which Everaert et al.
2017, pointed out as especially important to investigate): the SST is more cognitively demanding due to the time constraint and the cognitive load procedure, so less resources are left for volitional control and deliberate response strategies. Therefore, the SST may capture a more automatic (in terms of quick and effortless processing that occurs unintentionally and uncontrollably; cf. Beevers
2005; Teachman et al.
2012) and implicit aspect of interpretation. The AST, on the other hand, allows more reflection on one’s answers and might therefore be more susceptible to distorted responding, similarly to self-report measures (e.g., Gotlib and Joormann
2010). Hence, the AST presumably measures a more conscious and explicit aspect of interpretation (see Sfärlea et al.
2019, for more details). According to this assumption, our results suggest that an implicit interpretation bias can already be found in at-risk youth before onset of a depressive disorder and thus might act as a cognitive vulnerability or risk factor contributing to the development of depression (as suggested by theoretical models; e.g., Disner et al.
2011). The explicit interpretation bias, on the other hand, was only found in the currently depressed group, indicating that this type of bias may arise as a consequence of depressive symptomatology. The finding that these two aspects of interpretation operate differently with respect to the question of being present already in youth at risk for depression or only in currently depressed children and adolescents contributes to a more comprehensive and differentiated understanding of interpretation biases in youth depression. However, the cross-sectional design of the study does not allow any conclusions about time course or causality: we cannot determine the predictive value of interpretation biases for prospectively predicting the onset of an episode of MD, i.e., whether the more negative interpretation bias in the HR group compared to the LR group indeed acts as a risk factor for the development of MD. Likewise, we cannot conclude if the more negative interpretation biases we found in the MD group compared to the HR group are consequences of the depressive disorder or had already characterized those individuals that developed MD before disorder onset. Longitudinal research is needed to address these important questions as well as to investigate what role negative interpretation biases play in the maintenance of depressive symptoms.
Clinical Implications
We found strong negative interpretation biases in children and adolescents with MD on explicit as well as implicit levels. This suggests that therapeutic attempts to modify these biases in depressed youth might be more efficient if they address interpretation biases not only explicitly via Cognitive Behavioral Therapy (e.g., J. S. Beck
2011) but also implicitly, for example via Cognitive Bias Modification interventions that have been shown to successfully modify interpretation biases not only in healthy (Lothmann et al.
2011) but also in depressed adolescents (LeMoult et al.
2018; Micco et al.
2014).
The presence of negative implicit interpretation biases also in youth at high risk for depression, on the other hand, indicates that this kind of interpretation bias might also be a target for preventive approaches trying to reduce the impact of cognitive vulnerabilities in children of depressed parents. Modifying cognitive processes using implicit methods might enhance the efficacy of prevention programs in this high-risk group, whose effects are rather small and short-term (Loechner et al.
2018). However, as some studies implementing Cognitive Bias Modification interventions for interpretation bias reported that those lacked transfer effects (e.g., LeMoult et al.
2018; Yiend et al.
2014), these interventions clearly need to be refined and improved before representing useful therapeutic tools for treatment and prevention of depressive disorders. Moreover, as any intervention intended for younger age groups, Cognitive Bias Modification interventions for children and adolescents need to be age-adapted, e.g., by using picture-based instead of text-based stimuli for younger children.
Furthermore, as the two measures of interpretation bias presumably capture different aspects of interpretation, the AST and the SST could be useful tools for assessing the extent to which existing interventions are able to change interpretation biases in children and adolescents with MD separately on conscious as well as automatic levels.
Strengths
The present study makes a significant contribution to our knowledge of the role of interpretation biases in youth depression holding several methodological strengths.
Two different tasks were administered to experimentally assess interpretation biases. The reliability of the tasks was determined and turned out to be at least good for both measures (corresponding to e.g., Micco et al.
2014; Novović et al.
2014). Furthermore, the correlations between bias scores and depressive symptoms underline the construct validity of the measures as indicators of depressive processing.
Moreover, not only did all participants included in the study undergo extensive diagnostic assessment, psychopathology was also carefully assessed in one (HR group) or both (LR group) of their parents via a diagnostic interview instead of relying on self-report of mental disorder history only.
Limitations
One limitation of the present study is that the three groups investigated differed in age with participants in the MD group being significantly older than participants in the HR and LR groups. This probably results from the prevalence of depression being rather low in childhood and rising substantially with puberty (Thapar et al.
2012) and therefore the majority of the participants in the MD group being 12 to 14 years old. However, as age was not related to bias scores, it is unlikely that the age difference accounts for the group differences we found.
Another limitation results from nearly half of the participants in the MD group having a comorbid anxiety disorder. Also, not only depressive but also anxiety symptoms were related to interpretation biases, which was to be expected considering that the stimuli used in the tasks – even though adapted to our study population – were not entirely depression-specific due to the symptom overlap between depression and particular anxiety disorders like social anxiety disorder or generalized anxiety disorder. Therefore, it cannot be ruled out that comorbid psychopathology contributed to our results. However, the association with depressive symptoms was stronger than the association with anxiety symptoms (for the SST, which is the more depression-specific measure), suggesting at least partial specificity.
Furthermore, it remains unknown if group differences in interpretation bias, particularly the difference between HR and LR groups in the SST, can also be observed during baseline mood and without the cognitive load, as interpretation biases were only assessed following a negative mood induction and the SST was not applied without the cognitive load procedure. These possibilities should be addressed by future studies as they have important implications for cognitive models of depression.
Finally, since most participants in the MD group were recruited at a Department of Child and Adolescent Psychiatry or through licensed outpatient psychotherapists, it is likely that most of them were receiving some form of psychotherapy at the time of their participation (unfortunately, this was not assessed systematically). Since psychotherapy, particularly Cognitive Behavioral Therapy, targets negative interpretation biases, our effect sizes might be underestimates of the effect sizes in untreated youth depression. Furthermore, since a considerable proportion of the participants in the HR group were recruited through a study evaluating a family-based prevention program for children of parents with a history of depression (Platt et al.
2014), our HR participants might have been less vulnerable to depression than the average offspring of depressed parents (see Sfärlea et al.
2019, for a more detailed discussion).
9 In summary, our MD and HR samples might not be entirely representative and group differences might be underestimated in our study.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.