Mood disorders, including depressive disorders and bipolar disorders, are common and impairing psychiatric conditions (Merikangas et al., 2010; Kessler et al., 2012; Walker et al., 2015). Mood disorders have been linked with cognitive impairments, which may contribute to worse functioning in daily life (Baune et al., 2010; Cotrena et al., 2016). In particular, mood disorders often present with impairments in executive function (EF; Snyder, 2013; Cotrena et al., 2016, 2020), the ability to plan and carry out goal-directed behavior. Mood disorders often emerge during adolescence and early adulthood (Paus et al., 2008), which also marks a time of considerable EF development (Best and Miller, 2010). A better understanding of the relationship between mood disorders and EF during this sensitive period may help us develop more effective early interventions to target cognitive impairments and improve functioning in daily life.

However, the nature of the association between mood disorders and EF remains unclear, in part due to the challenges of measuring these constructs. First, EF poses measurement challenges because EF involves controlling lower-level processes to carry out task-specific goals (Burgess, 1997; Miyake et al., 2000). Therefore, measures based on a single task can be contaminated by the specific lower-level processes required by the task. Second, EF is not a unitary construct, and tasks may vary in the extent to which they tap specific EFs (Miyake et al., 2000; Miyake and Friedman, 2012). For example, the Unity and Diversity model of EFs proposes that EFs can be broken down into shared and specific factors (Miyake et al., 2000; Miyake and Friedman, 2012). Common EF, which represents the variance shared across diverse EF tasks, has been theorized to relate to the general ability to develop and maintain a goal, and use that goal to direct ongoing behavior (Miyake and Friedman, 2012; Friedman and Miyake, 2017). However, performance on an individual task not only reflects common EF but also specific EFs (e.g., shifting between task sets, or updating the contents of working memory) and lower level perceptual processes (Miyake et al., 2000; Miyake and Friedman, 2012; Friedman and Miyake, 2017). Therefore, studies that only use a single task to measure EF face issues with task impurity and potential biases towards specific EFs.

Multi-task approaches combined with factor analysis can aid in addressing these issues in measuring EF by allowing for the separation of common and specific dimensions of EF and reducing task-specific impurities (Friedman and Miyake, 2017). Indeed, prior research adopting a latent variable approach to measuring EF has suggested that deficits in common EF correlate with higher severity of internalizing symptoms, including depressed mood as well as anxiety (Hatoum et al., 2018; Snyder et al., 2019). In contrast, specific EFs such as updating and shifting abilities have not been consistently linked to internalizing psychopathology (Hatoum et al., 2018; Snyder et al., 2019). Lower common EF may have significant potential to cause problems in everyday life due to its involvement in goal-directed behavior, which could include ignoring distracting thoughts, regulating emotions, and planning everyday tasks (Snyder et al., 2019). Other studies not explicitly deriving a common EF factor have found links between mood disorders and worse performance across a variety of EF tasks, which aligns with the hypothesis that mood pathology relates to what is shared across various EF tasks (for reviews, see Snyder, 2013; Cotrena et al., 2020). In sum, mood disorders appear broadly associated with deficits in common EF, reflecting general deficits in goal-directed behavior.

Whereas prior work has distinguished between common vs. specific dimensions of EF, it has tended to treat mood symptoms as unitary, ignoring the marked heterogeneity of symptoms that can characterize any individual’s clinical profile. Mood disorders are highly heterogeneous categories. Based on criteria from the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; American Psychiatric Association, 2013), 227 unique symptom profiles of depression exist, and two individuals with that diagnosis may not have any shared symptoms. For example, some individuals diagnosed with a mood disorder display higher levels of anhedonia, defined as the loss of interest or pleasure (Pizzagalli, 2014). Others may report elevated physiological arousal or anxiety (Ionescu et al., 2013), while others may experience manic or hypomanic symptoms even if they do not meet the criteria for a bipolar diagnosis (McIntyre et al., 2017). Many psychiatric symptoms are also transdiagnostic, and appear across a range of disorders and at subthreshold levels in the general population (Dalgleish et al., 2020). It is unknown which of these symptom dimensions are associated with common EF, yet answering this question may help to explain why some individuals diagnosed with mood disorders experience higher levels of cognitive impairment.

However, investigating these issues requires facing a critical measurement challenge associated with the study of psychopathology: symptoms tend to be highly correlated (Caspi et al., 2014). Thus, even when studies have employed continuous measures of symptom dimensions, it remains ambiguous whether correlations are driven by unique attributes of that symptom dimension. The shared variance across symptom dimensions may itself represent a meaningful dimension of psychopathology that can help explain the association between mood disorders and EF (Caspi et al., 2014; Hatoum et al., 2018; Snyder et al., 2019). For example, the shared variance might reflect greater distress, general negative affect, or symptom burden. Deficits in EF may influence or reflect an individual’s ability to regulate their emotions (Zelazo and Cunningham, 2007; Pruessner et al., 2020), thereby increasing an individual’s susceptibility to experiencing greater distress or multiple types of mood symptoms (Joormann and Stanton, 2016). Lower common EF has also been linked with increased dependent stress (i.e., stressful events that an individual has some influence over) and subsequent rumination, which in turn may lead to a variety of mood symptoms (Snyder and Hankin, 2016; Snyder et al., 2019). In the opposite causal direction, a greater symptom burden may impair EF. For example, individuals experiencing increased symptom load may be more prone to negative, self-focused patterns of thought (Calvete et al., 2015), which may result in worse performance on measures of EF by consuming attentional resources (Watkins and Brown, 2002). However, to our knowledge, no studies have simultaneously examined how shared and specific mood symptom dimensions may relate to common EF. Therefore, the question remains: are associations between mood disorders and common EF driven primarily by what is shared across mood symptoms, or do certain symptom dimensions have additional unique associations with common EF?

To help address this question, bifactor models can serve as a means of simultaneously examining the shared and specific features of psychopathology symptoms. Bifactor models are a class of structural equation models that can partition a set of measures into shared and specific variance. A general factor captures the shared variance across all dimensions, and orthogonal specific factors capture variance unique to a particular dimension that remains after taking into account the general factor. We note that bifactor models have also come under critique; for example, these models can result in irregular or inadmissible results such as small or negative factor variances or loadings (Eid et al., 2017). Bifactor S-1 models have been proposed as a means of mitigating these issues (Eid et al., 2017; Burke and Johnston, 2020; Burns et al., 2020; Heinrich et al., 2020). These models use an additional set of indicators as a reference domain, which loads exclusively onto the general factor but does not have its own specific factor. Together, bifactor models have value as tools for partitioning variance, and bifactor S-1 models may help further address technical issues typically associated with bifactor models.

The current study used a bifactor S-1 model to examine associations between common EF and general and specific dimensions of mood symptoms in a sample of adolescents and young adults (ages 13–25 years) with high variance in symptoms of mood disorders. We chose to focus on this age group as it marks a critical developmental window that captures the intersection of EF development (Best and Miller, 2010; Friedman et al., 2016) and higher risk of emerging mood pathology (Paus et al., 2008). We pooled the sample across two study sites and study protocols to maximize the power and generalizability of the analyses. In confirmatory factor analyses, we examined associations between a common EF factor and a general depression factor representing shared variance across mood symptom dimensions, as well as specific factors anchored to items from canonical Anxious Arousal, Loss of Interest, Loss of Pleasure, and Mania scales. These symptom dimensions are prevalent and separable symptoms of psychopathology that often characterize or correspond with depressed mood (Watson et al., 1995a, b; Watson and Naragon-Gainey, 2010; Kendler, 2016). While anxious arousal is commonly associated with anxiety disorders (e.g., panic disorder) we note that altered physiological arousal and heightened anxiety have historically been considered central features of mood pathology (Kendler, 2016, 2017). Anxious arousal has also been previously used to predict cognitive functioning in individuals with depression (Quinn et al., 2012). Additionally, we evaluated exploratory bifactor models to compare data-driven symptom factors and their associations with Common EF against the confirmatory model. Based on prior research, we hypothesized that lower Common EF would be significantly associated with higher levels of the general depression factor from the confirmatory symptom model, and a similar general factor from the exploratory symptom model. Secondary analyses included tests of relationships between Common EF and each symptom dimension in its own model, to identify potential differences in results when shared symptom variance remained in the model.

This study examined associations between common EF and general or specific dimensions of mood symptoms, in adolescents and young adults with high variance in symptom severity. In both confirmatory and exploratory bifactor models, higher levels of a general depression factor capturing shared variance across mood symptom dimensions was significantly associated with lower executive function across tasks. In addition, higher levels of a specific factor primarily reflecting physiological arousal were significantly associated with lower common EF, in both confirmatory and exploratory analyses. Altogether, results indicate that general deficits in goal-directed behavior are broadly associated with features shared across mood symptom dimensions, but also specifically associated with physiological arousal.

The association between higher levels of general depression and lower common EF (identified in both confirmatory and exploratory factor analyses) aligns with previous studies, which have found associations between the shared variance across depression and anxiety symptoms and common EF at both the behavioral and genetic level (Hatoum et al., 2018; Gustavson et al., 2019; Snyder et al., 2019). In our confirmatory factor analyses, the general depression factor was anchored by the MASQ general distress—depression and loss of interest items, which together reflect the cardinal symptoms of depression (sad mood and anhedonia). In our exploratory factor analysis, the General Depression factor showed a similar pattern of loadings, with the highest loadings from the MASQ general distress—depression and loss of interest scales. The association between general depression and common EF can be interpreted in several ways. Lower overall EF may represent vulnerability to experience multiple types of (often, co-occurring) symptoms that are together captured by the general depression factor, perhaps especially under conditions of stress (Snyder et al., 2019; Peterson et al., 2021). Alternatively, the negative affective state captured in the general depression factor may have an impairing effect on common EF. For example, greater distress or negative effect may exacerbate the use of maladaptive response styles such as rumination (Nolen-Hoeksema et al., 1999; Calvete et al., 2015), which in turn can have detrimental effects on task performance (Watkins and Brown, 2002). Recent work supports the existence of bidirectional pathways, in which EF may act as a vulnerability factor for psychopathology, and psychopathology may in turn exacerbate dysfunctions in EF (Romer and Pizzagalli, 2021; Zainal and Newman, 2021). Additional longitudinal studies extending this work to other age groups and further exploring potential moderators and mediators may help illuminate the mechanistic pathways underlying this association.

These results highlight a significant association between common EF and general depression in adolescence and young adulthood. Prior research has supported such an association in children, adolescents, and adults (Caspi et al., 2014; Martel et al., 2017; Snyder et al., 2019), suggesting that this relationship is present throughout the lifespan. However, adolescence may be a particularly salient developmental period for evaluating the association between common EF and mood pathology, given that this is a time of profound EF development as well as a period of heightened risk for mood problems. An interesting direction for this work will be investigating the interactions between cognitive development and mood pathology. For example, one study examining adolescents and young adults found that the association between common EF and general psychopathology increased with age (Snyder et al., 2019); another study identified a similar pattern in which the association between mania and common EF strengthened during this developmental period (Kaiser et al., 2021). While outside the scope of the current study, examination of the dual and potentially bidirectional effects of cognitive development and psychopathology warrants continued exploration, especially in longitudinal samples.

Consistent with prior research (Warren et al., 2021), we also identified a significant association between a symptom dimension of physiological arousal and lower common EF. In the confirmatory model, this factor was restricted to canonical anxious arousal items (which primarily reflect sympathetic nervous system arousal). In the exploratory model, a specific factor emerged that predominantly captured items from the canonical anxious arousal scale (e.g., “Muscles twitched or trembled”). However, it also captured small but significant loadings from other items reflecting greater positive affect and lower negative affect. This pattern could reflect the fact that the general factor already captures negatively valenced forms of arousal, and what remains simply reflects physiological arousal. Across both confirmatory and exploratory factor analyses, lower physiological arousal was associated with better common EF, suggesting that EF may serve as a protective factor that helps individuals modulate their level of arousal through its role in self-regulation (Nigg, 2017), or that excessively high arousal disrupts task performance (Yerkes and Dodson, 1908). Notably, the modification indices also suggested a potential correlation between the mania-specific and arousal-specific factors. While this correlation was not added to retain orthogonality of the factors, it may reflect the fact that heightened sympathetic nervous system arousal often accompanies mania (Henry et al., 2010). Thus, physiological arousal in this sample may be occurring in the context of mania as well as anxiety. Altogether, the association between physiological arousal and common EF suggests that this symptom dimension may relate to common EF through a separate mechanistic pathway from general depression.

Other specific factors were not associated with common EF in either the confirmatory or exploratory bifactor models. However, all symptom dimensions were individually associated with common EF when modeled separately—with the exception of the standalone loss of pleasure factor. One interpretation is that low positive affect does not interfere with EF; some research has suggested that positive mood can in fact impair EF performance (Mitchell and Phillips, 2007). Alternately, common EF may primarily protect individuals from negative affect without necessarily boosting positive affect (e.g., Short et al., 2016). Null effects could also originate from methodological differences: the canonical loss of pleasure scale is the only measure in the MASQ that consists entirely of reverse-coded statements; thus, its specific variance may in part reflect method variance. Finally, while this scale may capture the presence or absence of positive effect, it might not effectively capture the inability to experience the pleasure that defines consummatory anhedonia (Treadway and Zald, 2011). In contrast, the standalone loss of interest factor was individually associated with EF, highlighting the importance of carefully decomposing the symptom dimension of anhedonia.

Overall, the confirmatory and exploratory analyses revealed consistent patterns of correlations between common EF and symptom dimensions: both identified significant associations between common EF and factors reflecting general depression and physiological arousal. However, an additional comment is warranted regarding differences in the symptom dimensions that emerged from these two modeling approaches. We note that while both approaches yielded conceptually similar symptom structure, there were also key differences: e.g., the exploratory model yielded four specific factors, whereas the confirmatory model fit best with three specific factors. Moreover, the nature of the factors derived from each analysis differed in several ways. For example, in the confirmatory model, the canonical loss of interest items were subsumed by the general factor, which aligns with prior models also using the MASQ (Banich et al., 2020). However, in the exploratory model, a specific factor emerged which included a number of items from the loss of interest scale, but also items from the canonical loss of pleasure, general distress—depression, and anxious arousal scales (Specific Factor 1). This exploratory specific factor seemed to capture a mixture of low motivation, low positive effect, and sad mood, but also had negative associations with symptoms of fatigue and physiological arousal. These patterns point to a divergence between the MASQ conceptualization of the loss of interest dimension, and how a data-driven approach clusters these items. Given the exploratory nature of these analyses, future studies should examine the replicability of these dimensions in other samples. Alternative measures of anhedonia may also help identify separable components of this construct.

Several limitations of this study should be addressed. First, we utilized a cross-sectional design, which does not allow for clarification of the temporal relationship between common EF and various symptom dimensions. Future studies examining changes in the associations between general and specific mood symptoms and common EF over time, or comparing high-risk, currently symptomatic, and remitted individuals, may help illuminate the causal mechanisms underlying these relationships. In addition, this study focused on mood disorders rather than other forms of psychopathology, leaving it unknown whether other specific symptom dimensions have similar or different associations with common EF. This approach allowed a targeted investigation of mood symptom dimensions, but future research may extend to additional clinical samples. For example, studies that include more individuals with primary anxiety disorders, as well as mood disorders, may be better suited to looking at internalizing symptoms more generally. Additionally, while our sample was transdiagnostic across mood disorders, there was a higher proportion of individuals with unipolar vs. bipolar forms of depression (Table 2). However, the sample still exhibited a significant variance in manic symptoms, consistent with prior studies showing that individuals with unipolar disorders frequently experience subclinical manic symptoms (Carlson and Kashani, 1988; McIntyre et al., 2015). A valuable next step would be to explore these models in larger samples of individuals with bipolar disorders. These models could be augmented in a variety of ways, such as by expanding the EF model, incorporating additional symptom dimensions, or controlling for other variables that may influence these relationships. Here, we chose to focus on common EF based on prior research suggesting that this dimension is most consistently linked with psychopathology (Hatoum et al., 2018; Snyder et al., 2019). However, future studies utilizing a larger suite of EF tasks may wish to extract specific EF factors, including but not limited to shifting and updating, and examine their relationship with various symptoms of psychopathology. Additionally, a next step in testing this model will be to incorporate objective measures (e.g., of physiological arousal) and compare them with subjective self-report measures of symptom dimensions. In this study, we identified an association between higher self-reported sympathetic arousal and lower common EF; however, prior work has suggested that self-report and physiological measures of arousal can diverge (Miers et al., 2011). Therefore, a multi-modal evaluation of these constructs may provide a more nuanced understanding of psychiatric health and cognitive functioning. Future analyses may also benefit from controlling for factors such as processing speed, since prior work has shown that processing speed is correlated with, but separable from, Common EF (Friedman et al., 2008). Finally, the model could be modified by using a different reference domain for the general factor. For example, we chose in our confirmatory factor analysis to anchor the general factor with items from the canonical general distress—depression scale, based on the predominance of individuals with mood disorders in our sample. However, the most appropriate reference domain for a general factor in a bifactor S-1 may vary depending on the research question, study sample, and theoretical interpretations of the meaning of the general factor.

In sum, this study identified significant associations between lower executive functioning ability (common EF) and symptom dimensions representing general depression and physiological arousal. These results suggest that general depressive symptoms and physiological arousal may be associated with common EF via different mechanisms, and highlight the value of distinguishing between the shared and specific features of symptom dimensions.

The datasets presented in this article are not readily available because evaluation of other experimental hypotheses are ongoing in this dataset. Data will be available upon request after other a priori analyses have been completed. Requests to access the datasets should be directed to RK, roselinde.kaiser@colorado.edu.

The studies involving human participants were reviewed and approved by the University of Colorado Boulder Institutional Review Board and the University of Los Angeles California Institutional Review Board. Written informed consent to participate in this study was obtained from legal adults (ages 18 and older) and parental consent and child assent were obtained from legal minors (ages 17 and younger).

The studies involving human participants were reviewed and approved by the University of Colorado Boulder Institutional Review Board and the University of Los Angeles California Institutional Review Board. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

RK and EP developed the study concept and design. EP, CN, BR, CH, CS, LO, SG, JK, JF, CR, AG, and SS performed data collection. EP, CN, BR, CH, and CS performed clinical evaluation, under RK’s supervision. DM provided consultation on recruitment and clinical evaluation. EP performed analyses with assistance from RK, NF, and HS. The manuscript was drafted by EP and all other co-authors provided critical revisions. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.