Intergenerational Transmission of Depression: Testing A Comprehensive Set of Putative Mediators

Abstract

Offspring of depressed parents at increased risk for developing depression. They also differ from offspring of non-depressed parents on numerous risk factors, including personality, cognitive biases, neural processing of emotional stimuli, subthreshold depression and anxiety and irritability symptoms, and interpersonal functioning (Goodman, 2020; Gotlib et al., 2023). However, few studies have tested whether these risk factors mediate the transmission of depression from parents to offspring. In a longitudinal study of a community sample of youth and parents (n=481), we examined if these risk factors mediated the relationship between parental and youth depression using structural equation modeling. Separate models revealed significant indirect effects for negative emotionality, cognitive biases, and depressive, anxiety and irritability symptoms. However, when these independent mediators were entered in a model that also included subthreshold depressive symptoms only depression had unique indirect effects. Our findings are contextualized based on our analytic approach and the highly intertwined and overlapping nature of the mediating risk factors assessed with depressive symptoms.

General Scientific Summary

Offspring of depressed parents are at increased risk for becoming depressed themselves, but little is known about the factors that mediate this intergenerational transmission. We examined a number of putative mediators and found that youth temperamental negative emotionality, cognitive biases, subclinical depressive symptoms, and anxiety and irritability symptoms at age 12 mediated the association between prior parental depression and youth depression in mid-late adolescence, although these effects were not unique when depressive symptoms were also considered.

Introduction

One of the best-established findings in the depression literature is that depressive disorders are transmitted across generations. Indeed, offspring of a depressed parent have a 2–4-fold increase in risk for developing a depressive disorder themselves (Hammen, 2017; Weissman et al., 2016). In part, this is due to genetic influences, which have a small-medium size effect (Kendler et al., 2018). In addition, a number of other risk factors, which may, in part, reflect these genetic influences, have been hypothesized to mediate the transmission of depression from parents to children (e.g., Goodman, 2020; Gotlib et al., 2023; Hammen, 2017). Identifying processes involved in intergenerational transmission can contribute to prevention and early intervention. However, surprisingly few empirical studies have formally tested the indirect effects of plausible mediators. Moreover, we are unaware of any studies that have included multiple mediators and examined their unique effects.

In this paper, we draw on an ongoing longitudinal study of a community sample (Klein & Finsaas, 2017) to examine the indirect effects of a wide range of putative mediators on the association between parental depression and the subsequent development of depressive disorders in adolescence. Although multiple types of factors may link parent and offspring depression, this paper focuses on offspring characteristics and leave contextual factors for future investigation. In order to provide a clear temporal ordering of independent, mediating, and dependent variables, we assessed maternal and paternal histories of depressive disorders through offspring age 9; mediators were assessed when offspring were 12 years old, prior to the beginning of the period of maximal risk for depression; and offspring depression was assessed at ages 15 and 18, when risk for depression dramatically increases (Rohde et al., 2013). Assessments used a multi-method, multi-informant, multiple-indicator approach.

We used mediators from the age 12 wave of our study for two reasons: (1) assessing mediators that are proximal to the increase in the incidence of depression, which begins around age 14 (Solmi et al., 2022), should maximize the power to detect effects; and (2) by age 12, youth have become relatively good reporters of their own experiences and mental states (e.g., Hyland et al., 2022), allowing us to augment informant and laboratory data with self-report assessments that are not feasible or of questionable validity at younger ages. Offspring who had already developed depressive disorders by age 12 were excluded from the study.

Putative mediators were selected based on evidence that they are both associated with parental depression and prospectively predict depressive disorders or symptoms in childhood or adolescence. The putative mediators fell into 5 domains: 1) personality (or temperament) traits; 2) cognitive biases; 3) neural processing of emotional stimuli; 4) symptoms, specifically of depression, anxiety and irritability, that frequently precede the onset of depression; and 5) interpersonal functioning. To examine the unique effects of the hypothesized mediators on change in depressive symptoms from early to mid/late adolescence, we also conducted a second set of analyses for all putative mediators except depressive symptoms in which subthreshold depressive symptoms at age 12 were included as a second, competing mediator in each individual mediation model.

Personality.

Depression may be transmitted across generations through shared individual difference traits (Mackin et al., 2022). Numerous cross-sectional studies (Klein et al., 2011; Kotov et al., 2010), have linked depression to higher levels of neuroticism (or negative emotionality; NE), lower extraversion (or the related construct of positive emotionality; PE), and higher disinhibition (or lower levels of the related constructs of conscientiousness and effortful control). A handful of studies have reported that offspring of depressed parents exhibit higher levels of neuroticism (Goodman et al., 2011; Lauer et al., 1997) and lower levels of extraversion (Durbin et al., 2005; Goodman et al., 2011), than offspring of nondepressed parents. Neuroticism, and to a lesser degree, low extraversion, also prospectively predict the development of depressive disorders and symptoms in youth (e.g., Kotelnikova, et al., 2015; Michelini et al., 2021). To our knowledge, only two studies have formally tested whether offspring personality traits mediate the association between parental depression and offspring internalizing psychopathology. In a sample of young children, Allen et al. (2019) reported that higher neuroticism and lower conscientiousness at age 5 mediated the association of parental depressive symptoms with offspring internalizing symptoms at age 9. In contrast, Mackin et al. (2022) did not find significant indirect effects of offspring neuroticism and extraversion on the relationship between parent and offspring depressive disorders in adolescent females.

Cognitive Biases.

Cognitive models of depression emphasize the role of maladaptive cognitions and biases in the development of depressive disorders (Clark et al., 1999). Numerous cross-sectional studies have shown that individuals with depression view themselves in a more negative and critical light, recall more negative and less self-referential positive information, and engage in more rumination and brooding than non-depressed individuals (LeMoult & Gotlib, 2019). Importantly, offspring of depressed parents also exhibit higher levels of rumination and brooding (Gibb et al., 2012), negative inferential styles (Shankar & Gibb, 2024), and attentional biases (Joormann et al., 2007) than offspring of non-depressed parents. Moreover, negative cognitive style (Alloy et al., 2000), self-criticism (Kopola-Sibley et al, 2017), negative self-referential processing (Connolly et al., 2016), and rumination (Abela & Hankin, 2011) prospectively predict depressive symptoms or disorders in youth. However, to our knowledge, only one study has examined whether cognitive factors mediate the association between parent and offspring depression. In a large sample of adolescent girls, Dunning et al. (2021) reported that maternal depression history predicted adolescent depressive symptoms via negative views of the self.

Neural Processing of Emotional Stimuli.

Neural processing of, and reactivity to, emotional stimuli may also contribute to the intergenerational transmission of depression (Kujawa & Burkhouse, 2017). This has been supported by studies using both task-based functional magnetic imaging and event-related potentials (ERPs) (Burkhouse & Kujawa, 2023). For example, offspring of depressed mothers show a blunted late positive potential (LPP) to emotional stimuli compared to children of nondepressed mothers (Kujawa & Burkhouse, 2017). Additionally, offspring of depressed mothers show a blunted reward positivity (RewP) in response to monetary and social incentive tasks compared to children of non-depressed mothers (e.g., Freeman et al., 2022). Blunted LPP and RewP also prospectively predict the onset of depressive disorders in adolescents (Michelini et al., 2021; Nelson et al., 2016). Although these data suggest that alterations in the neural processing of emotional stimuli may mediate the intergenerational transmission of depression, we are unaware of any studies formally testing this conjecture.

Depression, Anxiety, and Irritability Symptoms.

Parental depression has been consistently linked to increased levels of depressive symptoms in offspring (Goodman et al., 2011; Gotlib et al., 2024). Additionally, subthreshold depressive symptoms that do not meet criteria for a diagnosis are a potent precursor and/or risk factor for the subsequent onset of depressive disorders in adolescence and young adulthood (Bertha & Balázs, 2013; Klein et al., 2009; Lee et al., 2019). Together, this suggests that subclinical depressive symptoms may play a role in mediating intergenerational transmission.

In this paper, depressive symptoms are also accorded a second role in addition to being a potential mediator of intergenerational transmission. Thus, we conducted a second set of models in which depressive symptoms at age 12 were included as a competing mediator in the models for each of the other mediators. This enabled us to examine whether the other variables mediate change in depression from early to mid/late adolescence.

Children of depressed parents also frequently experience significant anxiety and irritability. For example, offspring of depressed parents are over three times more likely to exhibit anxiety disorders compared to children of non-depressed parents (Weissman et al., 2016). Parental depression is also associated with elevated levels of irritability in offspring (e.g., Whelan et al., 2015). Moreover, both anxiety (Cummings et al., 2014) and irritability (Vidal-Ribas et al., 2016) symptoms in youth predict subsequent depressive disorders and symptoms. Indeed, Rice et al. (2017) observed that anxiety and irritability, but not depression, symptoms independently predicted the first onset of depressive disorders in the offspring of depressed parents. However, Rice and colleagues could not test mediation due to the lack of a comparison group of non-depressed parents. We are unaware of any studies testing whether subthreshold depression and anxiety and irritability symptoms mediate the intergenerational transmission of depression.

Interpersonal Functioning.

Interpersonal models of depression emphasize the role of relational functioning in the onset and maintenance of depressive symptoms (Hames et al., 2013). In youth, depression is associated with low social support, peer rejection and victimization, and less closeness and more conflict in relationships with parents, siblings, friends, and other peers (Rudolph & Flynn, 2014). Children of depressed parents also exhibit poorer interpersonal functioning than offspring of non-depressed parents (Hammen, 2017). Furthermore, poor interpersonal functioning predicts subsequent depressive disorders and symptoms in adolescents (e.g., Borelli & Prinstein, 2006; Letkiewicz et al., 2023; Platt et al., 2013). In a study of a large community sample, Hammen and colleagues (Hammen et al., 2004, 2012) found that the quality of youth interpersonal relationships mediated the association between maternal and offspring depressive disorders. More recently, Israel and Gibb (2024) reported that maternal criticism of the child mediated the relationship between maternal depression and youth depressive symptoms.

The Current Study

We examined whether 5 domains of youth characteristics (personality/temperament; cognitive biases; neural processing of emotional stimuli; subthreshold depressive and anxiety and irritability symptoms; and interpersonal functioning) mediate the association of parental depressive disorders and the subsequent development of depression in offspring. We used a wide range of multi-informant, multimethod indicators to create a set of sensitive and robust latent constructs representing the putative mediators at age 12 and depression at ages 15 and 18. Then we tested indirect effects separately for each candidate mediator. Finally, in order test whether the putative mediators influence change in depressive symptoms from early to mid/late adolescence, we conducted a second set of analyses in which age 12 depressive symptoms were included as a second mediator in models testing the other candidate mediators.

Methods

Participants

Adolescents and their biological parents were drawn from the Stony Brook Temperament Study, a multi-wave longitudinal study examining early antecedents and pathways to psychopathology from preschool through adolescence (Klein & Finsaas, 2017). Families with a 3-year-old child (N=559) in the Long Island, New York area were identified via commercial mailing lists and invited to participate. An additional 50 families were recruited in the second (age 6) wave to increase the diversity of the sample, for a total N of 609. Children had to live with at least one English-speaking biological parent and could not have significant medical or developmental disorders. Only one child per family was included. Of the 609 youth enrolled in the study, we collected data on parental lifetime history of depression for 591 mothers and 584 fathers. Of the 584 youth with data on history of depression in both parents, 71 youth were missing all indicators for depression outcomes and mediators of interest, reducing the N to 513. Females (46.6%, n=224) were less likely to be included than males (53.4%, n=257), χ²(1, 584) = 6.08, p = .014 and youth who were included had a greater number of parents with a college degree (M = 0.99, SD = 0.80) than youth who were excluded (M = 0.72, SD=0.76), t(548) = 2.74, p = .006. However, youth who were included versus excluded did not different on race/ethnicity, χ²(1, 552) = 0.47, p= .499 or number of parents with a lifetime history of depression (included M =0.56, SD=0.67; excluded M = 0.44, SD=0.65), t(550) = 1.45, p = .146.

Adolescents with a history of depressive disorder including not otherwise specified (D-NOS) prior to age 12 were excluded (n=32) to ensure that youth did not have clinically significant depression prior to the assessment of the putative mediators. This resulted in an analytic sample of 481 youth and their parents. A summary of sample characteristics is provided in Table 1. Assessments were conducted every three years from ages 3–18. Parents provided written consent at each wave through age 15, and children provided written assent at ages 12 and 15. At age 18, youth provided written informed consent. Participants were compensated for their time. All procedures were approved by the Stony Brook University Institutional Review Board. We report how we determined our sample size, all data exclusions (if any), all manipulations, and all measures in the study¹.

Table 1.

Analysis Sample Characteristics

Variable/Measure	n	%	M	SD	Range
Youth Demographics
Sex at Birth (female)	277	45.30	-	-	-
White and Non-Hispanic	389	80.90	-	-	-
Number of Parents with Bachelor’s degree at study entry
0 Parents	23	4.80	-	-	-
1 Parent	303	63.30	-	-	-
2 Parents	153	31.90	-	-	-
Parental Lifetime History of Depression (through youth age 9)
0 Depressed Parents	259	53.80	-	-	-
1 Depressed Parent	175	36.40	-	-	-
2 Depressed Parents	47	9.80	-	-	-
Youth Depression at age 15
K-SADS-PL Depression diagnosis	43	8.90	-	-
CDI-Child Report	-	-	5.45	5.20	0–31.00
CDI-Mother Report	-	-	7.82	5.30	0–34.00
CDI-Father Report	-	-	7.92	5.03	0–36.00
Youth Depression at age 18
K-SADS-PL Depression diagnosis	90	18.07	-	-	-
CDI-Child Report	-	-	7.88	7.16	0–37.00
IDAS-Child Report	-	-	37.02	12.83	20–90.00

Note: K-SADS-PL =Kiddie Schedule for Affective Disorders and Schizophrenia for School-Aged Children, Present and Lifetime Version; CDI=Children’s Depression Inventory; IDAS=Inventory of Depression and Anxiety Symptoms. This table presents data prior to model testing and estimation of missing data. Two parents were missing education data.

Measures

Table 2 summarizes the hypothesized mediating and outcome constructs and their indicators. More detailed information about the measures and their corresponding reliability estimates, means, standard deviations and ranges can be found in the Supplementary Materials.

Table 2.

Summary of Youth Latent Factors and Indicators

Youth Latent Factors	Indicators
Negative Emotionality (NE)	Schedule for Nonadaptive and Adaptive Personality-Youth (SNAP-Y; child-, mother-, and father-report) Affect and Arousal Scale (AAS; child-report)
Positive Emotionality (PE)	SNAP-Y (child-, mother-, and father-report) AAS (child-report)
Disinhibition	SNAP-Y (child-, mother-, and father-report)
Cognitive Biases	Rumination: Children’s Response Styles Questionnaire (child-report) Brooding: Children’s Response Scale (child-report) Self-Criticism: Depressive Experiences Questionnaire (child-report) Self-Referent Encoding Task: Negative Endorsement, Negative Processing, Positive Endorsement, Positive Processing
Neural Processing of Emotional Stimuli	Doors - Monetary Reward Positivity (RewP) Gain-to-Loss Residual Island Getaway - Social RewP Accept-to-Reject Residual Emotion Interrupt Late Positive Potential (LPP) Negative-to-Neutral Residual
Subthreshold Depressive Symptoms	Children’s Depression Inventory-2nd edition (child-, mother-, and father-report)
Anxiety Symptoms	Screen for Child Anxiety Related Disorders (child-, mother-, and father-report)
Irritability	Affective Reactivity Index (child-, mother-, and father-report)
Interpersonal Functioning	Peer Experiences Questionnaire (child-report) Multidimensional Scale of Perceived Social Support-Family and Friends (child-report) Network Relationship Inventory-Closeness and Discordance, child-report UCLA Life Stress Interview - friends, social, mother and father (child and parent interviews)
Depression Outcome	Kiddie Schedule for Affective Disorders and Schizophrenia (child and parent interviews) Children’s Depression Inventory-2nd Edition (child-, mother-, and father-report) The Inventory of Depression and Anxiety Symptoms (child-report)

Parental depression

Parental lifetime history of depression was assessed with the Structured Clinical Interview for DSM-IV, Nonpatient version (SCID-NP; First et al., 1996). Parents were interviewed in-person or by telephone by master’s and doctoral level raters at entry into the study and again when children were 9 years old. The initial interview used a lifetime time frame; the interview at age 9 focused on the interval since the prior assessment. Parents were considered to have a lifetime history of depressive disorder if they met criteria for major depressive disorder (MDD) or dysthymic disorder (DY) at either assessment. We operationalized parental depression as the number of parents with a history of depressive disorders (ranging from 0–2), as offspring of two depressed parents are at greater risk than those with only one depressed parent (Brennan et al., 2002). The SCID-NP was administered only to biological parents, regardless of their current marital status.

Youth depression

Three measures from the age 15 and 18 waves were used to model latent youth depression outcome. The Kiddie Schedule for Affective Disorders and Schizophrenia for School-Age Children, Present and Lifetime Version (K-SADS-PL; Kaufman et al., 1997) is a semi-structured diagnostic interview which was used to assess MDD, DY, and D-NOS. It was administered to both a parent and the child at the age 9, 12, and 15 waves, and to the child alone at age 18. Discrepancies between parent and child reports were resolved by further questioning and interviewer judgment. At age 9, a lifetime reporting period was used. Subsequent interviews focused on the time since the last assessment wave. Diagnoses of depressive disorders in the age 15 and 18 waves (spanning ages 13–18) were included.

The Children’s Depression Inventory, 2^nd edition (CDI-2; Kovacs, 2011) is a self-report measure that was administered to the youth and both parents at the age 12 and 15 waves and solely to youth at the age 18 wave.

The Inventory of Depression and Anxiety Symptoms—Expanded Version (IDAS II; Watson et al., 2012) is a self-report measure that was administered to youth at the age 18 wave. We used the General Depression subscale in our analyses.

Hypothesized Mediators

Measures administered at the age 12 wave were used to create latent variables for 9 hypothesized mediators. Three constructs were from the personality/temperament domain (NE, PE, and disinhibition). Three were symptom constructs that are common in offspring of depressed parents: depression; anxiety, which is the most common comorbidity in depression (Cummings et al., 2014); and irritability, which is a highly transdiagnostic symptom that is included in the diagnostic criteria for MDD in youth but is also included in the criteria for many other disorders (Klein et al., 2021). The remaining three putative mediators, cognitive biases, neural processing of emotional stimuli, and interpersonal functioning, were represented by one latent construct each.

Personality/Temperament.

The Schedule for Nonadaptive and Adaptive Personality-Youth Version (SNAP-Y; Linde et al., 2013) was used to assess NE, PE and disinhibition (vs. constraint). The SNAP-Y was completed by the youth and both parents. The Affect and Arousal Scale for Children (AFARS, Chorpita et al., 2000) provided an additional youth-reported measure of PE and NE.

Cognitive Biases.

Cognitive biases were assessed with one laboratory and three self-report measures. The Self-Referential Encoding Task (SRET, Derry & Kuiper, 1981) is a lab task that assesses endorsement and recall of positive and negative self-descriptive adjectives. Following a sad mood induction, youth were presented with a series of adjectives and asked whether each word characterized them. Then, after a short delay, youth were asked to recall as many words as possible (see Goldstein et al., 2015 for details). Endorsement scores reflect the number of self-descriptive positive and negative adjectives endorsed and can be interpreted as indexing explicit self-concept. Processing scores, believed to reflect cognitive schemas, were derived by dividing the number of positive or negative words endorsed and correctly recalled by the total number of both positive and negative words endorsed.

Child rumination and brooding, a dimension of rumination, were assessed using the Children’s Response Styles Questionnaire (CRSQ; Abela et al., 2007), and the Children’s Response Scale—Brooding Subscale (Ziegert & Kistner, 2002), respectively. Self-criticism was assessed using the adolescent version of the Depressive Experiences Questionnaire self-criticism scale (DEQ-A; Fichman et al., 1994). These scales were completed by the youth only.

Neural Processing of Emotional Stimuli.

We used three tasks to examine event-related potentials (ERPs) in response to emotionally and motivationally salient stimuli. The LPP to negative (relative to neutral) stimuli was assessed with the emotion interrupt paradigm (Kujawa et al., 2016) using developmentally appropriate images from the International Affective Picture System (IAPS). Two tasks examined the RewP. First, we examined response to monetary reward (gain versus loss) using the Doors task (Kujawa, Proudfit, et al., 2014). Second, responses to social reward (acceptance versus rejection) were assessed using the Island Getaway task (Kujawa, Arfer, et al., 2014). Detailed descriptions of these tasks and EEG processing and scoring are provided in the Supplementary Material.

Symptoms.

As described above, subthreshold depressive symptoms at age 12 were assessed with self-, mother-, and father-reports on the CDI. These symptoms can be regarded as subthreshold because youth with a history of diagnosable depression (including D-NOS) were excluded from the sample. Anxiety and irritability symptoms at age 12 were assessed with self-, mother-, and father-reports on the Screen for Child Anxiety and Related Disorders (SCARED; Birhamer et al., 1997) and the Affective Reactivity Index (ARI; Stringaris et al., 2012), respectively.

Interpersonal Functioning.

Youth interpersonal functioning was assessed using three youth self-report scales and a semi-structured interview that was administered to both the youth and a parent. The Multidimensional Scale of Perceived Social Support (MSPSS; Zimet et al., 1988) is a self-report measure of perceived social support from family and friends. The Network of Relationship Inventory: Relationship Quality Version (NRI: RQV, Buhrmester & Furman, 2008) is a self-report measure of the quality of the youth’s relationships with each parent and their best friend. We summed the closeness and discord scores across the three relationships to create composite closeness and discord scores for youth. The Revised Peer Experiences Questionnaire (RPEQ; Prinstein et al., 2001) is a self-report measure of overt, relational, and reputational peer victimization.

Finally, the UCLA Life Stress Interview (LSI; Rudolph & Hammen, 1999) was used to assess interpersonal functioning in the previous 12 months. The LSI is a semi-structured interview that was administered separately to a parent (generally mothers) and youth. Interviewers used their judgement to resolve discrepancies. We summed the ratings for interpersonal functioning in the areas of close friends, broader peer relationships, and each parent.

Data Analytic Strategy

All analyses were conducted in Mplus 8.3 (Muthen & Muthen, 2007) using full maximum likelihood estimation (ML) to account for missing data. This data estimation approach allows youth with data on the independent variable, number of depressed parents, and at least one indicator for a latent mediator or depression outcome to be included in model estimation.

Confirmatory factor analysis (CFA) was used to model latent factors for each of the nine mediator constructs at the age 12 wave and youth depressive outcome (collapsed across the age 15 and 18 waves). Then we used structural equation modeling (SEM) to test whether the hypothesized mediators in the age 12 wave exhibited significant indirect effects between the number of parents with a lifetime history of depression through the age 9 wave and youth latent depression outcomes in the age 15 and 18 waves; see Figure 1. In these models, we included direct effects from: 1) parental depression to each latent mediator, 2) each latent mediator to latent youth depressive outcome, and 3) parental depression to youth depressive outcome. First, we tested each mediator construct in a separate model. Next, we examined the effects of the hypothesized mediators on change in depression by entering youth depressive symptoms as a second, competing mediator.

Figure 1. Structural Equation Model Representation.

Note: Observed variables are depicted with boxes. Latent variables are depicted with ovals. Regression paths are depicted with single headed arrows. Direct effects are indicated with solid lines and indirect effects are indicated with dashed lines. SCID-NP = The Structured Clinical Interview for DSM-IV, Nonpatient version; K-SADS-PL = Kiddie Schedule for Affective Disorders and Schizophrenia, Present and Lifetime Version; CDI=Children’s Depression Inventory; IDAS-II= The Inventory of Depression and Anxiety Symptoms.

Covariances were included in the CFA and the SEM models to account for method variance shared by indicators. Specifically, we considered inclusion of correlated residuals for: a) measures from the same task (e.g., SRET positive and negative processing); b) questionnaires assessing the same construct completed by the same informant (e.g., child reports on AFARS and SNAP-Y NE); c) questionnaires assessing the same construct completed by different informants (.e.g., mother- and child-reports on the CDI); and d) questionnaires assessing different constructs completed by the same informant (e.g., mother reports on SNAP-Y NE and SCARED). We examined modification indices and included suggested covariance estimates if their presence led to a significant improvement in model fit, as indicated by chi-square difference tests. Finally, we included covariances between the hypothesized latent mediating factor and subthreshold depressive symptoms at age 12 in the models that included both latent mediators.

CFA and SEM model fit were evaluated using the Comparative Fit Index (CFI) and the Root Mean Square Error of Approximation (RMSEA). Using conventional guidelines, adequate model fit was indicated by a comparative fit index (CFI) ≥ .90 and root mean square error of approximation (RMSEA) ≤ .08; good model fit was indicated by CFI ≥ .95 and RMSEA ≤ .06 (Hu & Bentler, 1999). We conducted one thousand resamples to yield bootstrapped confidence intervals for all models. All indicators of latent factors except for neural processing were Z-scored to assist model scaling and convergence. ERP variables were not Z-scored because they had been calculated as residualized scores between task conditions (e.g., Doors RewP gain vs. loss residual) and were already on the scaling of the other indicators.

Results

Summary of Confirmatory Factor Analyses

A summary of factor loadings, model fit statistics, and covariance estimates (when applicable) for the latent mediator (age 12) and latent youth depression outcome (across ages 15 and 18) measurement models are provided in the Supplementary Material (see Supplementary Tables 3-12). The measurement models for disinhibition, neural processing of emotional stimuli, and depression, anxiety, and irritability symptoms were just-identified, therefore model fit could not be determined. For the over-identified measurement models (NE, PE, cognitive biases, interpersonal functioning, and depression outcome), all CFIs ≥ .947 and RMSEAs ≤ .068. All indicators in all measurement models loaded significantly on their latent factors.

Structural Models for Each Mediator

A series of separate structural models were estimated to examine the intergenerational transmission of depression via each hypothesized mediator construct (see Table 3). All 9 structural models exhibited at least adequate model fit (all CFIs ≥ .930 and all RMSEAs ≤ .082). Significant direct effects of number of depressed parents on the mediators were observed for youth NE, cognitive biases, subthreshold depressive symptoms, and anxiety and irritability symptoms (all βs ≧ .11, ps ≦ .022). The direct effects of mediators on youth depression outcomes were significant for all variables (all βs ≥ −.26, ps ≤ .008) except neural processing of emotional stimuli (β = −.12, p = .521). The number of parents with a lifetime history of depression did not have direct effects on youth depression after taking the mediators into account in any of the models (all βs ≤ .10, ps ≥ .064). Finally, significant indirect effects were observed for youth NE (β = .10, p = .021, R²=.22), cognitive biases (β = .12, p = .005, R²=.23), subthreshold depressive symptoms (β = .11, p = .018, R²=.38), and anxiety (β = .07, p = .006, R²=.12), and irritability (β = .08, p = .025, R²=.16) symptoms.

Table 3.

Separate Structural Equation Models Testing Mediation in the Intergenerational Transmission of Depression

Mediator	Path/Effect	β	SE(β)	p	RMSEA	CFI	R2 Youth Depression
Negative Emotionality	IV to MED	.21	.06	.004	0.074	0.930	.22
	MED to DV	.47	.10	≦ t.001
	IV to DV	.01	.06	.815
	Indirect Effect	.10	.04	.021
Positive Emotionality	IV to MED	-.07	.07	.391	0.059	0.949	.07
	MED to DV	-.26	.09	.008
	IV to DV	.09	.05	.124
	Indirect Effect	.02	.02	.459
Disinhibition	IV to MED	.08	.06	.182	0.069	0.950	.05
	MED to DV	.20	.07	.013
	IV to DV	.09	.06	.103
	Indirect Effect	.02	.01	.263
Cognitive Biases	IV to MED	.25	.66	.007	0.044	0.968	.23
	MED to DV	.48	.09	≦ .001
	IV to DV	-.02	.05	.709
	Indirect Effect	.12	.04	.005
Neural Processing of Emotional Stimuli	IV to MED	-.05	.09	.519	0.040	0.973	.03
	MED to DV	-.12	.01	.521
	IV to DV	.10	.05	.094
	Indirect Effect	.01	.02	.682
Subthreshold Depressive Symptoms	IV to MED	.17	.07	.009	0.082	0.930	.38
	MED to DV	.62	.08	≤ .001
	IV to DV	<.01	.05	.943
	Indirect Effect	.11	.04	.018
Anxiety Symptoms	IV to MED	.22	.05	≤ .001	0.055	0.963	.12
	MED to DV	.33	.08	≤ .001
	IV to DV	.05	.06	.425
	Indirect Effect	.07	.02	.006
Irritability	IV to MED	.19	.06	.022	0.068	0.943	.16
	MED to DV	.39	.10	.002
	IV to DV	.03	.06	.588
	Indirect Effect	.08	.03	.025
Interpersonal Functioning	IV to MED	.04	.06	.480	0.060	0.931	.14
	MED to DV	.36	.10	.002
	IV to DV	.10	.05	.064
	Indirect Effect	.02	.02	.517

Note: IV to MED = Path for number of depressed parents to hypothesized mediator; MED to DV= Path for mediator to youth depression; IV to DV = Path for parental depression to youth depression; Indirect Effect=Effect for number of depressed parents to youth depression via hypothesized mediator; β=standardized beta; p=unstandardized p-value; R² Youth Depression=Total variance explained in predicting youth depression across pathways tested; RMSEA = Root Mean Square Error of Approximation; CFI = Comparative Fit Index.

Mediation of Change in Depressive Symptoms from Early to Mid/Late Adolescence

A series of separate structural models were estimated to examine the uniqueness of each hypothesized mediator construct while considering the role of youth subthreshold depressive symptoms as a second mediator in the model in order to account for earlier depression (see Table 4 and Figure 2). All 9 structural models exhibited at least adequate model fit (all CFIs ≥ .901 and all RMSEAs ≤ .068). Similar to the previous models, significant direct effects of number of depressed parents on the mediators were observed for youth NE, cognitive biases², subthreshold depressive symptoms, and anxiety and irritability symptoms (all βs ≧ .19, ps ≦ .015). In contrast, the direct effects of mediators on youth depression outcomes were not significant for any mediators (βs ≦ .13, ps ≧ .247), except for disinhbition (β= −.16, p=.046). The number of parents with a lifetime history of depression did not have direct effects on youth depression (all βs ≦ .03, ps ≧ .801)³. The previously significant indirect effects for youth NE (β = .10, p = .021, R²=.22), cognitive biases (β = .12, p = .005, R²=.23), and both anxiety (β = .07, p = .006, R²=.12) and irritability symptoms attenuated to non-significance after including youth subthreshold depressive symptoms as a covariate (all βs ≦ .03, all ps ≧ .276, all R² ≥ .36).

Table 4.

Structural Equation Model Examining the Unique Effects of Mediators of the Intergenerational Transmission of Depression

Mediator	Path/Effect	β	SE(β)	p	RMSEA	CFI	R2 Youth Depression
Negative Emotionality	IV to MED	.19	.06	.004	0.062	0.953	.36
	IV to SX	.16	.06	.013
	MED to DV	.01	.16	.930
	SX to DV	.59	.16	≤.001
	IV to DV	.01	.05	.889
	Indirect Effect MED	.03	.02	.932
	Indirect Effect SX	.09	.05	.049
	MED with SX	.72	.06	≤.001
Positive Emotionality	IV to MED	-.06	.343	.380	0.063	0.943	.38
	IV to SX	.17	.06	.009
	MED to DV	.13	.10	.247
	SX to DV	.67	.11	≤.001
	IV to DV	≤.01	.05	.993
	Indirect Effect MED	-.01	.01	.525
	Indirect Effect SX	.12	.05	.021
	MED with SX	-.58	.06	≤.001
Disinhibition	IV to MED	.08	.06	.174	0.068	0.944	.39
	IV to SX	.17	.06	.008
	MED to DV	-.16	.08	.046
	SX to DV	.70	.10	≤.001
	IV to DV	≤.01	.05	.962
	Indirect Effect MED	-.01	.01	.276
	Indirect Effect SX	.12	.05	.012
	MED with SX	.58	.06	≤.001
Cognitive Biases	IV to MED	.24	.06	≤.001	0.059	0.938	.36
	IV to SX	.12	.06	.003
	MED to DV	-.06	.10	.576
	SX to DV	.65	.04	≤.001
	IV to DV	.03	.05	.402
	Indirect Effect MED	-.01	.03	.591
	Indirect Effect SX	.08	.04	.038
	MED with SX	.99	.02	≤.001
Neural Processing of Emotional Stimuli	IV to MED	.05	.09	.523	0.056	0.939	.39
	IV to SX	.16	.06	.010
	MED to DV	-.05	.09	.825
	SX to DV	.61	.08	≤.001
	IV to DV	≤.01	.05	.972
	Indirect Effect MED	≤.01	.01	.831
	Indirect Effect SX	.10	.04	≤.001
	MED with SX	-.11	.11	.269
Anxiety Symptoms	IV to MED	.21	.05	≤.001	0.067	0.942	.39
	IV to SX	.20	.06	.003
	MED to DV	.06	.08	.487
	SX to DV	.59	.11	≤.001
	IV to DV	.01	.05	.801
	Indirect Effect MED	.01	.02	.492
	Indirect Effect SX	.11	.05	.016
	MED with SX	.47	.09	.015
Irritability	IV to MED	.19	.06	.015	0.066	0.944	.36
	IV to SX	.17	.06	.009
	MED to DV	.06	.13	.649
	SX to DV	.56	.13	≤.001
	IV to DV	≤.01	.02	.975
	Indirect Effect MED	.01	.05	.656
	Indirect Effect SX	.10	,04	.039
	MED with SX	.69	.07	.010
Interpersonal Functioning	IV to MED	.06	.06	.345	0.067	0.907	.42
	IV to SX	.17	.06	.005
	MED to DV	-.15	.24	.565
	SX to DV	.74	.24	.004
	IV to DV	.01	.05	.912
	Indirect Effect MED	-.01	.02	.710
	Indirect Effect SX	.13	.06	.054
	MED with SX	.72	.06	≤.001

Note: IV to MED = Path for number of depressed parents to the mediator; IV to DEP SX = Path for number of parents depressed to youth subthreshold depressive symptoms; MED to DV= Path for mediator to youth depression; DEP SX to MED = Path for youth subthreshold depressive symptoms to youth depression; IV to DV = Path for number of depressed parents to youth depression; Indirect Effect MED= Effect for number of depressed parents to youth depression via hypothesized mediator; Indirect Effect DEP SX=Effect of number of depressed parents to youth depression via youth subthreshold depressive symptoms; MED with SX=Correlation between hypothesized mediator and youth subthreshold depressive symptoms; β=standardized beta; p=unstandardized p-value; R² Youth Depression=Total variance in predicting youth depression across pathways tested; RMSEA = Root Mean Square Error of Approximation; CFI = Comparative Fit Index.

Figure 2. Structural Equation Model Examining the Unique Effects of Latent Mediators in the Intergenerational Transmission of Depression.

Note: Observed variables are depicted with boxes. Latent variables are depicted with ovals. Regression paths are depicted with single headed arrows. Covariance estimates are depicted with curved, double-headed arrows. Direct effects and covariance estimates are indicated with solid lines and indirect effects are indicated with dashed lines. SCID=The Structured Clinical Interview for DSM-IV, Nonpatient version; K-SADS-PL Depression= Kiddie Schedule for Affective Disorders and Schizophrenia, Present and Lifetime Version; CDI=Children’s Depression Inventory; IDAS-II= The Inventory of Depression and Anxiety Symptoms.

Finally, similar to the independent model for youth subthreshold depressive symptoms, the number of parents with a lifetime history of depression significantly predicted youth subthreshold depressive symptoms in all models (all βs ≧ .12, ps ≦ .013), which, in turn, significantly predicted youth depressive outcomes in all models (all βs ≧ .56, ps ≦ .004). The indirect effect for youth subthreshold depressive symptoms was significant in all models (all βs ≧ .08, ps ≦ .039), except for interpersonal functioning, which approached significance (β=.13, p=.054).

Secondary Analyses

We re-ran the models for each mediator covarying for youth sex assigned at birth. None of the results changed substantively after controlling for youth sex. We also examined whether youth sex moderated the relationships between the number of parents with a lifetime history of depression and the mediators in our initial, separate models. Moderation effects were not observed for the indirect path in any of the models tested. Lastly, given the high comorbidity between depression and anxiety disorders, we included the number of parents with a lifetime history of anxiety disorder as a covariate in each initial separate model. In the individual models, indirect effects remained significant for cognitive biases and anxiety symptoms, but attenuated to trend levels for negative emotionality, subthreshold depressive symptoms, and irritability (see Supplementary Materials).

Discussion

Although the intergenerational transmission of depression is well documented, less is known about the processes involved. We used a longitudinal multi-method design to formally test putative mediators of the association between parent and offspring depression. In separate mediational models, we found evidence for indirect effects between lifetime history of depression in parents and subsequent depression in adolescent offspring via youth’s negative emotionality, cognitive biases, subthreshold depressive symptoms, anxiety, and irritability at age 12. However, when youth subthreshold depressive symptoms were included in models with the other candidate mediators in order to examine mediators of change in depressive symptoms from early to mid/late adolescence, only age 12 depressive symptoms exhibited significant unique indirect effects.

The results of the single mediator models are broadly consistent with and extend the limited existing literature. Allen et al. (2019) observed that higher neuroticism at age 5 mediated the association between depressive symptoms and offspring internalizing symptoms at age 9, although Mackin et al. (2022) failed to find evidence of personality mediating intergenerational transmission in an adolescent sample. Additionally, Dunning et al. (2021) observed that negative views of the self-mediated the association between parent and adolescent offspring depression across a 3-year window. We are unaware of any studies that formally tested whether subthreshold depressive symptoms, anxiety symptoms, or irritability mediate the intergenerational transmission of depression. However, Rice et al. (2017) found that anxiety and irritability, but not depression, symptoms independently predicted the first onset of depressive disorders in the offspring of depressed parents.

Despite observing significant indirect effects in separate models, NE, cognitive biases, anxiety, and irritability did not exhibit significant unique effects when we added youth subthreshold depressive symptoms as a second, competing mediator. Including subthreshold depressive symptoms tests whether the other mediators predict change in depression symptoms from early to mid/late adolescence. In addition, it conservatively guards against the possibility that concurrent depressive symptoms influence assessment of the other mediators (e.g., due to mood state effects). However, this introduces some interpretive complexities. First, it shifts the substantive question from whether the candidate mediators have indirect effects on later depression to whether they have indirect effects on change in depression. Second, these models may underestimate the effects of the other mediators, as age 12 depression likely absorbed the variance that subclinical depressive symptoms and the other putative mediators shared with age 15/18 depression. Third, as age 12 depression was measured concurrently with the putative mediators, it is conceivable that these models are misspecified (Wysocki, Lawson, & Rhemtulla, 2022) as we cannot exclude the possibility that the other candidate mediators influenced age 12 depressive symptoms. If that is the case, the correct model would be parental depression influencing the other mediators, which in turn influence age 12 depression, which influences age 15/18 depression. Unfortunately, we cannot test this alternative model as the other mediators and age 12 depression were assessed at the same time.

Surprisingly, we did not observe a significant direct effect of parental depression on youth depression at ages 15/18, although it was a trend level effect in bivariate analyses. Shrout and Bolger (2002) argue that mediation can be meaningfully interpreted even in the absence of a significant direct effect, particularly in the context of distal developmental processes. As they note, the effect of a distal predictor such as parental depression may be transmitted through multiple, temporally extended mechanisms, such that the direct path is attenuated over time. In our case, the emergence of significant indirect effects in the absence of a significant direct effect between parental history of depression and youth depression in mid/late adolescence underscores the relevance of early-emerging psychological vulnerabilities and symptoms in shaping trajectories of adolescent depression.

Comparing the individual models, the construct that accounted for the most variance was subthreshold depressive symptoms. Importantly, this was even after excluding offspring with a lifetime history of depressive disorders through age 12, which notably included D-NOS. This suggests that in prevention efforts designed to target mediators to interrupt the intergenerational transmission of depression, it might be sufficient to focus on subclinical depressive symptoms and multimodal interventions targeting multiple mediating processes may not be necessary. A caveat, however, is that unlike the indicators for any of the other mediating constructs, the CDI was used as indicators for both the subthreshold depression symptoms mediator and the depression outcome construct. Thus, shared method variance may have inflated the effects of age 12 subthreshold depression symptoms on outcomes.

Neural processing of emotional stimuli was the one construct that did not predict subsequent depression in this study (this was true at the indicator level as well; see Supplemental Materials). Other research, including work using this sample, has reported altered neural processing of reward cues and negatively valenced images in the offspring of depressed parents and in youth who subsequently develop depression (Burkhouse & Kujawa, 2023). A possible explanation is that these alterations are present only in subgroups of depressed and depression-prone youth (e.g., Kujawa, Proudfit, et al., 2014) and that they are moderated by other risk factors, such as life stress (see Goldstein et al., 2020; Kujawa et al., 2016).

We failed to find evidence that parental depression predicted offspring low PE, disinhibition, and poor interpersonal functioning, despite the fact that these constructs predicted later depression in offspring. Although some studies have reported links between maternal depression and low PE and impaired relational functioning (e.g., Goodman et al., 2011; Hammen, 2017), our findings suggest that, at least in this sample, these risk factors arise from influences that are distinct from parental depression.

The present study has several notable strengths. First, we focused on developmentally significant periods, assessing mediators just prior to the age when the incidence of depression begins to rise and examining outcomes during a critical window of heightened risk. Second, the use of multiple waves of data allowed for temporal disaggregation of parental depression, putative mediators, and offspring depression. Third, by limiting the sample to offspring with no history of diagnosable depression through age 12, we ensured that mediators were not influenced by prior clinically significant depressive symptoms. Fourth, we employed a multi-method, multi-informant approach to assess a broad range of cognitive, affective, social, and biological mediators. Fifth, we operationalized offspring depression using a latent factor that integrated both categorical diagnoses and dimensional symptom data. Sixth, we obtained diagnostic information from both mothers and fathers, the latter of whom are often underrepresented in the literature (Dachew et al., 2023), and this allowed us to consider the additive risk of having two depressed parents (Brennan et al., 2002). Seventh, we assessed parents at two separate time points, addressing the tendency of single assessments to underestimate depression prevalence (Olino et al., 2012).

Despite this study’s many strengths, it has several limitations that suggest directions for future research. First, we focused on offspring characteristics as mediators of intergenerational transmission and did not examine genetic or contextual pathways such as parenting practices or early life adversity, which also play key roles (Goodman, 2020; Lau & Eley, 2008; Loechner et al., 2018). Second, although the study spans a broad developmental period, we were limited in our ability to examine more fine-grained developmental changes. Different pathways of risk may unfold at distinct stages across childhood and adolescence (Goodman, 2020). Third, we did not assess whether the effects of parental depression varied based on the sex or primary caregiving status of the parent, nor did we distinguish whether parental depression occurred before or after the child’s birth. Fourth, while we accounted for parental anxiety in secondary analyses, we did not examine other forms of psychopathology or extend our outcome scope beyond depression, leaving the specificity of the findings unclear. Fifth, our follow-up was only through age 18. The peak period of risk for the onset of depression continues into early adulthood. Hence, some participants with low levels of depression will subsequently develop depressive disorders. A more extended follow-up would likely provide greater statistical power to detect effects. Finally, as with many longitudinal studies, we observed greater attrition among males and families with lower educational attainment. This may have introduced some bias, and our predominantly white, non-Hispanic, working- and middle-class sample may limit generalizability to more diverse populations.

In conclusion, we examined mediators of the intergenerational transmission of depression, focusing on personality/temperament, cognitive biases, neural processing of emotional stimuli, relevant symptoms, and interpersonal functioning in offspring with no prior history of diagnosable depression. Taken singly, we found significant indirect effects for NE, cognitive biases, subthreshold depression symptoms, anxiety, and irritability. However, when these constructs were modeled alongside age 12 subthreshold depressive symptoms, none emerged as unique mediators, while depressive symptoms exhibited significant unique indirect effects in all but the model with interpersonal functioning (where it was a trend). These findings suggest that the observed indirect effects of the first set of models reflect variance that is largely shared with depressive symptoms. This interpretation is consistent with structural models of psychopathology that emphasize shared variance among internalizing traits and symptoms (Zinbarg et al., 2016). However, our study design cannot disentangle the direction of the influences between depressive symptoms and the other putative mediators assessed at age 12. Future work using alternative model structures and finer temporal resolution may help disentangle the unique versus shared contributions of factors that mediate the intergenerational transmission of depression.