The Degree of Fluctuations in Maternal Depressive Symptom in Early Childhood is Associated with Children’s Depression Risk: Initial Evidence and Replication between Two Independent Samples

Abstract

Guided by life history theory, the present study examined whether the degree of fluctuations in maternal depressive symptoms in early childhood was prospectively linked to children’s risk for depression. This was the first study to present preliminary evidence on this topic and replicated main findings across two large, independent longitudinal samples. Study 1 included 1,364 families where maternal depressive symptoms were longitudinally assessed at child ages 1, 6, 15, 24, and 36, and 54 months, where child depressed/anxious behaviors at Grade 1 were reported. Study 2 included 1,292 families where maternal depressive symptoms were assessed at child ages 2, 6, 15, and 24 months. At 36 months, child internalizing symptoms and inhibitory control were assessed. In Study 1, findings revealed that the degree of fluctuations in maternal depressive symptoms over 54 months was associated with higher child depressed/anxious behaviors at Grade 1, only when mothers had higher but decreasing depressive symptoms. Study 2 revealed that the degree of fluctuations in maternal depressive symptoms over 24 months was related to higher child internalizing symptoms at 36 months, for mothers whose depressive symptoms were higher but decreasing, higher and increasing, and lower and decreasing. In addition, the degree of fluctuations in maternal depressive symptoms over 24 months was related to lower child inhibitory control at 36 months, for mothers who had higher but decreasing depressive symptoms. Findings highlighted the degree of fluctuations in maternal depressive symptoms during early childhood can contribute to environmental unpredictability, which can increase children’s depression risks.

Depression is the most significant mental health risk for women during the perinatal period (Bauman et al., 2020; Segre et al., 2007). One in every eight U.S. women develop clinically significant levels of postpartum depression and the symptoms can persist and fluctuate for years after giving birth (Bauman et al., 2020). Given its prevalence, postpartum depression is a significant issue as it can impact maternal parenting behaviors and, subsequently, increase children's risk of developing depression. Children of depressed mothers tend to show lower levels of emotion regulation and higher levels of behavioral problems, which in turn increase their risk of developing psychiatric disorders, such as depression (Feldman et al., 2009; Slomian et al., 2019; Walker et al., 2013).

Decades of research suggests a robust link between the severity of maternal depressive symptoms and children’s less optimal development and risks for future depression (Goodman et al., 2020; Sutherland et al., 2022; Walker et al., 2013). However, despite the episodic nature of depressive symptoms, seldom are maternal symptoms of depression studied in terms of its course, longitudinal changes, fluctuations, or variations over time. Yet, even sub-clinically elevated depressive symptoms in mothers interrupt normal caregiving behaviors and induce risk to children’s regulatory processes (Dix et al., 2014; Premo & Kiel, 2016; Wu et al., 2019). Guided by life history theory, this study presented preliminary evidence whether the degree of fluctuations in maternal depressive symptoms in early childhood was prospectively linked to children’s risk for depression and replicated main findings across two large, independent longitudinal samples.

Fluctuations in Maternal Depression – An Unpredictable Developmental Context

Life history theory explains how individuals strategize to allocate limited resources based on their life experiences (Stearns, 1992). Ellis et al. (2009) extended this discussion to focus on environmental harshness and unpredictability as key aspects based on which individuals make life history strategies that are adaptive to their unique ecological context. Evolutionally, harshness (the average rates of presence of adverse events) programs resource allocation towards faster life history strategies. In comparison, unpredictability (stochastic variations over time in life-history environmental conditions) increases difficulty of learning and adaptation to the environment, as the individual will be overburdened to make effective predictions about the environment and adjust their life strategies accordingly (Ellis et al., 2009; Young, 2020). Current evidence on environmental unpredictability has mostly focused on relatively severe forms of deprivation or threat, such as economic adversity, family violence, crime, and death, showing that an unpredictable early environment shapes children’s physical growth, cognitive competence, and stress regulation (e.g., Chang et al., 2019; Copping & Campbell, 2015; Lam et al., 2022; McCullough et al., 2013; Mittal et al., 2015; Li et al., 2023).

Although the harshness-unpredictability hypothesis has not been applied to maternal depression research, several indicators suggest that the area of maternal depression can be a good candidate to apply this theory to. First, maternal postpartum depression is related to infant mortality and morbidity (Slomian et al., 2019), a key factor based on which individuals make life history strategies. Second, maternal depression can contribute to both environmental threat (such as maternal abuse) and deprivation (such as maternal neglect), increasing early adversity (Choi et al., 2010). Finally, beyond the environmental harshness brought by maternal depression (e.g., as due to maladaptive parenting and family conflict), symptoms of depression can persist, fluctuate, exacerbate, and regress over the years, which provides a great source of environmental unpredictability. Current models on the intergenerational transmission of depression suggest a few pathways but focus on harshness such as heritability, neuroregulatory dysfunction, affect and behavioral learning, and environmental stress (e.g., Field, 2010; Goodman, 2007; Moses-Kolko et al., 2014). However, seldom do these models point to the possibility that environmental unpredictability due to fluctuations in maternal depressive symptoms constitutes an additional source of stress.

When depressive symptoms elevate, mothers can be unable to provide for their families (e.g., Wu et al., 2018), carry out day-to-day parenting activities (e.g., Feldman et al., 2009; McLearn et al., 2006), and be intrusive or unresponsive towards their children’s emotional needs (e.g., Kujawa et al., 2014; Wu et al., 2019). Studies also revealed inconsistency in depressed mothers’ parenting, such they may selectively respond to and over-exaggerate children’s high-intensity emotions and ignore the less salient cues (e.g., Dix et al., 2014; Premo & Kiel, 2016; Morgan et al., 2020; Wu, 2021). The variations in depressed parenting are additionally supported by the heterogeneity within depressed mothers: some enduring chronic stress and undergoing persistent levels of depression, flat affect, and withdrawal behaviors; some encountering acute symptoms and experiencing a surge in negativity and harsh parenting; some having a mixture of both; yet some showing adaptive parenting behaviors at times (e.g., Field, 2010; Hooper et al., 2015; Wang & Dix, 2013).

Given these findings, intriguing questions remain in relation to how researchers currently quantify maternal depressive symptoms (mostly in terms of harshness, as discussed above) and how this affects our understanding. In particular, two commonly assessed factors of depressive symptoms indicating harshness are the mean (or the severity level over time; e.g., Gartstein et al., 2010; Gross et al., 2009; Premo & Kiel, 2016) and the slope (or the progression, the rate of linear change; e.g., Hughes et al., 2013; Wu, 2021). However, no known study to date has examined the additional dimension in maternal depressive symptoms, i.e., symptom fluctuations, which theoretically can interfere mothers’ normal caregiving behaviors, disrupt children’s stress regulation systems, interrupt children’s development of self-regulation, thus contributing to the risk of depression (e.g., Feldman et al., 2009; Gross et al., 2009; Wu et al., 2019).

Supporting this, recent studies observed higher internalizing behaviors among school-aged children (Yan et al., 2021) and less successful emotion regulation among infants and toddlers (Wu & Gazelle, 2021), when maternal depressive symptoms were elevated at a certain time point (beyond one’s average levels over time). Whereas these two studies revealed time-specific associations, such that one-time elevation in maternal depressive symptoms would be related to higher children’s risks immediately after, important questions remain - whether a mother’s general tendency to experience fluctuations in depressive symptoms across time, in comparison to their peers who may not experience such fluctuations, can contribute to additional risks towards their children beyond their average depression severity or progression (or the rate of change).

Another important yet under-studied question in the harshness-unpredictability framework is the harshness-by-unpredictability interaction. Only one study examined this interaction and found that income unpredictability was a risk factor when income harshness was low (Li et al., 2018). It is unclear whether the same moderation pattern applies to maternal depressive symptoms. Theoretically, severe and progressing maternal depression can increase environmental harshness, in which accurate learning and prediction of the environment become more vital (Young et al., 2020). It will thus be interesting to investigate whether the severity and progression of depression can exacerbate the effects of maternal symptom fluctuations on children’s depression risks.

Children’s Development of Depression Risk.

Life history theory conceptualizes that fluctuations in maternal postpartum depressive symptoms constitute an unpredictable early environment. The classic learned helplessness theory of depression (Maier & Seligman, 1976) would suggest that when an individual have learnt that the environment is unpredictable and uncontrollable, depression-like symptoms can occur. In this way, fluctuations in maternal depressive symptoms can cascade into children’s future depression risk. Yet this effect may manifest differently at different developmental stages. In school-aged children, one of the most significant indicators lies in behavioral problems, especially DSM-oriented depressed/anxious behaviors, given the comorbidity in this age group (Beauchaine & Hinshaw, 2020). For preschoolers, a more common indicator is internalizing symptoms (e.g., Luby et al., 2004), as behavioral problems are less differentiated (as separate elements of depression, anxiety, and somatic symptoms) at this age. Past research supported the examination of these indicators by revealing that maternal depressive symptoms are related to reduced emotion regulation, higher negative affect, and higher behavioral problems among preschoolers and depressive symptoms among school-aged children (e.g., Feldman et al., 2009; Gross et al., 2008; Slomian et al., 2019; Yan et al., 2021). One additional indicator can be preschooler’s self-regulatory abilities, such as inhibitory control (i.e., being able to suppress a dominant response to allow for a subdominant response), a core capacity shielding children from developing problem behaviors and depressive symptoms (Hermansen et al., 2022). In fact, literature has revealed that an unpredictable early environment can be related to impaired inhibitory control among offsprings (e.g., Davis et al., 2019; Mittal et al., 2015; Sturge-Apple et al., 2016), however this body of literature has not yet been extended to the examination in unpredictability within maternal depressive symptoms.

The Present Study

Guided by life history theory (Ellis et al., 2009; Young, 2020), the current study is the first one to empirically test whether the degree of fluctuations in maternal depressive symptoms during early childhood contributed to additional depression risks among children. This study used two independent longitudinal samples and focused on children transitioning into the school age (Study 1) and into the preschool age (Study 2). This study focused on whether fluctuations in maternal depressive symptoms conferred additional risks to children beyond two commonly assessed factors: the mean and the slope of depressive symptoms. Additionally, this study examined whether the mean and the slope of maternal depressive symptoms moderated the link between the fluctuation in maternal depressive symptoms and child depression risks. Child depression risks were operationalized as higher depressed/anxious behaviors among school-aged children, and higher internalizing symptoms and lower inhibitory control among preschoolers.

Although the multi-study design is a strength of the current study, it is noteworthy that the rates of maternal depression and the effects of maternal depression can vary according to important sample characteristics, such as race and SES. For example, U.S. Black mothers were more likely to endorse perinatal depressed mood compared to U.S. White mothers (Segre et al., 2006). This difference could be attributed to factors such as SES, racial discrimination, experienced adversities, and neighborhood risk levels, as suggested by various studies (Ertel et al., 2011; Mukherjee et al., 2016; Liu & Tronick, 2014; Pachter et al., 2006). In terms of the effects of maternal depression, a U.S. study found that postpartum depression predicted lower maternal sensitivity only in the context of high SES risk, with no significant observed differences by maternal race (White or Black; Norcross et al., 2020). Additionally, Pachter et al. (2006) observed that maternal depression had an impact on child behavioral problems during middle childhood through parenting practices, but this effect was noted only among U.S. White and Hispanic mothers, not among Black mothers. It was speculated that sharing childrearing responsibilities with extended family members served as a buffer against maternal depression among U.S. Black families. Together, these studies suggest possible differences in associations and transmission mechanisms of maternal depression to children by demographic factors such as SES levels and race.

Figure 1a presents the conceptual model of the current study. It was hypothesized that the degree of fluctuations in maternal depressive symptoms, in addition to the severity (the mean) or the progression (the slope), would contribute to depression risks among children. In addition, a higher mean level and an increase in depressive symptoms over time, would exacerbate the effect of fluctuations in maternal postpartum depressive symptoms on children’s depression risks.

Figure 1.

The Conceptual and Statistical Models.

(a) The Conceptual Model.

(b) The Statistical Model of Study 2. Standardized Coefficients are Shown. Covariates are Omitted for Simplicity.

Study 1

Participants

Study 1 examined fluctuations in maternal depressive symptoms across early childhood (54 months or the first 4.5 years) and the effect on children at the school-entry age. This study employed data from the National Institute of Child Health and Human Development (NICHD) Study of Early Child Care and Youth Development (SECCYD; N = 1,364; 50.4% boys). The study's participants were infants who were born in 1991 from ten U.S. states, and their mothers were over 18 years old, healthy, spoke English as their primary language, did not intend to put their child up for adoption, lived within an hour of the research site, and had no plans to move within the next 3 years (NICHD ECCRN, 2002). The average age of the mothers was 28.11 years (SD = 5.6). The majority (78.9%) of the mothers were White, 11.2% Black, 5.6% Hispanic, 4.3% other. Most of the mothers (85.2%) were living with their partner at study enrollment. The average length of maternal education was 14.4 years (SD = 2.5). The average family income-to-needs ratio was 3.7 (SD = 2.8).

Procedures

Mothers provided consent and parental permission for data collection at study enrollment. At 1, 6, 15, 24, and 36, and 54 months, mothers self-rated their depressive symptoms. At 54 months and in Grade 1, mothers rated children’s depressed/anxious behaviors. Florida State University Institutional Review Board provided ethical approval.

Measures

Maternal depressive symptoms were self-rated using the Center for Epidemiological Study-Depression Inventory (CES-D; Radloff, 1977) for experienced symptoms during the past week, on a 4-point scale (0 = less than 1 day, 3 = 5–7 days). The total score from the 20 items were used. The CES-D is a validated tool to be used in non-clinical populations. Reliability and validation studies have indicated the CES-D to be internally consistent (α > .90) and moderately stable over several months (rs > .50; Vilagut et al. 2016; Yang et al. 2017). In the current study, the associations of the CES-D between neighboring time points were above .52. The Cronbach's α ranged from .85 to .90 in the current sample.

Child depressed/anxious behaviors were reported by mothers using DSM-oriented anxious/depressed behaviors in the Child Behaviors Checklist (CBCL; ages 4-18; Achenbach, 1991). Mothers rated children’s behaviors on a 3-point scale (0 = not true, 2 = very true). A higher score indicated more severe symptoms. The scale showed a satisfactory internal consistency (.74 and .74 for 54 months and Grade 1, respectively).

Covariates included child sex, race, and birth order; maternal marital status, non-marital cohabitation, employment, age, years of education, and family income-to-needs ratio, reported by mothers at 1-month postpartum. Child negativity was reported by mothers using the Early Infant Temperament Questionnaire (EITQ; Medoff-Cooper et al., 1993) at 6 months old (Cronbach's α = .81). A higher score indicated more temperamental negativity.

Data Preparation and Analytic Strategies

Little’s MCAR test (Little, 1988) revealed that data were not missing completely at random, χ²(1195) = 1889.99, p < .001. Missingness in maternal depressive symptoms and child depressed/anxious behaviors were related to being racially minority, not married, not cohabiting, not employed, mother younger in age, and having a lower income-to-needs ratio, ts ≥ 2.0, ps ≤ .05. As missingness of the key study variables can be explained by covariates, full information likelihood (FIML) estimation with robust standard errors (MLR) was used to estimate for the missing information (Enders & Bandalos, 2001). Analysis was conducted using the lavaan package (Rosseel, 2012) in R.

Considering that prior research has generally overlooked symptom fluctuation as a third factor, beyond the severity and progression of maternal depressive symptoms, that impacts children, an analytic approach that extracts the stochastic variations in symptoms was applied. This approach was chosen over other person-centered longitudinal methods, such as latent growth modeling and latent growth mixture modeling, because it mathematically separates within-individual variability from the individual mean and slope. First, following the procedures by Wang et al. (2012), a longitudinal multilevel model (with an intercept, a slope on time, and random effects) was estimated with FIML in Mplus program, through which the linear rate of change (i.e., the slope) in maternal depressive symptoms over early childhood (from 1 to 54 months) for each mother was extracted and saved. Then the detrended mean (referred to as the “mean” hereafter) was calculated, by averaging the detrended data, which involved subtracting the rate of change multiplied by time from depressive symptom scores at each time point from 1 to 54 months for each mother. This method provides a “true” mean score, eliminating the impact of linear changes over time. The equation is as follows. ${\mathrm{X}}_i$ indicates the depressive symptom scores at time point $i$ (from 1 to 54 months, for a maximum of six time points) for each mother; ${\mathrm{s}}_0$ indicates the slope for each mother derived from Mplus; ${\mathrm{X}}_{dei}$ indicates the detrended depressive symptom scores at time point $i$; $n$ indicates the number of available time points (adjusted for each mother due to missing measurement points); and ${\mathrm{X}}_{mean}$ indicates the detrended mean for each mother.

$$X_{de\_i}=X_i-S_0\ast i$$

$$X_{mean}=\frac{{\sum }_{i=1}^n{X}_{de\_i}}{n}$$

Next, an intraindividual standard deviation ($ISD$) was calculated with detrended depressive symptom data from 1 to 54 months for each mother through the following equation, to indicate the extent of variation or fluctuation in depressive symptoms over time beyond the mean and the slope.

$$ISD=\sqrt{\frac{{\sum }_{i=1}^n{(X_{de\_i}-X_{mean})}^2}{n-1}}$$

The three variables (i.e., the mean, the slope, and the $ISD$) of maternal depressive symptoms are mathematically independent, and were used as predictors towards child depressed/anxious behaviors in Grade 1, whereas the same variable at 54 months was controlled for along with other covariates, in a multiple regression model. In probing the interactions, simple slope tests were conducted at lower and higher levels of the moderating variables (i.e., the mean and the slope of depressive symptoms), centered at one standard deviation ($SD$) below the average and above the average, respectively. That is, simple slope tests were conducted at a combination of the depressive symptoms centered at lower mean (the average mean − 1 $SD$) or higher mean (the average mean + 1 $SD$), and fast declining (the average slope − 1 $SD$) or fast increasing (the average slope + 1 $SD$). Of note, the higher mean depressive symptoms were cantered at 16.99, and a score of 16 and above on the CES-D can be deemed clinically depressed. These groups were named “higher” and “lower” as a relative level in the community-based sample. For the ease of interpretation, the combinations of lower and higher levels of the mean and the slope were labeled as healthy (lower mean, fast declining), recovering (higher mean, fast declining), impending (lower mean, fast increasing), and deteriorating (higher mean, fast increasing).

Results

Descriptive statistics of variables are presented in Table 1. The regression model (Table 3) suggested that beyond the covariates, the mean of maternal depressive symptoms was associated with higher child depressed/anxious behaviors in Grade 1 (B = 0.09, SE = 0.03, t = 2.93, p = .003). Additionally, an interaction between the slope and the ISV of maternal depressive symptoms was found (B = −0.22, SE = 0.07, t = 3.00, p = .003). Only mothers who were recovering (i.e., with higher mean and fast declining depressive symptoms), the ISV of maternal depressive symptoms was related to higher child depressed/anxious behaviors (B = 0.19, SE = 0.09, t = 2.24, p = .03; Figure 2a).

Table 1.

Descriptive Statistics and Bivariate Correlation for Study 1 Variables.

		1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21
1.	Child Sex	--
2.	Child Race	−.02	--
3.	Child Birth Order	−.02	−.01	--
4.	Married	−.02	.41**	−.09**	--
5.	Cohabitation	−.01	.35**	−.12**	.74**	--
6.	Employment	−.02	.11**	.16**	.20**	.19**	--
7.	Mother's Age	.03	.26**	−.23**	.43**	.34**	.23**	--
8.	Mother's Education	.04	.20**	.05	.40**	.31**	.30**	.55**	--
9.	Income-to-Needs Ratio	.01	.20**	.08**	.34**	.28**	.13**	.39**	.42**	--
10.	Child Negativity	.05	−.19**	.00	−.17**	−.11**	−.10**	−.18**	−.14**	−.14**	--
11.	Depression 1M	−.04	−.16**	.01	−.21**	−.16**	−.15**	−.21**	−.23**	−.17**	.23**	--
12.	Depression 6M	.03	−.17**	−.03	−.24**	−.18**	−.16**	−.18**	−.21**	−.17**	.22**	.52**	--
13.	Depression 15 M	.01	−.13**	−.01	−.19**	−.16**	−.16**	−.19**	−.25**	−.15**	.15**	.45**	.58**	--
14.	Depression 24 M	.01	−.17**	.02	−.27**	−.25**	−.10**	−.23**	−.28**	−.20**	.16**	.41**	.52**	.53**	--
15.	Depression 36 M	.04	−.10**	.02	−.21**	−.19**	−.11**	−.20**	−.24**	−.15**	.17**	.45**	.47**	.50**	.54**	--
16.	Depression 54 M	−.04	−.17**	−.01	−.28**	−.21**	−.10**	−.21**	−.26**	−.17**	.15**	.42**	.42**	.39**	.50**	.52**	--
17.	Depression Mean	.00	−.20**	.00	−.30**	−.24**	−.18**	−.27**	−.32**	−.23**	.25**	.79**	.82**	.78**	.76**	.73**	.64**	--
18.	Depression Slope	.00	.05	.01	.00	−.01	.08**	.02	.01	.02	−.09**	−.48**	−.37**	−.23**	.03	.22**	.51**	−.26**	--
19.	Depression ISV	−.03	−.12**	−.03	−.19**	−.18**	−.13**	−.16**	−.24**	−.16**	.12**	.55**	.41**	.46**	.48**	.51**	.51**	.62**	−.04	--
20.	Anxious Depression 54M	.00	−.08*	.11**	−.10**	−.07*	−.05	−.09**	−.09**	−.02	.12**	.18**	.20**	.21**	.20**	.25**	.19**	.27**	.01	.12**	--
21.	Anxious Depression G1	−.02	.00	.08**	−.03	.01	−.01	−.09**	−.04	.03	.10**	.17**	.16**	.18**	.19**	.22**	.18**	.23**	.02	.15**	.47**	--
N		1364	1364	1364	1362	1362	1364	1364	1363	1273	1279	1363	1278	1241	1119	1202	1077	1363	1363	1304	1061	1028
Minimum		1.00	0.00	0.00	0.00	0.00	0.00	18.00	7.00	0.08	1.54	0.00	0.00	0.00	0.00	0.00	0.00	−0.07	−2.87	0.03	50.00	50.00
Maximum		2.00	1.00	1.00	1.00	1.00	1.00	46.00	21.00	25.08	4.72	53.00	52.00	54.00	51.00	57.00	55.00	40.25	2.68	23.33	81.00	81.00
Mean		--	--	--	--	--	--	28.11	14.23	2.86	3.18	11.36	8.97	9.05	9.40	9.22	9.83	10.13	−0.16	4.86	52.09	52.76
$SD$		--	--	--	--	--	--	5.63	2.51	2.61	0.40	9.02	8.34	8.18	8.63	8.31	8.70	6.86	0.45	3.20	4.27	4.83

Note. Child sex: 1 = male, 2 = female. Child race: 1 = White, 0 = other. Birth order: 1 = first born, 0 = not first born. Mother married: 1 = married, 0 = not married. Cohabitation: 1 = cohabitating, 0 = not cohabitating. Employment: 1 = employed, 0 = not employed. 1- 54 M = 1- 54 months. G1 = Grade 1.

^*p < .05

^**p < .01.

Table 3.

Model Results.

	Study 1				Study 2
	Child Anxious Depression G1 R2 = .27				Internalizing Symptoms 36 M R2 = .18					Inhibitory Control 36 M R2 = .04
	B	SE	t	β	B	SE	t	β		B	SE	t	β
Child Sex	−0.16	0.26	−0.63	−.02	0.10	0.16	0.59	.02		−0.04	0.07	−0.52	−.02
Child Race	0.74	0.40	1.83†	.06	−0.31	0.22	−1.45	−.05		0.01	0.10	0.11	.01
Child Birth Order	0.35	0.29	1.22	.04	−0.07	0.19	−0.38	−.01		0.17	0.08	2.02*	.10
Married	−0.09	0.58	−0.16	−.01	−0.33	0.26	−1.27	−.06		0.10	0.11	0.95	.06
Cohabitation	0.99	0.68	1.46	.07	−0.34	0.27	−1.27	−.05		0.11	0.13	0.85	.05
Employment	0.01	0.28	0.04	.00	−0.01	0.17	−0.08	.00		0.06	0.08	0.77	.04
Mother's Age	−0.05	0.03	−1.52	−.06	0.00	0.02	−0.17	−.01		0.00	0.01	−0.62	−.03
Mother's Education	0.09	0.07	1.31	.04	−0.14	0.04	−3.61***	−.14		0.01	0.02	0.36	.02
Income-to-Needs Ratio	0.09	0.07	1.36	.05	−0.04	0.05	−0.74	−.02		0.02	0.03	0.58	.04
Child Negativity	0.52	0.35	1.50	.04	0.60	0.13	4.54***	.15		0.01	0.06	0.12	.01
Anxious Depression 54 M	0.49	0.05	10.15***	.43	--	--	--	--	--	--	--	--	--
Depression Mean	0.09	0.03	2.93**	.13	−0.14	0.08	−1.84	−.12		0.07	0.04	1.76†	.24
Depression Slope	1.34	0.40	3.35***	.12	1.44	0.38	3.76***	.26		−0.26	0.17	−1.51	−.17
Depression ISV	0.05	0.06	0.86	.03	0.26	0.11	2.39*	.15		−0.02	0.04	−0.38	−.03
Mean * ISV	0.01	0.01	0.78	.03	0.04	0.02	2.22*	.17		−0.01	0.01	−1.77†	−.22
Slope * ISV	−0.22	0.07	−3.00**	−.11	−0.27	0.07	−3.69***	−.24		0.08	0.03	2.31*	.25

Note. Child sex: 1 = male, 2 = female. Child race: 1 = White, 0 = other. Birth order: 1 = first born, 0 = not first born. Mother married: 1 = married, 0 = not married. Cohabitation: 1 = cohabitating, 0 = not cohabitating. Employment: 1 = employed, 0 = not employed. 36 M = 36 months. 54 M = 54 months. G1 = Grade 1.

^†p < .10

^*p < .05

^**p < .01

^***p < .001.

Figure 2.

Interaction plots between maternal depressive symptoms and children’s depression risk.

(a) Study 1: Maternal depression of 1-54 months predicting child depressed/anxious behaviors in Grade 1.

(b) Study 2: Maternal depression of 2-24 months predicting child internalizing symptoms at 36 months.

(c) Study 2: Maternal depression of 2-24 months predicting child inhibitory control at 36 months.

Study 2

Participants

Study 2 examined maternal depressive symptoms across the first 2 years postpartum and three-years-olds’ development among low-income, rural families. Data included 1,292 children (50.9% boys) with their families from the Family Life Project (FLP; Vernon-Feagans & Cox, 2013). The families were recruited from hospitals in U.S. areas with high poverty rates, oversampling low-income and Black families. Mothers’ mean age was 25.9 years ($SD$ = 6.0) at enrollment. Mothers were 58.8% White, 40.7% Black, and 0.5% other races. Half of the mothers (48.6%) were married at enrollment, and 40.9% of the mothers were employed. Mothers had less than a high school education (19.8%), a GED/high school certificate (39.2%), some college education (27.2%), and a college degree and above (13.7%).

Procedures

Mothers provided consent and parental permission for data collection at study enrollment. Mothers reported their depressive symptoms when children were 2-, 6-, 15-, and 24-month-old. Mothers rated children’s internalizing symptoms, and children’s inhibitory control was measured during the 36-month visit. Florida State University Institutional Review Board provided ethical approval.

Measures

Maternal depressive symptoms were self-rated on the Brief Symptom Inventory-18 (BSI-18; Derogatis, 2000) when children were 2-, 6-, 15-, and 24-month-old. The BSI includes 6 items measuring depressive symptoms experienced in the past week, e.g., feelings of “blue” and “worthlessness,” rated on a 5-point scale (0 = not at all, 4 = very much). A sum score was generated with higher scores indicating higher depressive symptoms. Cronbach’s α was between .80 and .86 over time. The longitudinal measurement invariance tests of the BSI-18 depression subscale are included in the supplemental material. The measurement of maternal depressive symptoms over time met the requirements of scalar invariance (Putnick & Bornstein, 2016).

Child internalizing symptoms was rated by mothers using the internalizing subscale on the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997). A sample item was “often unhappy, depressed, or tearful,” rated on a 3-point Likert scale (0 = not true, 2 = certainly true). A total score was used, with higher scores indicating higher internalizing symptoms (Cronbach's α = .82).

Child inhibitory control was assessed with the Silly Sounds Stroop task (17 items; Gerstadt et al., 1994). Children were asked to meow when seeing a dog picture and bark when seeing a cat picture. Children’s response accuracy was used to generate an expected a posteriori (EAP) score, with higher scores indicating better inhibitory control (Willoughby et al., 2010).

Covariates included child sex, race, and birth order; maternal marital status, non-marital cohabitation, employment, age, years of education, and family income-to-needs ratio, reported by mothers at 2-month postpartum. Child negativity was reported by mothers using three subscales on the Infant Behavior Questionnaire (IBQ; Rothbart, 1981): distress to limitation (16 items), distress to novelty (16 items), and recovery from distress (reversed, 4 items). A sample item included “distress at stranger approach,” rated on a 7-point scale (1 = never, 7 = always). Mean scores of the three subscales were standardized and averaged to generate a score of child negative reactivity (Cronbach's α = .85).

Data Preparation and Analytic Strategies

Analysis was conducted using the lavaan package (Rosseel, 2012) in R using a path model. The mean, the slope, and the ISV of maternal depressive symptoms across four time points (at 2, 6, 15, and 24 months) were calculated. Little’s MCAR test (Little, 1988) revealed that data were not missing completely at random, χ²(495) = 569.23, p = .01. Employed mothers were more likely to miss assessments in maternal depressive symptoms, ts ≥ 2.2, ps ≤ .03. Missingness in child internalizing symptoms and inhibitory control were related to being a boy, mother having low education, and mother having elevated depressive symptoms ts ≥ 2.1, ps ≤ .03. Missingness in child inhibitory control were also related to being racially minority, not married, cohabiting, not employed, mother younger in age, and having a lower income-to-needs ratio, ts ≥ 2.8, ps ≤ .01. FIML estimation with MLR was used to handle missingness (Enders & Bandalos, 2001). The model fit was evaluated by the root mean squared error of approximation (RMSEA, < .05), the comparative fit index (CFI, > .95) and the standardized root mean squared residual (SRMR, < .05).

Results

Descriptive statistics of variables are presented in Table 2. The path model (Table 3 and Figure 1b) was saturated, thus showing perfect model fit, χ²(0) = 0.00, CFI = 1.00, SRMR = .00, RMSEA = .00 (CI_90% [.00, .00]). In addition to the covariates, the slope (B = 1.44, SE = 0.38, t = 3.76, p < .001) and the ISV (B = 0.26, SE = 0.11, t = 2.39, p = .02) of maternal depressive symptoms were related to higher child internalizing symptoms. Additionally, interactions between the mean and the ISV (B = 0.04, SE = 0.02, t = 2.22, p = .03), and between the slope and the ISV (B = −0.27, SE = 0.07, t = −3.69, p < .001) of maternal depressive symptoms were observed. As shown in Figure 2b, the ISV of maternal depressive symptoms was related to higher child internalizing symptoms, for mothers who were recovering (B = 0.53, SE = 0.11, t = 4.90, p < .001), deteriorating (B = 0.26, SE = 0.08, t = 3.06, p = .002), and healthy (B = 0.33, SE = 0.12, t = 2.78, p = .005), with the largest coefficients when mothers were recovering.

Table 2.

Descriptive Statistics and Bivariate Correlation for Study 2 Variables.

	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19
1. Child Sex	--
2. Child Race	−.02	--
3. Child Birth Order	−.01	.04	--
4. Married	.01	.37**	−.24**	--
5. Cohabitation	−.01	.05	.08**	−.45**	--
6. Employment	.01	.04	.00	.10**	−.02	--
7. Mother's Age	−.01	.19**	−.33**	.48**	−.14**	.10**	--
8. Mother's Education	−.01	.22**	−.07*	.43**	−.14**	.26**	.46**	--
9. Income-to-Needs Ratio	−.05	.38**	.09**	.45**	−.11**	.28**	.35**	.52**	--
10.Child Negativity	.07*	−.38**	−.01	−.22**	.00	−.07*	−.20**	−.28**	−.29**	--
11.Depression 2M	.01	−.07*	.01	−.09**	.03	−.08**	−.10**	−.18**	−.15**	.14**	--
12.Depression 6M	.02	−.06*	−.03	−.09**	.04	−.09**	−.09**	−.13**	−.14**	.17**	.51**	--
13.Depression 15 M	.02	−.13**	−.02	−.11**	.04	−.08**	−.12**	−.18**	−.16**	.20**	.50**	.54**	--
14.Depression 24 M	.07*	−.06*	−.01	−.09**	.04	−.08*	−.08**	−.17**	−.14**	.12**	.36**	.43**	.49**	--
15.Depression Mean	.03	−.10**	−.02	−.11**	.04	−.10**	−.12**	−.21**	−.19**	.20**	.83**	.82**	.80**	.65**	--
16.Depression Slope	.05	−.07*	−.02	−.07*	.03	−.05	−.07*	−.12**	−.11**	.10**	.03	.25**	.57**	.90**	.43**	--
17.Depression ISV	.05	−.13**	.00	−.17**	.11**	−.10**	−.17**	−.25**	−.20**	.20**	.54**	.58**	.61**	.54**	.72**	.42**	--
18.Internalizing Problems 3Y	.04	−.21**	.00	−.20**	.03	−.08**	−.16**	−.28**	−.24**	.29**	.17**	.19**	.27**	.18**	.25**	.18**	.29**	--
19.Inhibitory Control 3Y	−.03	.05	.10*	.03	.02	.05	−.02	.05	.09	−.02	.02	.01	−.02	.04	.01	.02	−.01	−.01	--
N	1292	1292	1292	1292	1292	1292	1292	1292	1204	1186	1292	1191	1166	1124	1292	1292	1231	1093	479
Minimum	1.00	0.00	0.00	0.00	0.00	0.00	14.52	7.00	0.00	−1.99	0.00	0.00	0.00	0.00	−0.03	−1.99	0.01	0.00	−1.98
Maximum	2.00	1.00	1.00	1.00	1.00	1.00	69.52	21.00	16.49	2.95	22.00	20.00	24.00	24.00	17.00	3.48	10.32	16.00	1.37
Mean	--	--	--	--	--	--	25.91	14.42	1.81	0.00	2.08	2.29	2.71	2.42	2.16	0.20	1.66	3.89	−0.51
$SD$	--	--	--	--	--	--	6.03	2.82	1.68	0.74	3.15	3.31	3.86	3.81	2.54	0.52	1.66	2.94	0.80

Note. Child sex: 1 = male, 2 = female. Child race: 1 = White, 0 = other. Birth order: 1 = first born, 0 = not first born. Mother married: 1 = married, 0 = not married. Cohabitation: 1 = cohabitating, 0 = not cohabitating. Employment: 1 = employed, 0 = not employed. 2-24 M = 2-24 months.

^*p < .05

^**p < .01.

As for child inhibitory control, an interaction between the slope and the ISV of maternal depressive symptoms was found (B = 0.08, SE = 0.03, t = 2.31, p = .02). Only for mothers who were recovering, the ISV of maternal depressive symptoms was related to lower child inhibitory control (B = −0.11, SE = 0.05, t = −2.00, p = .045; Figure 2c). This pattern mirrored that in Study 1 by revealing the risk of symptom fluctuations for mothers having recovering depression.

Discussion

This study was the first to empirically test whether the degree of fluctuations in maternal depressive symptoms over early childhood was related to young children’s depression risks. Using two independent longitudinal samples, findings replicated each other in that a higher degree of fluctuations in maternal depressive symptoms conferred the highest risks when mothers were recovering from relatively higher levels of depressive symptoms. Findings highlighted the risks of fluctuations in maternal depressive symptoms in consideration of the progression and severity levels of maternal depressive symptoms.

Study 1 demonstrated that the degree of fluctuations in maternal depressive symptoms over the first 4.5 years of children’s life was related to children’s depressive/anxious behaviors at Grade 1 (or ages 6-7), when mothers were recovering from relatively higher levels of depression (i.e., higher mean and fast declining depressive symptoms). It is likely that when mothers are recovering from depressive symptoms, any fluctuations can be perceived as signs to degenerate, thus being threatening to young children’s well-being. This effect supports the harshness-by-unpredictability interaction. This pattern does not fully align with previous findings where unpredictability in income conferred higher risks under lower mean harshness but lower risks under higher mean harshness (Li et al., 2018). It is likely not only the mean harshness matters, how it progresses over time is also important. Differences in findings may also be explained by the operation of adversity and the feature of maternal depression, such as unpredictability in maternal depressive symptoms can contribute to learned helplessness and interrupt children’s socioemotional learning (Maier & Seligman, 1976), and this effect may be especially salient when there are signs of the hope of recovery yet setbacks of symptom fluctuations. Notably, this effect was observed beyond the previous level of children’s anxious depressive behaviors at age 4.5, revealing that this effect can be long-lasting, beyond the concurrent effects.

Study 2 revealed that the degree of fluctuations in maternal depressive symptoms over the first 2 years of children’s life was related to higher children’s internalizing symptoms at age 3. This effect was most significant when mothers were recovering (higher mean and fast declining depressive symptoms), and also deteriorating (higher mean and fast increasing symptoms), and healthy (lower mean and fast declining symptoms). The sharpest slope found when the mothers were recovering mimicked the pattern in Study 1. It seems that the only condition where maternal depressive symptoms were not a contributing factor was when mothers were impending (lower mean and fast increasing symptoms), possibly because in this scenario, the fast increase in depressive symptoms is a more powerful predictor (as evidenced by the strong connection between the slope of depressive symptoms and child internalizing symptoms), over and above the fluctuations. In other scenarios, even labeled as “healthy” (with depressive symptoms lower and reducing than average levels), maternal symptom fluctuations still conferred additional risks to preschoolers’ internalizing symptoms, demonstrating that the unpredictability of maternal mood potentially contributes to stress in the rearing environment of children, thereby linking to children’s behavioral indicators of sadness and withdrawal.

The links between maternal symptom fluctuations and children’s internalizing symptoms were significant in more than one scenario in Study 2. This may be sample-specific, as Study 2 utilized a low-income, rural sample that may have experienced more early adversity. Similarly, past research has identified that demographic risk factors, such as low SES levels, can interact with maternal depression to increase the risk for children (Norcross et al., 2020), which may explain the current findings. This may also be an age effect (i.e., stronger among 2-3 years olds than 4-6 years olds), which would be expected by life history theory (Ellis et al., 2009). However, this finding may also be due to not controlling for previous child internalizing symptoms (like in Study 1), for a lack of available measures. These hypotheses need to be replicated in future studies.

Study 2 additionally found that the degree of fluctuations in maternal depressive symptoms over the first 2 years of children’s life predicted lower inhibitory control at age 3, only when mothers were recovering from relatively higher levels of depressive symptoms. This pattern is similar to that in Study 1. This finding provided preliminary evidence that the increased depression risks can be due to impairment in regulatory processes, in particular, inhibitory control among children. As inhibitory control is a key factor in developing emotion and cognitive regulation and can serve as a protective factor against developing depression (Hermansen et al., 2022), it is likely that reduced inhibitory control will serve as a mechanism of future depression risks in face of maternal depression fluctuations. This finding aligns with past observations where environmental unpredictability compromised children’s development of cognitive or stress regulation (e.g., Davis et al., 2019; Li et al., 2023; Mittal et al., 2015; Sturge-Apple et al., 2016). It is suggested that in a developmental context where the future is uncertain, being able to act on the immediate impulses rather than refraining from these impulses to obtain future treats may be adaptive for children (Sturge-Apple et al., 2016). However, future studies are needed to test hypotheses regarding how regulatory functions are affected by the fluctuations in maternal depressive symptoms.

Together, this study revealed that the degree of fluctuations in maternal depressive symptoms in children’s early life was related to increased children’s depression risks in early and middle childhood. This effect is especially salient for mothers who were recovering from relatively higher levels of depressive symptoms, supporting a harshness-by-unpredictability interaction. Findings underscore life history theory in that an unpredictable environment introduces risks to children’s learning about the environment beyond the harshness level (Ellis et al., 2009; Del Giudice et al., 2015), and fluctuations in maternal depressive symptoms as one factor that contributes to such instability and unpredictability. Findings additionally support a potential mechanism of the effect of fluctuations in maternal depressive symptoms on children, such that this effect can be due to impairment in children’s inhibitory control functions. Moreover, findings provide evidence in suggesting additional pathways of the intergenerational transmission of depression risks that have not been fully captured by current literature (i.e., how maternal symptom fluctuations contribute to environmental unpredictability that undermine children’s inhibitory control and increase depression risks beyond average symptom levels). Together, findings of the present study have significant implications for designing targeted and cost-effective interventions towards families where mothers suffer from postpartum depression.

Strengths, Limitations, and Clinical Implications

This study has several noticeable strengths. This study employed a multi-study design with longitudinal samples to provide robust associations between fluctuations in maternal depressive symptoms and various types of child depression risks in key developmental stages. This study utilized a multi-method approach, including both maternal reports of themselves and their children, as well as children’s responses to the inhibitory control task. Maternal depressive symptoms were repeatedly assessed across 4.5 years (Study 1) and 2 years (Study 2) on a similar time schedule to obtain the degree of fluctuations over early childhood. Additionally, Study 2 included a low-income sample that was under-studied but tended to experience early adversity (Sarsour et al., 2011).

Several limitations need to be considered. First, Study 2 oversampled low-income, rural families, which, despite its unique strengths, might limit the generalization of the current findings given the heightened risk levels (e.g., Ertel et al., 2011; Norcross et al., 2020). Second, this study only assessed maternal depressive symptoms, therefore findings may not generalize to other caregivers such as fathers. This study did not report unpredictability of other caregivers in the family beyond the mother, which would be a direction of future studies. Third, child depressed/anxious behaviors and internalizing symptoms were reported by mothers, which may have been biased by maternal depressive symptoms. Finally, there can exist multiple mother-child exchanges during development, and mothers’ depression may change based on children’s behavioral problems (Gross et al., 2008; Wu et al., 2019). Thus, future studies should take this parent-child mutual influence into account when investigating the multifaceted associations among maternal depressive symptoms and child depression risks.

In conclusion, the current study indicates that fluctuations in maternal depressive symptoms in early childhood can create an unpredictable environment for children’s development, and this can be particularly challenging for mothers who are recovering from severe depressive symptoms. The effects of fluctuations in maternal depressive symptoms can introduce long-lasting risks for children to develop depression in early and middle childhood. Findings of this study provides insight into facilitating monitoring and management of maternal depressive symptoms over time through ongoing and continuous care (e.g., medication, counseling, and self-care practices). For example, technology tools can be developed to track maternal symptoms over time, and mothers experiencing greater fluctuations (especially during the recovery from postpartum depression) can receive targeted interventions to treat the symptoms and/or self-care tips to improve their overall mood. Children of mothers who experience greater symptom fluctuations can also be encouraged to engage in preventative programs to enhance their regulatory abilities and reduce emotional symptoms. Together, a more complete understanding of how and when maternal depressive symptoms affect children’s depression risks will inform more targeted intervention towards this population, with the promise to promote healthy development among their children.