Language Development in Children of Clinically Depressed Mothers in Remission: Early Experience Effects

Abstract

We compared language comprehension and production across the second year of life in children of clinically depressed mothers who later remitted with children of nondepressed mothers. Altogether, 157 mother–child dyads participated: 46 with mothers diagnosed at infant age 5 months as having major, minor, or other depressive disorders who fully remitted by 15 and 24 months and 111 with nondepressed mothers. The majority of mothers (mean [M] age = 32.8 years) were married, European American, and college educated. The groups did not differ on most sociodemographic indices, and analyses controlled for residual group differences in maternal education and child birth order. Children of early clinically depressed mothers understood fewer words at 15 months (by maternal report on the MacArthur Communicative Development Inventory [MCDI]) and 24 months (by experimenter-administered Reynell Developmental Language Scales [RDLS]) than children of nondepressed mothers. Children of nondepressed mothers increased in language comprehension from 15 to 24 months, whereas children of early clinically depressed mothers did not. Aggregating over 15 and 24 months, nondepressed mothers reported that their children expressed more words than clinically depressed mothers reported their children expressed; experimenter assessment revealed no difference between the two groups. Maternal clinical depression in the first 5 months is related to reduced language comprehension and production during the second year of life even in children whose mothers remit, indicating an early experience effect of maternal clinical depression in muting child language development. Future research should identify the mechanisms by which this early experience effect occurs to inform targeted early preventative interventions for at-risk children.

Depression affects over 300 million people worldwide and is a principal cause of disability in the general population (World Health Organization, 2017). Depression is especially prevalent among women of childbearing age (Gavin et al., 2005); indeed, depression is the most common mental health condition to affect perinatal women worldwide. Globally, perhaps 17% of new mothers suffer from symptoms and disorders of depression (Hahn-Holbrook et al., 2017). The prevalence of maternal depression in low- and middle-income countries is estimated to be between 15% and 57% (Shidhaye & Giri, 2014). In the United States, 7% of postpartum women experience a major depressive episode within the first 3 months after childbirth (Gavin et al., 2005), and as many as 13% experience symptoms of postpartum depression (Bauman et al., 2020; Hahn-Holbrook et al., 2017). Findings from the Centers for Disease Control and Prevention’s analysis of the Pregnancy Risk Assessment Monitoring System indicate that symptom rates are higher among women who report younger age, ethnic minority membership, lower educational attainment, single marital status, and participation in the Special Supplemental Nutrition Program for Women, Infants, and Children during pregnancy (Bauman et al., 2020). Although new motherhood may be blissful in many respects (Nelson-Coffey & Stewart, 2019), pregnancy and child-birth are risk factors for depression (O’Hara & McCabe, 2013), and depressed mothers are less equipped to meet their children’s needs across physical, socioemotional, and cognitive domains (Dix & Moed, 2019; Sohr-Preston & Scaramella, 2006). Maternal depression therefore likely has meaningful implications for children’s development (Goodman et al., 2011).

Language Development in Typically Developing Infants

Language acquisition in infancy instigates important developmental cascades (Oakes & Rakison, 2019). Early language skills predict later intelligence and academic functioning (Bornstein, Hahn, et al., 2016; Bornstein, Hahn, et al., 2016), have consequences for broader domains of cognitive development (Pruden et al., 2011), and extend to social and behavioral adjustment (Bornstein et al., 2013; Petersen et al., 2013). Reciprocally, early language difficulties predict diminished performance on academic, social, and behavioral measures of school readiness (Justice et al., 2009).

Language development is guided by a confluence of biological and environmental factors that unfold over time. Behavior genetic twin studies demonstrate that child language is affected by biological and experiential factors alike (Dehaene-Lambertz et al., 2006; Kovas et al., 2005; Stromswold, 2001), and progress in interrelated systems of phonology, perception, cognition, and socioemotional functioning provide the framework for ongoing language development (Kuhl et al., 2005). Importantly, language is inherently social and dyadic, and variation in social contexts profoundly alters language learning (Brookman et al., 2020; Kuhl, 2007). Thus, mother–infant interactions from birth shape systemic bases for child language development in addition to influencing language in structural ways (Golinkoff et al., 2015). Accordingly, infant-directed speech (Singh et al., 2009); parental warmth (Madigan et al., 2019); and overarching parental cognitions and practices, such as parents’ abilities to perceive, interpret, and respond contingently to children’s needs, have been identified as important predictors of child language (Prime et al., 2020; Tamis-LeMonda et al., 2001).

Language Development in Infants of Depressed Mothers

Qualities of maternal caregiving are critical factors linking postpartum depression with child language development. Research suggests that speech directed to infants differs qualitatively in depressed versus nondepressed mothers (Kaplan et al., 2001), and mothers with depression tend to display fewer warm and responsive, and more disengaged and intrusive, parenting behaviors than nondepressed mothers, a style that persists even after depression has remitted (Lovejoy et al., 2000). However, close analysis of the large body of research on the effects of maternal depression on child development reveals that empirical evidence for specific associations between maternal depression and child expressive and receptive language is surprisingly inconsistent (Slomian et al., 2019).

Some studies have reported associations between maternal depressive symptoms and reduced language expression (Kaplan et al., 2014; Treat et al., 2020) and language comprehension (Brennan et al., 2000) in children. However, other research has reported no effects of maternal depression on child language development (Cornish et al., 2005; Murray, 1992; Thorpe et al., 2003). For example, with regard to expression, one study reported that 24-month-olds whose mothers scored in the 90th percentile on the Center for Epidemiologic Studies Depression (CES-D) Scale produced four fewer word types than children whose mothers scored at the 10th percentile, a difference that grew to 20 word types at 36 months (Pan et al., 2005); however, another study reported that maternal depression on the CES-D at 9 months did not predict mothers’ reports of child language expression at 24 months (Paulson et al., 2009). However, expression was measured using automated analyses of videorecorded interactions in the first study, whereas mothers reported expressive language on the Mac-Arthur Communication Development Inventory (MCDI; Fenson et al., 1993) in the second study. Furthermore, an investigation in low- and middle-income countries in Africa of 34 factors thought to contribute to child language development found that child hemoglobin/iron status, access to play materials, and dietary diversity predicted child language at 18 months, but several maternal factors (e.g., depression, stress, education, hemoglobin/iron status) and other child factors (e.g., acute respiratory infections) did not (Prado et al., 2017). As potential explanations for these null effects, the authors suggested that early language may be resilient to some early risk factors, such as maternal depression.

Methodological Considerations in Research on Maternal Depression and Infant Language Development

Variation in methods constitutes one factor that renders comparing findings across studies difficult and drawing firm conclusions about effects of maternal depression on child expressive and receptive language development especially challenging. First, diagnoses of maternal depression may be clinical or confined to symptoms, and postpartum women who self-report depressive symptoms may not surpass thresholds for diagnoses of clinical depression (Campbell & Cohn, 1991; Stein et al., 2008; Zajicek-Farber, 2010). Subclinical depressive symptomatology alone may attenuate potential associations between maternal depression and child language outcomes. Second, substantive and methodological challenges arise in the measurement of child language. Substantively, either comprehension or production of vocabulary or grammar in children is commonly assessed, ignoring the multivariate nature of child language and a wholistic view of the child. Methodologically, some studies rely on maternal reports of children’s language, which is an indirect measure, whereas other studies collect language data from direct observations or tests of children. However, agreement between parent reports and independent observations of children are known to vary, especially perhaps in the context of maternal depression (De Los Reyes & Kazdin, 2005). For example, compared with nondysphoric mothers, mothers with higher levels of dysphoria showed less agreement with independent raters on their children’s negative behaviors and emotions (i.e., rating their children more negatively) and greater agreement with independent raters on their children’s expressed positive emotions (Youngstrom et al., 1999). Furthermore, infants may display different behaviors and capacities (e.g., greater gaze behavior, more smiling, less fussiness) when with their depressed mothers compared with unfamiliar adults (Pelaez-Nogueras et al., 1994). Third, sociodemographic characteristics tend to vary within and between samples: Perhaps socioeconomic status, rather than maternal depression, predicts child language, an effect partially mediated by caregiving practices (Prado et al., 2017).

Additionally, maternal depression and child language are both dynamic in the sense that depression may persist or remit over time, and language is continually acquired as children age. One example study examined cognition and language in 15-month-old children of nondepressed, briefly depressed (clinical depression within the first 4 months postpartum only), and chronically depressed (clinical depression at 4 and 12 months) mothers (Cornish et al., 2005). Maternal chronic depression was related to diminished child cognition, whereas children of briefly depressed mothers did not differ in cognition from the children of nondepressed mothers; no differences emerged in expressive or receptive language across the three groups, but language was measured only up to a very young age. Other investigations have also failed to detect an association between the timing of maternal depression symptoms and children’s language comprehension (Brennan et al., 2000). In contrast, diminished cognition has been detected in 11-year-old boys of mothers who were clinically depressed at 3 months postpartum and who had and had not experienced a recurrence of depression (Hay et al., 2001). Last, central developmental issues of stability (consistency or inconsistency in individual differences over time) and continuity (consistency or inconsistency in group mean level over time) are often overlooked in the field of depression research, limiting our understanding of the nature and course of depression’s purported effects on the overall ontogeny of child language (Bornstein et al., 2017).

The Current Study

Taken together, extant reports in the literature reinforce the need to examine multiple domains of child language from multiple informants in multiple contexts at multiple time points comparing otherwise-comparable children of clinically depressed with nondepressed mothers to reach a more comprehensive comparative understanding of the course of language development in children of clinically depressed mothers (De Los Reyes et al., 2020). We attempted to do so here. Specifically, we implemented a longitudinal study of the individual-order stability and group mean-level continuity of language in children of initially clinically depressed but later remitted mothers and in children of nondepressed mothers. Mothers in the depressed group were clinically depressed at 5 months but had fully remitted at 15 and 24 months, permitting the isolation and investigation of early experience effects. To evaluate the specificity or generality of depression’s effects on child language acquisition, we gathered data on child actions, gestures, comprehension, and production, expecting specific effects on language; to evaluate possible methodological variation, we gathered data on child language via maternal report and independent standardized assessments by trained administrators, expecting possible methodological variation; to evaluate possible developmental effects, we gathered data on child language at two waves across the first 2 years of life, expecting individual-order consistency but group discontinuity across time. Finally, because girls and boys are known to acquire language on different schedules (Bauer et al., 2002; Bornstein et al., 2004), we explored gender differences in child language development, expecting girls to outperform boys in language acquisition.

Method

Study Design, Participants, and Selection Criteria

Mothers

Mothers were recruited from the Washington, DC, metropolitan area through mass mailings, women’s groups, and newspaper advertisements. All participants were self-referred into the study. To identify similar groups of participants, 316 mothers between 1 and 5 months postpartum who screened with low (1–7) and high scores (>12) on the Beck Depression Inventory (BDI-II; Beck et al., 1996) were invited to participate in the Structured Clinical Interview for DSM–IV Axis I Disorders–Nonpatient research version (SCID-I; First et al., 2001) administered by trained licensed clinical psychologists at 5 months postpartum. At this interview, the definition of a “current” episode of depression was modified to “within the lifetime of the child” because this was the question of interest (Manian et al., 2013). Mothers with a primary diagnosis of depression (major depression, minor depression, dysthymia, or depressive disorder not otherwise specified [DDNOS]) on the SCID-I (n = 121) were assigned to the clinically depressed group. In the Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM–IV; American Psychiatric Association, 1994), major depression is defined as five or more symptoms of depression (including depressed mood or anhedonia) for 2 weeks or more, causing significant distress or impairment. For a diagnosis of minor depression, only two to four symptoms of depression are required. To meet the criteria for major or minor depression, there should be no evidence of manic or hypomanic episodes. Dysthymia is specified by the predominance of depressed mood for at least 2 years, without an interruption in symptoms for 2 months or more. In addition to depressed mood, two or more symptoms of dysthymia are required, causing significant distress or impairment. To meet criteria for dysthymia, there should be no evidence of manic, hypomanic, or mixed episodes or major depressive disorder in the first 2 years. DDNOS reflects presentation of depressed mood with clinically significant impairments that do not meet the criteria for duration or severity as specified in the diagnosis of major or minor depression or dysthymia.

Following these diagnostic criteria, the clinically depressed group consisted of 77 mothers with major depression, 29 with minor depression, 9 with dysthymia, and 6 with DDNOS. Mothers who were not diagnosed with any depressive disorder on the SCID-I at 5 months (n = 195) were assigned to the nondepressed group. Trained licensed clinical psychologists also administered the SCID-I to mothers at 15 and 24 months. Of the 121 mothers identified as clinically depressed at 5 months, 31 did not return, 13 were diagnosed with clinical depression, and 77 were nondepressed at 15 months; 43 did not return, 11 were diagnosed with clinical depression, and 67 were nondepressed at 24 months. Diagnostic classifications for the sample at 5, 15, and 24 months are presented in detail in Tables S1 and S2 in the online supplemental materials. To isolate early experience effects, we used only language data of children of mothers who were diagnosed as clinically depressed at 5 months and who had fully remitted by 15 and 24 months for the depressed group (n = 55). Mothers not diagnosed with any depressive disorder at 5, 15, and 24 months were selected for the nondepressed group (n = 132). Because language development was our main interest and the language assessments were administered in English, we excluded mothers whose primary language when interacting with the child was not English. Thus, the final samples consisted of data for 46 children whose mothers were clinically depressed at 5 months but remitted at 15 and 24 months and 111 children whose mothers were nondepressed. There was no attrition in the sample from 15 to 24 months. In terms of clinical presentation at 5 months, clinically depressed mothers had an average of 5.28 symptoms; approximately 67% had major depression, and 24% had minor depression; 74% had postpartum onset; 67% had two episodes or more; 28% were in therapy; and 37% were on medication. (Self-reported depressive symptoms of mothers are reported in Table S3 in the online supplemental materials.) At 15 months, the average number of SCID-I symptoms, out of 9 possible, for mothers in remission had reduced to .36 (range = 0–3, median [Mdn] = 0); at 24 months, the average was .50 (range = 0–5, Mdn = 0). For the nondepressed group, the average number of SCID-I symptoms was .04 at 5 months (range = 0–1, Mdn = 0), .05 at 15 months (range = 0–2, Mdn = 0), and .06 at 24 months (range = 0–2, Mdn = 0).

Descriptive statistics and group differences for sociodemographic variables are reported in Table 1. The two groups of participating mothers did not differ in age, marital status, work status, family income, distributions of ethnicity, or verbal intelligence. However, the two groups differed in educational attainment; in consequence, we controlled for maternal education in the statistical analyses of group differences that follow.

Table 1.

Sociodemographic Descriptive Statistics for Maternal Early Depressed but Remitted and Nondepressed Groups at 5 Months

Variable	Depressed (n = 46)	Nondepressed (n = 111)	Group differences
Maternal age (years)	32.82 (4.94)	32.82 (4.47)	F(1, 152) = 0.17, ns
Marital status (married)	84.4%	92.7%	F(1, 153) = 1.08, ns
Maternal ethnicity			χ2 (3, N = 154) = 4.23, ns
European American	34	89
African American	10	14
Asian American	1	1
Biethnic	0	5
Maternal education			χ2(2, N = 153) = 7.17*
Partial college or less	34.1%	16.5%
Undergraduate degree	38.6%	37.6%
Graduate degree	27.3%	45.9%
Mother employed outside of home (% yes)	40.9%	39.4%	χ2(1, N = 153) = 0.87, ns
Number of hrs mother worked	12.84 (16.42)	13.81 (17.26)	F(1, 151) = 0.10, ns
Family income ($)	100,454 (47,149)	109,266 (51,634)	F(1, 151) = 0.96, ns
Maternal verbal intelligence	106.51 (9.76)	109.36 (11.54)	F(1, 146) = 1.95, ns
Child age (months)	5.21 (.31)	5.14 (.22)	F(1, 156) = 3.11, ns
Child gender (% male)	54.3%	57.7%	χ2(1, N = 157) = 0.15, ns
Child birthweight (lbs)	7.29 (.89)	7.60 (1.0)	F(1, 151) = 3.20, ns
Child birth order (% firstborns)	66.1%	47.7%	χ2(1, N = 153) = 4.42*
Child in daycare (M hrs per month)	33.1 (41.6)	23.8 (34.6)	F(1, 143) = 1.91, ns

Note. ns = not statistically significant. M (SD) reported.

^*p < .05.

Children

All infants were term, healthy singletons with no known genetic disorders or birth complications (Rutter et al., 2003). Children’s average ages were mean (M) = 5.16 months (standard deviation [SD] = .25), M = 14.97 months (SD = .33), and M = 24.35 months (SD = .43) at the three waves of data collection. Children did not differ between the two groups in ages at testing or in the distributions of their gender. Birth order was significantly different between the two groups; in consequence, we controlled for child birth order in the statistical analyses of group differences that follow.

Family recruitment and the conduct of the study “Dyadic Interactions in Depressed and Non-Depressed Mothers and Their Infants” were approved by National Institutes of Health (NIH) Clinical Center, Eunice Kennedy Shriver National Institute of Child Health and Human Development Institutional Review Board (IRB), under Protocol 02-CH-0278 (NIH Clinical Trials Identifier: NCT00044174).

Maternal Depression Measures

The BDI-II (Beck et al., 1996) is a 21-item (4-point scale ranging from 0 to 3) self-report measure of the presence and degree of depressive symptoms consistent with the DSM–IV. The SCID-I (First et al., 2001) is a semistructured interview for making DSM–IV Axis I diagnoses. The SCID-I is administered by trained and certified mental health professionals familiar with the DSM–IV classification and diagnostic criteria (American Psychiatric Association, 1994). Four pilot cases were audio recorded and scored by the two administrators; diagnostic consensus was 100%.

Child Language Measures

Fifteen Months

Mothers completed the MCDI (Fenson et al., 1993) infant version, a parent report of child actions, gestures, and language comprehension and expression designed for use with 8- to 16-month-olds. Part 1 asks the parent to specify which of 396 words organized in 19 semantic categories the child understands or says and understands. Comprehension is the sum of words the child understands or understands and says, and Expression is the sum of words the child understands and says. Part 2 focuses on actions and gestures for appraising early communicative and representational skills not dependent on verbal expression per se (e.g., pointing, giving, and showing). Actions or gestures that the parent indicates the child performs are summed to create a total score. The internal consistency reliabilities of the MCDI (Cronbach’s alpha) were .93 for Comprehension, .94 for Expression, and .74 for Action and Gestures.

At 15 months, experimenters administered the Reynell Developmental Language Scales (RDLS; Reynell & Gruber, 2003), an assessment of language comprehension and expression designed for use with 1-to 6-year-olds. In Verbal Comprehension Scale A, the child is asked to demonstrate an understanding of increasingly difficult verbal expressions ranging from labeled objects to higher-order concepts. The Expressive Language Scale uses two subscales (appropriate to age level): structure and vocabulary. The structure subscale is based on the child’s spontaneous expressions during the assessment and scored on a scale from vocalizations other than crying to the use of complex sentences. In the vocabulary subscale, the child is asked to name familiar objects and actions from pictures. Scores represent the sum of items the experimenter recorded that the child comprehended or expressed. Administrators of the RDLS were blind to the depression status of children’s mothers.

Twenty-Four Months

Mothers completed the MCDI toddler version for 16- to 30-month-olds, a 680-word language expression checklist organized into 22 semantic categories (Fenson et al., 1993). Vocabulary Expression is the sum of words the child says. In our data, the Cronbach reliability was .94 for Vocabulary Expression.

At 24 months, a separate group of administrators, also blind to the depression status of children’s mothers, retested children with the RDLS. Split-half reliability coefficients for the Comprehension and Expressive Scales for children between the ages of 2 years and 2 years 5 months are both .93 in the normative sample (Reynell & Gruber, 2003).

Analytic Plan

Distributions of all variables were first examined for normality, outliers, and influential cases, and transformations were applied to resolve problems of nonnormality (Tabachnick & Fidell, 2012). MCDI Expression and RDLS Expression scores at 15 and 24 months required log transformation, and MCDI Actions and Gestures required a square transformation. For ease of interpretation, descriptive statistics are presented in the variables’ original metrics.

Child birth order and birth weight, maternal age, education, verbal intelligence, marital status, ethnicity, and social desirability bias were all evaluated as potential covariates. Only maternal education and child birth order differed between the groups and qualified as covariates for all analyses (see the online supplemental materials). We also evaluated patterns of nonmissing data (see the online supplemental materials) and repeated the analyses with imputed data (see Table S4 in the online supplemental materials). Given that missing data were missing at random and that the results did not differ, we present the results with listwise deletion of data. That is, within a single analysis, we included only those cases that had nonmissing data for all the variables for that particular analysis.

For scales administered at one time point, we tested effects of group and gender using analysis of covariance (ANCOVA), controlling for maternal education and child birth order. For scales administered at 15 and 24 months, we tested continuity across child age and the effects of group and gender as well as 2-way interactions employing repeated-measures analysis of covariance (RM-ANCOVA), controlling for maternal education and child birth order. Child age was treated as a within-subjects variable, and group status (clinically depressed vs. nondepressed) and gender (girls vs. boys) were treated as a between-subjects variable. Stability was assessed by partial zero-order Pearson correlation coefficients, controlling for maternal education and child birth order.

Results

Descriptive Statistics and Stability of Child Language

Table 2 presents descriptive statistics and stability across time for language measures by group. All language measures were moderately to strongly stable, but children showed differential stability by measure and depression group. Stability was higher for children of depressed mothers on language expression by maternal report on the MCDI than for children of nondepressed mothers (z = 2.69, p = .007).

Table 2.

Descriptive Statistics and Stability Across Time for Child Language

	Descriptive statistics				15- to 24-month stability
	15 months		24 months
Language scale	M	SD	M	SD	pr
MCDI
Comprehension
Depressed	101.88	70.86	n/aa	n/aa	n/aa
Nondepressed	154.32	82.44
Actions and Gestures
Depressed	28.59	13.65	n/aa	n/aa	n/aa
Nondepressed	31.78	13.85
Expression
Depressed	16.24	17.58	254.39	173.01	.77 ***
Nondepressed	25.84	34.74	327.53	159.38	.40 ***
RDLS
Comprehension
Depressed	7.47	3.22	23.68	9.21	.39*
Nondepressed	8.01	3.41	29.04	9.25	.47***
Expression
Depressed	8.81	2.91	22.23	5.85	.56*
Nondepressed	9.47	4.01	24.72	5.41	.39***

Note. MCDI = MacArthur Communicative Development Inventory; RDLS = Reynell Developmental Language Scales; pr = partial correlation; n/a = not applicable; did not administer this measure at 24 months. Descriptive statistics represent raw scores. Ns for stability statistics for clinically depressed group = 37–43; nondepressed group = 87–106. Descriptive statistics represent full sample available at each time point; stability estimates are partial correlations controlling for maternal education and child birth order. Bold statistics represent a significant difference between clinically depressed and nondepressed groups.

^aMCDI Actions and Gestures and Comprehension scales are developmentally inappropriate for children age 24 months (see the text).

^*p < .05.

^***p < .001.

Effects of Child Age, Group, and Gender on Child Language

ANCOVA and RM-ANCOVA results are presented in Table 3. Only significant effects are reported here.

Table 3.

Age, Group, Gender, and Interaction Results for Child Language

	MCDI						RDLS
	Comprehensiona		Actions and Gesturesa		Expressionb		Comprehensionb		Expressionb
Predictor	F	${\mathrm{\eta }}_p^2$	F	${\mathrm{\eta }}_p^2$	F	${\mathrm{\eta }}_p^2$	F	${\mathrm{\eta }}_p^2$	F	${\mathrm{\eta }}_p^2$
Age	n/a	n/a	n/a	n/a	249.45***	.68	143.70***	.51	149.08**	.53
Group	10.50**	.07	2.75	.02	7.35**	.06	4.21*	.03	1.08	.00
Gender	3.58	.02	3.58	.03	15.67***	.12	19.78***	.13	14.94***	.10
Age × Group	n/a	n/a	n/a	n/a	.07	.00	6.87*	.05	.24	.00
Age × Gender	n/a	n/a	n/a	n/a	.37	.00	10.43**	.07	.01	.00

Note. MCDI = MacArthur Communicative Development Inventory; RDLS = Reynell Developmental Language Scales; n/a = not applicable; did not administer this measure at 24 months. All analyses control for maternal education and child birth order. Ns for clinically depressed group = 37–43; nondepressed group = 87–106.

^aStatistics based on analysis of covariance (ANCOVA) at 15 months.

^bStatistics based on repeated-measures ANCOVA from 15 to 24 months.

^*p < .05.

^**p < .01.

^***p < .001.

MCDI Comprehension at 15 Months

ANCOVA revealed a main effect of group: Nondepressed mothers reported that their children understood more words than depressed mothers reported for their children (Estimated Marginal Mean [EMM] = 155.32, standard error [SE] = 7.95; EMM = 105.90, SE = 12.86), respectively. There were no significant interactions.

MCDI Actions and Gestures at 15 Months

ANCOVA revealed no main effects or significant interactions.

MCDI Expression at 15 and 24 Months

RM-ANCOVA revealed main effects of child age, group, and gender. Expectedly, children were reported to express more words at 24 months than at 15 months (EMM = 301.70, SE = 14.72; EMM = 23.77, SE = 2.87, respectively), nondepressed mothers reported that their children produced a greater number of words than clinically depressed mothers reported for their children (EMM = 178.64, SE = 9.10; EMM = 133.60, SE = 13.93, respectively), and girls reportedly produced more words than boys (EMM = 186.62, SE = 11.70; EMM = 125.61, SE = 10.87, respectively). There were no significant interactions.

RDLS Comprehension at 15 and 24 Months

RM-ANCOVA revealed main effects for child age, group, and gender, as well as Age × Group and Age × Gender interactions. Child comprehension increased from 15 to 24 months (EMM = 7.97, SE = .32; EMM = 27.40, SE = .82, respectively). Children of nondepressed mothers comprehended more language than children of depressed mothers (EMM = 18.72, SE = .51; EMM = 16.64, SE = .86, respectively). Girls comprehended more language than boys (EMM = 19.62, SE = .63; EMM = 15.75, SE = .63, respectively). However, in consideration of the Age × Group interaction, we explored the effect of group at each of the two ages. No group mean difference in tested comprehension was found at 15 months, F(1, 152) = .83, not statistically significant (ns), but a group difference emerged at 24 months, F(1, 146) = 9.83, p = .002, ${\mathrm{\eta }}_{\text{p}}^2=.06$: Two-year-olds of nondepressed mothers comprehended more language than 2-year-olds of previously depressed mothers (EMM = 29.04, SE = .89; EMM = 23.68, SE = 1.46, respectively). In consideration of the Age × Gender interaction, we explored the effect of child gender at each of the two ages. Gender had significant effects at 15 months, F(1, 152) = 13.14, p < .001, ${\mathrm{\eta }}_{\text{p}}^2=.08$, and 24 months, F(1, 148) = 16.99 p < .001, ${\mathrm{\eta }}_{\text{p}}^2=.10$. However, there was a greater difference in tested comprehension between girls and boys at 24 months (EMM = 31.09, SE = 1.13; EMM = 24.92, SE = 1.13, respectively) than at 15 months (EMM = 8.93, SE = .39; EMM = 7.02, SE = .35, respectively).

RDLS Expression at 15 and 24 Months

RM-ANCOVA revealed main effects of age and gender. Child tested language expression expectedly increased from 15 to 24 months (EMM = 9.40, SE = .35; EMM = 24.24, SE = .50, respectively). Girls spoke more than boys (EMM = 18.20, SE = .51; EMM = 15.34, SE = .46, respectively) in the two groups. There were no significant interactions.

Discussion

We examined stability, continuity, and group differences of multiple measures of child language assessed by different informants using different methods at two ages in children of mothers who were clinically depressed in their child’s infancy but had remitted around the end of their child’s first year through the second year with children of nondepressed mothers. Three results accord with normative language development across the second year and affirm the validity of the study protocol and measures: Children of depressed and nondepressed mothers alike spoke more words at 24 months than they did at 15 months, whether measured by maternal report or experimenter assessment; children in both groups were stable in their language from 15 to 24 months; and girls enjoyed early linguistic advantages over boys in expression, whether measured by maternal report or experimenter assessment (effects based on differential maturational timetables, gender-typed interests, or learning opportunities; Bornstein, 2013a).

As to the main aim of the study, across the second year of life, young children of once clinically depressed mothers were disadvantaged in their language development relative to otherwise-equivalent young children of mothers who were not depressed. Specifically, children of early depressed mothers understood fewer words at 15 months (based on maternal report), and they spoke fewer words by maternal reports aggregated over 15 and 24 months. Children of early depressed mothers also understood fewer words by experimenter assessment at 24 months than children of nondepressed mothers. Importantly, these results appear to be specific to early verbal language development because the children of clinically depressed and nondepressed mothers did not differ in their nonverbal actions and gestures.

Previous research has established that chronic maternal psychopathology affects children’s cognition (Campbell & Cohn, 1997) and has documented associations between remission of maternal depression and reduction in mental and behavioral symptoms and disorders of offspring (Weissman et al., 2006). Here, however, we found reduced language comprehension and expression at 15 and 24 months in children of early depressed but remitted mothers. This result points to lingering “early experience effects” of maternal depression and suggests that benefits of maternal depression remission may not extend to child language through year 2. This result is also noteworthy from a public health perspective. As indicated at the outset, depression is endemic and pervasive in newly parturient mothers, and depression is known to continue in substantial proportions of postpartum depressed women (Goodman, 2004). Here, we found that mothers’ clinical depression in the first 5 months cast a shadow on their children’s language into the second year when mothers themselves remitted from their depression. About one third of women with a major postnatal depressive episode fully remit by 6 months (Torres et al., 2019).

Such early experience effects might reflect a sensitive period, a time in development when a system is specifically and lastingly influenced by experience (Bornstein, 1989). Previous work has suggested that maternal depressive symptoms in the first postnatal year, but not at 36 months, are associated with poor child language at 36 months (Stein et al., 2008). That is, early maternal depression might have lingering consequences as children grow (Radke-Yarrow et al., 1992), and child language development may be affected by early and/or chronic maternal depression (Hay et al., 2001). Our results further localize some effects of very early experience of maternal depression on child verbal language development.

Developmental scientists have pursued basically three avenues to study children: report, test, and observation (Bornstein & Haynes, 1998). In this study, we employed two of the three, report and test. Each mode of study has advantages and disadvantages, and (however desirable) there is no guarantee that any two modes will yield identical results. Furthermore, the comparability of results may depend on whether the developmental comparison is of mean level or rank order. Earlier, we noted that parent reports and independent observations of children are known to differ, especially in the context of maternal depression and when children are with their mothers versus strangers. In two comparative assessments here, we found that the two methods painted somewhat divergent pictures of child language. At 15 months, nondepressed mothers reported that their children comprehended more words than clinically depressed mothers reported their children comprehended, whereas experimenter testing revealed no difference between the two groups. Aggregating over 15 and 24 months, nondepressed mothers reported that their children expressed more words than clinically depressed mothers reported their children expressed, whereas experimenter testing revealed no difference between the two groups. Discerning possible reason(s) behind these differences is more speculative: It could be, as has often been argued, that mothers know their children better, or that children of depressed mothers are muted in their language performance with stranger experimenters, or that the RDLS is more restricted than the MCDI and so structurally minimizes possible differences between means, among other reasons. We note, however, that experimenter testing (RDLS) at 24 months yielded a group difference in comprehension that paralleled maternal report (MCDI) at 15 months. The homily here for researchers (although most already know it) is that measures matter and that results may be specific to the measure (Bornstein, 2013b, 2017).

Although children of nondepressed and previously depressed mothers were both stable in their language development between 15 and 24 months, children of previously depressed mothers were more stable. (Notably, very preterm infants are also more stable in their early language development than moderate-late preterm and term infants, reinforcing the notion that at-risk infants may be more rigid or less flexible in their early development and disadvantaged in later catch-up; Putnick et al., 2017.) Between 15 and 24 months, children are relatively stable in their language skills (i.e., children who do better in language skills at 15 months continue to do better at 24 months; Bornstein, Hahn, et al., 2016; Bornstein, Hahn, et al., 2016), but stability is still far from perfect (e.g., some reorganization in language development is expectable as toddlers reach a certain level of language proficiency and, at around 18 months, show an explosion of vocabulary development; Huttenlocher et al., 1991; Reznick & Goldfield, 1992). The greater stability in children of depressed mothers could indicate some increased rigidity or diminished flexibility as they grow. Whether this enhanced developmental stability is limited to language skills or has broader generalizability (e.g., to other cognitive functions or socioemotional characteristics) is an open question that should prompt future research.

Strengths of This Study

Several design aspects distinguish the current study from other investigations of child language in depressed families. Diagnoses rendered by a licensed clinician increased the specificity of our results to clinical populations and disambiguated our findings from any impact of “postpartum blues.” The National Institute of Child Health and Human Development, Child Care Research Network (1999) examined RDLS scores on both Comprehension and Expression at 36 months as a function of maternal depression chronicity. Limiting our depressed sample to mothers who had remitted by 15 and 24 months allowed us to distinguish early-experience effects of maternal depression on child language development from putative effects of chronic depression. The literature suggests that both early and chronic depression are associated with children’s developmental outcomes; future research should compare early with continuing effects of maternal depression for their respective consequences for multiple developing systems in young children. Assessing multiple language variables and using multiple reporters increased the precision and robustness of our estimates of child language. The examination of comprehension and production, stability and continuity, and gender and longitudinality in children’s language at a crucial period in early development broadened our understanding of the nature and course of language acquisition in children of once clinically depressed compared with nondepressed mothers. Finally, evaluating multiple sociodemographic characteristics (see the online supplemental materials) as potential covariates and controlling for maternal education and child birth order reduced bias in our estimates.

Limitations of This Study

At 24 months, we collected maternal reports (MCDI) of expression only, and a balanced design would have benefited from inclusion of a language measure that examined continuity in mother-reported child language comprehension. However, the MCDI Actions and Gestures and Comprehension scales are developmentally inappropriate for children over 16 months (Feldman et al., 2000; Fenson et al., 1993). Some effect-size differences between groups were small. However, small differences early in development can (many likely) diverge into later larger differences (Bornstein, 2014; Walberg & Tsai, 1983). Language development exemplifies “K” economies or the “Matthew effect.” Matthew 13:12: “For whoever has, to him more will be given, and he will have abundance; but whoever does not have, even what he has will be taken away from him.” In essence, children whose language environments are rich will be advantaged in language, and children whose language environments are poor will be disadvantaged. Indeed, the Matthew-effect model was first applied in developmental science as a theoretical framework to study and explain individual differences in language abilities (Bast & Reitsma, 1998; Stanovich, 1986). Furthermore, mothers in our study were primarily European American and middle socioeconomic status with term, healthy infants, and language assessments were administered in English; these decisions precluded the inclusion of non–native-English-speaking mothers. Consequently, the generalizability of our findings is limited to similar populations.

Future Directions

We previously alluded to future research that would assess the generalizability of early-experience language effects to other developing systems and compare early with continuing effects of maternal depression on child development. Future research might also evaluate the separate effects of paternal depression as well as possible synergistic effects of maternal and paternal depression. Identifying the mechanisms by which maternal depression affects child language development was not the primary aim of the current study. The effects of parental depression doubtlessly involve multifactorial processes that include genetic and biological mechanisms and correlated risk factors (e.g., stress, reciprocal effects between child behavior and maternal mood; Goodman & Gotlib, 1999), environmental exposures (e.g., caregiving behaviors, infant-directed speech; Madigan et al., 2019), and associations among them (e.g., gene–environment correlations, Gene × Environment interactions; Clark et al., 2018). Future research should endeavor to examine the processes by which maternal depression specifically affects child language development. Last, several factors that were unaccounted for in the current study could influence the observed effects. Future research should examine how differences in the course (e.g., severity, chronicity) and treatment (i.e., therapy, medication) of depression might moderate associations between mothers’ early depression and their children’s language development.

Implications

Our findings illuminate critical scientific issues in relating children’s early experience with maternal clinical depression to their language development. In addition, the current study underscores the importance of early identification of depression in pregnant and postpartum women and early intervention (Bickford-Smith et al., 2005). It is possible to mitigate the impact of depression on the mother–infant relationship and child development, but much work remains to be done in identifying the effective components of interventions (Nylen et al., 2006). Because one key pathway from postnatal depression to language that is amenable to intervention may be through caregiving behaviors and literacy-oriented stimulation activities (Paulson et al., 2009; Stein et al., 2008; Zajicek-Farber, 2010), interventions might focus on enhancing depressed mothers’ verbal responsiveness to their infants and the vital opportunities for language growth that they provide (Tamis-LeMonda & Bornstein, 2002). Disruptions to the foundations of language likely have cascading impacts on development, and independent longitudinal studies support causal prospective associations between early language skills and later behavioral adjustment in children (Bornstein et al., 2013; Petersen et al., 2013). In this regard, interventions designed to enhance language skills have also been shown to augment behavioral regulation and other domains of psychological well-being, which may be especially beneficial for children of depressed mothers, who often suffer from attentional, psychopathological, and social deficits in addition to language delays (Barnett et al., 2008). Taken together, our findings highlight the need for systemic changes in designing multipronged and coordinated interventions that first identify and then support mothers with depression and their children (Dodge, 2018). Family Connects offers one potential blueprint, providing all families giving birth in a community with postnatal home visitation to assess their needs and connect them with appropriate resources (Dodge et al., 2019). In a randomized controlled trial, families receiving Family Connects showed decreased maternal depression (Dodge et al., 2019; Goodman et al., 2019).

Conclusion

Maternal depression interferes with typical psychological, emotional, and behavioral development in children, and we found that those conclusions extend to child language development. Even in mothers who had remitted by the start of the second year, depression was associated with reduced child language comprehension and production afterward, evidencing an early-experience effect. Notably, ours was a relatively educated and advantaged clinically depressed sample, and so our results may reflect optimal outcomes, and even poorer results may be expected with less educated and more disadvantaged clinically depressed samples. Given the significant number of children and families affected by maternal depression worldwide and the degree to which they suffer, researchers should continue to examine the processes by which maternal depression influences child development to better inform the implementation of targeted preventative interventions for clinically at-risk mothers and children.