Prenatal maternal mood instability and child externalizing problems: An ecological momentary assessment study

Abstract

This study examines whether exposure to affect fluctuations during fetal development have implications for child psychopathology. To evaluate maternal mood instability, participants (N = 154) completed Ecological Momentary Assessment (EMA) eight times per day across three days at 15-, 25-, and 35-weeks’ gestation, and then at two months postpartum. The EMA measured depressive symptoms (Center for Epidemiologic Studies Depression Scale), anxiety (Profile of Mood States), and stress (Perceived Stress Scale). Mood instability was quantified by calculating root mean square of successive differences (RMSSD), which captures both variability and temporal dependency of mood states. When the children were 3 years old, mothers reported on their child’s externalizing symptoms with the Child Behavior Checklist (CBCL) and then again at 4.5 years with the Strengths and Difficulties Questionnaire (SDQ). Bivariate correlations indicated a positive association between prenatal mood instability and child externalizing problems at both ages (r = 0.34, r = 0.29, respectively; all p < .05). Sensitivity analyses indicated that mood instability was predictive of child externalizing problems, even after consideration of potential confounding factors, including prenatal and postnatal negative mood level as well as postnatal mood instability. Examination of maternal mood instability, in addition to mood level, provides additional and unique information regarding how maternal mental health affects child psychopathology and underscores the developmental importance of patterns of parental signals even in the very earliest stages of neurodevelopment.

1. Introduction

Over half a century ago, Wiesel and Hubel (1963a, 1963b) covered the eye of a kitten briefly during the postnatal period, altering the morphology of the lateral geniculate body and responsivity of cortical cells resulting in functional blindness in the deprived eye. In doing so, this series of studies documented the fundamental importance of sensory inputs during critical periods for the development of the primary visual cortex. In the decades since, clear evidence describes the presence of both critical and sensitive periods in the sensory systems of multiple species, as well as the mechanisms underlying this experience-dependent circuit development (Cheng et al., 2022; Khazipov et al., 2004; Knudsen, 1998; Takesian et al., 2018). Although the central importance of sensory signal inputs for maturation of visual, auditory, and somatosensory circuits is clear, more recent research highlights that patterns of signals in early life similarly influence the development of cognitive and emotional circuits (Birnie and Baram, 2022). Specifically, experimental animal models demonstrate that unpredictable patterns of parental sensory signals influence synaptic connectivity in emotional brain circuits, resulting in abnormal circuit maturation, cognitive problems, aberrant stress responding, and disrupted reward behaviors in the offspring (Baram et al., 2012; Birnie and Baram, 2022; Bolton et al., 2022; Brunson et al., 2005; Ivy et al., 2010; Levis et al., 2021; Molet et al., 2016; Short et al., 2023). As with the sensory circuits, the fundamental importance of patterns of signals for the developing brain appear to be evolutionarily conserved, with patterns of parental sensory signals associating with emotional and cognitive phenotypes, as well as brain development in human and non-human primates (Davis and Glynn, 2024; Davis et al., 2022; Davis et al., 2017; Granger et al., 2021).

More recently, investigations have probed whether patterns of signals shape neurodevelopment, beginning in utero. These studies, using a novel indicator of mood dysregulation, mood entropy, demonstrate that instability or unpredictability of prenatal maternal mood, independent of level or valence, associates with child neurodevelopment (Glynn et al., 2018; Howland et al., 2021). Across multiple independent cohorts, higher prenatal mood entropy (more unpredictable or variable mood) predicts increased negative affectivity from infancy through late childhood (Davis and Glynn, 2024; Glynn et al., 2018). Consistent with links with elevated negative affectivity, exposure to more maternal prenatal mood entropy predicts self-reports of anxiety and depression in early adolescence (Glynn et al., 2018). Prenatal mood instability assessed with the entropy index also associates with less advanced cognitive and language development at 2 and 6–9 years of age (Howland et al., 2021), as well as poorer inhibitory control in late childhood (Davis and Glynn, 2024). It is important to note that these findings persist after accounting for levels of maternal negative mood (e.g., depression, anxiety, and stress), underscoring the unique contributions of unpredictable parental signals in the earliest stages of neurodevelopment.

The initial investigations described above examined patterns of maternal mood using mood entropy, which is quantified by the degree of unpredictability (entropy) of item-specific responses in standardized mood assessments (e.g., Edinburgh Postnatal Depression Index, State Anxiety Index) administered at a single time point (Glynn et al., 2018). Although mood entropy is associated with mood instability assessed in daily life, it is not a direct measure. In contrast, a robust methodology, which enables a strong test of the developmental relevance of patterns of maternal mood, is ecological momentary assessment (EMA). EMA (Stone and Shiffman, 1994), or experience sampling methods (Csikszentmihalyi and Larson, 1987), refers to approaches characterized by repeated collection of moment to moment, real-time measures of individuals’ experiences in their natural environments, providing data that are not available through more traditional measurement techniques. The advantages of EMA for the assessment of mood include increased accuracy, reduced retrospective bias, enhanced generalizability due to assessment in real world contexts, and characterization of dynamic processes in real-time (Ebner-Priemer and Trull, 2009; Shiffman et al., 2008). We are aware of only one study in which EMA was used to examine mood dynamics in pregnancy. MacNeill et al. (2023) assessed perceived stress up to four times a day over a 14-week period, finding that greater instability was associated with higher levels of infant negative affectivity at three months of age.

Accumulating evidence indicates that among the most reliably detected consequences of exposure to unpredictability in early life is altered effortful control (Davis and Glynn, 2024; Davis et al., 2019; Holmberg et al., 2022; Short et al., 2023), a component of executive function involved in the purposeful overriding of a prepotent or dominant response (Rothbart et al., 2011; Santens et al., 2020). Given that impairments in self-regulation are a transdiagnostic risk factor for psychopathology, and also specifically indicate increased propensity for behaviors such as disobedience, rule-breaking, and aggression (Eisenberg et al., 2015; Fergusson et al., 2005; Schoppe-Sullivan et al., 2009; Wills et al., 2016), here we provide an initial investigation of whether unpredictable prenatal maternal mood portends risk for child externalizing problems. The present prospective longitudinal study implements EMA to test whether maternal mood instability during gestation predicts externalizing behaviors in early childhood and whether these associations are independent from mood levels.

2. Methods

2.1. Study overview

Fig. 1 overviews data collection. Participants completed EMA of mood eight times a day for three consecutive days three times during pregnancy at 15-, 25-, and 35-weeks’. Assessments were conducted in early, mid, and late gestation to determine whether mood instability might differ as a function of pregnancy, as is the case for stress responding (Barron et al., 1986; Glynn et al., 2004; Glynn et al., 2001; Matthews and Rodin, 1992; Schulte et al., 1990), and also to ensure that if sensitive periods for fetal exposure to mood instability exist, they would be detectable (Davis and Sandman, 2010; Glynn and Sandman, 2012). Participants then reported on their child’s externalizing problems at 45 and again at 54 months of age (SD = 5.8 and 11.3 months, respectively).

Fig. 1.

Overview of Data Collection. EMA = Ecological Momentary Assessment; CBCL = Child Behavior Checklist; SDQ = Strengths and Difficulties Questionnaire. Note. A smaller subset of the cohort (n = 75) completed the postnatal EMAs.

2.2. Participants

Study participants were 154 mother-child dyads who were recruited from obstetric clinics when mothers were in their first trimester of pregnancy (see Table 1 for demographic characteristics). Initial recruitment criteria included: singleton pregnancy, over age 18, English-speaking, and absence of major medical conditions that might affect endocrine function. Additional criteria for inclusion in this study were: completion of a sufficient number of EMAs (>3/day) and at least one of the relevant postnatal assessments (see Fig. S1 for details). Those who were included in this study did not differ in race/ethnicity, parity, household income-to-needs ratio or maternal education level from those who were not. It was the case, however, that the mothers in this study were slightly older on average (See Table S1 for the demographic characteristics of the two groups, as well as statistical comparisons).

Table 1.

Participant characteristics (N = 154).

Maternal race/ethnicity (%)
Latina	46.8
Black	3.2
Asian	11.0
Non-Hispanic White	31.8
Other	7.1
Maternal age (mean years)	29.6
Income-to-needs ratio (median %)	219
Cohabitation with child's father (% yes)	87.0
Birth order (% first born)	48.1
Child sex (% female)	46.8

2.3. Measures

2.3.1. Assessment of maternal mood

Participants completed EMAs on a smart phone (via mEMA Ilumivu). Those without a personal iPhone were provided with a phone and cell service for the duration of the study. Approximately 89 participants (58.6 %) were provided with a smart phone. They were prompted to complete eight EMAs per day across three days, at three gestational time points (Fig. 1). These prompts were delivered at semi-random times between the hours of 9 am and 9 pm. In addition to being restricted to the hours of 9 am and 9 pm, surveys were scheduled to be delivered 30–120 minutes after the completion of the previous one. This approach ensured that the surveys were spaced throughout the day (rather than all in the afternoon, for instance), while still maintaining randomization. Each assessment measured depression, stress, and anxiety symptoms and took approximately three minutes to complete. A smaller subset of the cohort (n = 75) completed an identical EMA at 2 months postpartum.

2.3.1.1. Depressive symptoms.

Maternal depressive symptoms were assessed with the very short visual analog form of the Center for Epidemiologic Studies Depression Scale, or the CES-D-VAS-VS (Radloff, 1977). The CES-D-VAS-VS has been established as a valid and reliable measure of prenatal depression for EMA (Moullec et al., 2011). Participants rated their level of depressive symptoms, using a slider ranging from 0 (Not at all) to 10 (Absolutely).

2.3.1.2. Anxiety.

The Anxious mood subscale of the 15-item Profile of Mood States (POMS) was used to assess maternal anxiety and is the average of 3 items (“anxious”, “on edge”, and “uneasy”) (Cranford et al., 2006; McNair et al., 1992). The POMS is a reliable measure of mood, can detect within-person change processes, and has been used in previous EMA studies (Cranford et al., 2006; Engel et al., 2005).

2.3.1.3. Stress.

The Perceived Stress Scale (PSS) was used to assess perceptions of stress (Cohen et al., 1983). The 4-item PSS has been shown to have strong validity in measuring perceived stress and has been validated for use in EMA contexts (Cohen, 1988; Murray et al., 2023).

2.3.2. Child externalizing problems

Mothers completed the Child Behavior Checklist (CBCL) when the children were 3.8 years old (SD = 5.8 months). The Externalizing scale consists of the attention problems and aggressive behavior syndrome scales (Achenbach and Rescorla, 2000). The CBCL DSM-Oriented scales have strong reliability and convergent validity with DSM clinical diagnoses (Nakamura et al., 2009). For the present study, the scale showed strong internal reliability, with a Cronbach’s alpha of 0.93.

When the children were 4.5 years of age (SD = 11.3 months), externalizing problems were assessed with the Conduct Problems subscale of the Strengths and Difficulties Questionnaire (SDQ; parent report; Goodman, 1997). The SDQ demonstrates good validity and reliability, as indicated by an internal consistency score of 0.63–0.85 and a high McDonald’s omega value (Deighton et al., 2014; Stone et al., 2015). The Cronbach’s alpha of this sample was 0.75.

2.4. Data analysis

2.4.1. Calculation of mood instability

Mood instability was quantified by calculating the root mean square of successive differences (RMSSD) scores. RMSSD is a reliable and valid index of instability as it considers both variability and temporal order of scores (Ebner-Priemer et al., 2009; Solhan et al., 2009). RMSSD is calculated by taking the square root of the average of the squared differences between affect at measurement i and i + 1 (Ebner-Priemer et al., 2009). An individual RMSSD score for each mood indicator (CES-D, POMS, PSS) was first calculated for participants for each of the three days at each gestational timepoint and a mean was calculated for each measure. Then, the RMSSD scores for each mood indicator were standardized and averaged for a mean RMSSD score for each gestational assessment. Because mean RMSSD scores did not differ across gestation (15-, 25-, and 35-weeks’; all p’s > 0.16) and because they were robustly correlated (range 0.38 to 0.59; See Supplemental Table S3 for all correlations and significance values), they were averaged to create a single composite pregnancy RMSSD score. A composite RMSSD score was similarly computed for the subset with the postnatal assessment. Two participants had RMSSD scores that were >3 standard deviations above the mean and they were excluded from the analyses. The participants completed 11 EMAs on average at each gestational timepoint. Because of this missingness, we assessed whether the amount of missingness correlated with the child outcomes. However, the amount of missingness did not correlate with the CBCL (r = −0.07; p = .49) nor the SDQ (r = 0.01; p = .91) and so was not included in further analyses.

To determine if mood instability was a unique predictor of child externalizing behaviors, and independent from measures of mood level (i.e., anxiety, stress, and depression), we calculated a negative mood level score for each participant. As with RMSSD, the prenatal negative mood level score was calculated by first taking the average of the EMA measure scores (CES-D, POMS, PSS) for each day. Then, the mood level score for each mood measure was standardized and averaged to get one composite mood level score for each gestational timepoint. Finally, these timepoint composites were averaged to create a single prenatal negative mood level composite. Additionally, a postnatal negative mood level score was calculated with the POMS Anxious mood subscale (Cranford et al., 2006), the 9-item short form CES-D (Santor and Coyne, 1997), and the 10-item PSS (Cohen et al., 1983), which were administered in the laboratory at two months postpartum. If participants were missing one of the three postnatal mood measures, the median of the existing measures was imputed to get a two-month postnatal mood level score. In total, 15 participants received imputed scores.

2.4.2. Analyses

First, Pearson correlations were conducted to examine associations between prenatal mood instability and child externalizing problems. Then, sensitivity analyses with linear regressions were conducted to determine whether mood instability predicted externalizing problems after consideration of possible third variable explanations of income-to-needs ratio (INR), postnatal mood instability, and pre and postnatal mood levels. These covariates were chosen given the well-established associations with child psychopathology (Glasheen et al., 2010; Goodman et al., 2011; Peverill et al., 2021). INR is a measure of poverty status, calculated by comparing income to the federal poverty line, taking into account the number of individuals dependent on that household income. Ratios below 1.0 indicate that income is below the federal poverty line, while ratios above 1.0 indicate that income is above the federal poverty line (Grieger et al., 2009; United States Census Bureau, 2025). Child age and sex at birth were also included in all models.

Because of the well-documented sex-dependent associations between exposures to early life adversity and neurodevelopment (Assini-Meytin et al., 2022; Sandman et al., 2013; Sutherland and Brunwasser, 2018; Wright and Schwartz, 2021), we examined whether associations between prenatal mood instability and externalizing problems differed by sex. No clear sex differences emerged and so these results are not discussed further (See Table S2).

3. Results

As shown in Table 2, higher levels of prenatal mood instability predicted more child externalizing problems at ages 3.8 and 4.5 (r = 0.34 and 0.29, respectively; both p < .05; Fig. 2). As expected, maternal negative mood level, both prenatal and postpartum, also were positively associated with externalizing symptoms (Table 2). Importantly, the sensitivity analyses indicated that mood instability remained predictive of child externalizing problems, when adjusting for both pre and postnatal mood levels (see Table 3). Accounting for INR also did not affect the association between mood instability and externalizing problems at either age.

Table 2.

Bivariate correlations mood instability, negative mood level, and child externalizing problems.

	1.	2.	3.	4.	5.
1. Prenatal mood RMSSD	-
2. Postnatal mood RMSSD	0.75	-
3. Prenatal mood level	0.53*	0.52*	-
4. Postnatal mood level	0.36*	0.42*	0.70*	-
5. Child externalizing problems at 3 years	0.34*	0.36*	0.38*	0.35*	-
6. Child externalizing problems at 4.5 years	0.29*	0.31*	0.27*	0.25*	0.44*

^*p < .05.

Fig. 2.

Bivariate associations between prenatal maternal mood instability and externalizing problems at 3.75 (measured by the CBCL) and 4.5 years of age (measured by the SDQ).

Table 3.

Linear regressions testing associations between mood instability and child externalizing problems.

	CBCL externalizing problems									SDQ Conduct
	Model 1a			Model 1b			Model 1c			Model 2a			Model 2b			Model 2c			Model 2d
	B	SE	β	B	SE	β	B	SE	β	B	SE	β	B	SE	β	B	SE	β	B	SE	β
Prenatal RMSSD	7.08	1.81	0.36**	4.13	2.14	0.21*	5.83	1.91	0.29**	1.04	0.31	0.29**	0.75	0.36	0.21**	1.10	0.33	0.29**	1.09	0.60	0.30*
Child sex	1.34	2.13	0.06	0.93	2.10	0.04	0.75	2.14	0.03	0.45	0.38	0.10	0.43	0.37	0.10	0.42	0.38	0.09	0.56	0.54	0.12
Child age	0.07	0.19	0.04	0.05	0.18	0.02	0.05	0.18	0.02	−0.00	0.02	−0.02	−0.00	0.02	−0.02	−0.01	0.02	−0.05	0.00	0.03	0.01
INR	0.00	0.00	0.02	0.00	0.00	0.03	0.00	0.00	0.03	0.00	0.00	−0.05	0.00	0.00	−0.03	0.00	0.00	−0.04	0.00	0.00	0.04
Prenatal mood level	-	-	-	3.96	1.61	0.27**	-	-	-	-	-	-	0.42	0.29	0.15	-	-	-	-	-	-
Postnatal mood level	-	-	-	-	-	-	3.33	1.25	0.25**	-	-	-	-	-	-	0.34	0.23	0.13	-	-	-
Postnatal RMSSD	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	0.38	0.45	0.14

Note. Because only 50 participants had both postnatal mood instability and externalizing behavior data at 3 years of age, postnatal mood instability sensitivity analyses were only conducted in the model testing externalizing behaviors at 4.5 years of age, where a larger number had the paired data (n = 70).

^*p < .10.

^**p < .05.

To elucidate whether mood instability during pregnancy uniquely predicts child externalizing symptoms and cannot be attributed to postnatal mood instability, a last sensitivity analysis was conducted (see Table 3). When modeled together with postnatal mood instability, prenatal mood instability continued to predict externalizing problems, supporting its independent role.

4. Discussion

The findings from this study, utilizing a robust measure of mood instability in the form of EMA, demonstrate that patterns of prenatal maternal mood may play a role in the development of child externalizing problems. Previous work on the influence of maternal mental health during pregnancy has focused almost exclusively on profiles of mood levels (e.g., depression, anxiety, stress) (c.f. Beydoun and Saftlas, 2008; Manzari et al., 2019; Schuurmans and Kurrasch, 2013; Van den Bergh et al., 2018). Here we show that the effects of mood instability are independent of pre and postnatal negative mood level, thus identifying a novel dimension of the prenatal environment with lasting developmental consequences. It is well-established that mood lability is a transdiagnostic risk factor for mental illness within an individual (Aldao et al., 2010; Fernandez et al., 2016; Stringaris and Goodman, 2009); these data provide new evidence that it may also play a role in intergenerational transmission of psychopathology.

It has previously been documented that exposure to prenatal mood fluctuations portends risk for internalizing disorders (Davis and Glynn, 2024; Glynn et al., 2018; MacNeill et al., 2023). Here we document links between unpredictability and child externalizing problems. This is a noteworthy observation as externalizing behaviors in childhood predict juvenile delinquency, as well as higher levels of crime, substance use, mental health problems, and adverse sexual or partner relationship outcomes, such as teenage pregnancy or involvement in domestic violence, in adulthood (Fergusson et al., 2005; Vitaro et al., 1992). These findings also dovetail with the existing literature indicating that unpredictability is consistently predictive of poorer effortful control (Davis and Glynn, 2024; Davis et al., 2019; Mittal et al., 2015; Short et al., 2023; Sturge-Apple et al., 2016), which is an established risk factor for externalizing problems (Eisenberg et al., 2015; Schoppe-Sullivan et al., 2009; Wills et al., 2016).

The strengths of this study include the prospective longitudinal design with assessments spanning the prenatal period through early childhood. Further, the use of EMA to assess mood instability in real world contexts represents an important addition to the small existing literature, which to date has largely employed mood entropy, computed by standardized mood questionnaires administered at a single time point. One limitation of this investigation is the moderate sample size for the assessment of postnatal mood instability. Despite this relatively small sample size, it was still possible to discern independent contributions of prenatal mood instability beyond postnatal mood instability, which is consistent with previous work examining perinatal influences of prenatal maternal mood lability on child development (Glynn et al., 2018; Howland et al., 2021). A second limitation concerns the correlational nature of the study, which relies on natural variations in maternal mood instability, rather than those that are experimentally manipulated. We cannot therefore rule out alternative explanations such as inheritance of psychopathology associated with mood dysregulation. We do believe that this concern is somewhat mitigated by the fact that mood instability predicted child externalizing problems after adjustment for dimensional assessment of maternal mood symptoms and by findings from experimental animal models demonstrating that exposure to unpredictable maternal signals in early development exert persisting influences on offspring brain and behavior (Brunson et al., 2005; Ivy et al., 2008; Molet et al., 2016).

A key next step awaiting investigation is identification of mechanisms underlying the association between prenatal maternal mood instability and child neurodevelopment. A reasonable approach is to leverage the efforts in this domain related to how mood levels (e.g., maternal stress or depression) shape fetal development with possible mechanisms including endocrine, immune, epigenetic, and vascular pathways (Christian, 2014; Hilmert et al., 2014; O’Donnell et al., 2009; Shea et al., 2008). Relatedly, probing of the neurobiological changes that underly behavioral phenotypes, such as externalizing behaviors, are warranted. To this end, recently we reported evidence that exposure to greater maternal prenatal mood entropy is associated with salience network integrity in adolescence (Jirsaraie et al., 2024). These associations were specific to the salience network, which facilitates socioemotional regulation (Toller et al., 2018), as well as switching between attentional states (Goulden et al., 2014), and were not observed for other intrinsic connectivity networks. A last exciting next direction involves more detailed examination of additional emotion dynamical features (e.g., inertia, emotional granularity) which could enrich the understanding of how parental emotional states shape child development and also speak more specifically to the potential added value of the study of affect dynamics beyond a focus on more conventional measures of mood (e.g., levels) in this domain (Dejonckheere et al., 2019; Kuppens and Verduyn, 2015).

There is little dispute that exposures in early life, both deleterious and salutary, exert lasting influences on lifetime health and development. This awareness begs attention focused on prevention and intervention to promote healthy development. Here we identify a novel exposure, maternal mood instability, that appears to exert persisting influences on child neurodevelopment. These observations suggest a new opportunity to improve child health and development because mood lability is amenable to intervention (Agapoff et al., 2023; Hafeman et al., 2020) and therefore represents an additional potential target to augment current therapies aimed at improving parental mental health (Acri and Hoagwood, 2015; Cluxton-Keller et al., 2015).

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jad.2025.119947.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.