Maternal Posttraumatic Stress Predicts Mother-Child Symptom Flare-Ups Over Time

Abstract

Although concurrent associations between parent and child posttraumatic stress symptoms (PTSS) have been well-documented, few longitudinal studies have examined bidirectional influences by modeling the effects of both parent and child PTSS simultaneously over time. The current study examines patterns of PTSS in children and their mothers beginning in preschool and continuing through elementary school age (ages 4–9 years) in a large, heterogeneous sample (N = 331 mother-child dyads). Mothers reported on their own and their child’s posttraumatic stress symptoms. A random intercept cross-lagged panel model (RI-CLPM) was used to examine associations between symptoms across six time points. Results indicated that maternal and child symptoms were associated with each other at concurrent time points and tended to fluctuate in a synchronized manner relative to their overall mean symptom levels. Longitudinal cross-lagged paths were significant from mother to child, but non-significant from child to mother, suggesting that mothers’ symptom fluctuation at one time point predicted significant fluctuation in children’s symptoms at the subsequent time point. The concurrent co-variation of maternal and child symptoms and the predictive nature of maternal symptom flare-ups have important implications for both maternal and child mental health interventions and underscore the importance of attending to mothers’ symptomatology early in treatment.

Introduction

Associations between parent and child posttraumatic stress symptoms (PTSS) have been identified and well-documented in a number of meta-analyses and reviews (Lambert et al., 2014; Leen-Feldner et al., 2013; Morris et al., 2012). The numerous cross-sectional studies included in these reviews provide strong support for intergenerational trauma-related risk, but do not shed light on the direction of associations or reciprocal influences that trauma-related symptom patterns may have within a family system.

In an early effort to understand relational effects in trauma exposed families, (Scheeringa & Zeanah, 2001) proposed a model of posttraumatic symptomatology for parents and young children. Acknowledging that parent and child symptoms may be driven by shared traumatic experiences (e.g., domestic violence exposure), they also suggest a compound pattern of interaction in which parents’ symptoms and responses might exacerbate those of their children, regardless of whether they experienced the same or different traumatic events. The authors describe three possible patterns. First, children’s distress may be exacerbated if their parents become withdrawn – as a direct result of their PTSS or because their child’s trauma serves as a painful reminder of their own traumatic experience. For example, when a mother’s traumatic impairment causes her to be less available to respond sensitively to her child’s distress, this might serve to intensify her child’s symptomatology. Second, parents may engage in overprotective or restrictive parenting behaviors as a result of fears generated from their own traumatic experiences or from worry that their child may be re-victimized; e.g., a mother in response to her child’s sexual trauma. Third, a parent’s preoccupation with and non-constructive references to the child’s traumatic experience may serve to trigger and maintain the child’s PTSS. We suggest two additional, child-driven, patterns. One, children’s trauma-related symptoms may lead to parents’ feelings of distress, guilt or helplessness and contribute to or exacerbate their parent’s PTSS. Two, children’s PTSS may manifest in behavior that is more aggressive, irritable, and/or less positive toward the parent. Of course, these theorized mechanisms are not mutually exclusive.

Consistent with a relational model, parent and child PTSS tend to be correlated at synchronous time points (Lambert et al., 2014; Morris et al., 2012), with some suggestion that these associations may become stronger at later time points (Koplewicz et al., 2002; Ostrowski et al., 2006). Further, numerous studies have documented associations among parents’ traumatic experiences and impaired parenting behaviors and parent responsivity (Greene, Haisley, et al., 2020; van Ee et al., 2016), thus providing empirical support for an indirect mechanism by which the model hypothesizes that parents’ symptoms may influence children’s symptoms. Surprisingly, although emotional contagion (Hatfield et al., 2009), or the co-regulation of affective states, has been documented with PTSS among couples in the context of a child’s cancer diagnosis and treatment (Wikman et al., 2017), it has not been examined among parents and children. Indeed, few longitudinal studies have explored the effects of both parent and child posttraumatic stress symptoms simultaneously over time in order to examine these bidirectional influences among parent-child dyads (Koplewicz et al., 2002; Landolt et al., 2012; Nugent et al., 2006; Shi et al., 2018).

Among this small body of literature, there have been mixed findings. In two studies of pediatric injury or illness diagnosis, parent PTSS predicted children’s PTSS up to 1 year later, but children’s symptoms were not found to predict parent symptoms (Landolt et al., 2012; Nugent et al., 2006). In contrast, among 22 school-age children who were trapped without their parents during the 1993 World Trade Center bombing, children’s symptoms assessed 3 months following the attack predicted their parents’ symptoms 6 months later; however, parent symptoms at the initial assessment did not predict their children’s subsequent symptoms (Koplewicz et al., 2002). One study reported bidirectional effects. Among a large sample (n = 688) of parents and adolescents with moderate PTSS assessed 12- and 18-months following a natural disaster in China (Shi et al., 2018), maternal and paternal symptoms prospectively predicted youth symptoms, whereas youth symptoms predicted maternal, but not paternal, symptoms.

The timing of assessments might be contributing to the varied results among these longitudinal patterns of PTSS. The presentation and intensity of individuals’ symptoms are not typically static over time, as periods of exacerbation and remittance are common with PTSS (Bryant et al., 2013). Experiences of symptom fluctuations or “flare-ups” are particularly important when considering bidirectional effects within a family environment, where there is the potential for symptoms in one dyad member to influence or exacerbate symptoms in the other. Yet, in each of the parent-child PTSS studies described above, symptoms were assessed at only two time points, precluding the examination of fluctuations in symptoms over time.

It is also possible that the extent to which parents’ and children’s symptoms influence each other differs across development. Previous bidirectional studies of parents’ and children’s posttraumatic stress symptoms have frequently included samples of children ranging from school age through adolescence (Landolt et al., 2012; Nugent et al., 2006). Yet the interactions between parents and younger and older children can differ greatly. Examining children and adolescents together in the same study may obscure these developmental variations.

Current Study

The current study builds upon the previous literature in this area and addresses the abovementioned limitations by examining bidirectional patterns of PTSS in children and their mothers over the course of early and middle childhood in a large, heterogeneous sample. Specifically, ratings of maternal and child PTSS were collected from preschool age - elementary school age (ages 4–9). Both mothers’ and children’s symptoms were assessed, and their bidirectional influence on each other were evaluated, across the six time points.

Importantly, we approached this question using an innovative analytic design, the random intercept cross-lagged panel model (RI-CLPM) which allows for the examination of bidirectional relationships between mothers’ and children’s PTSS, while also accounting for their overall average level of PTSS, an important feature that is not captured in traditional cross-lagged models (Hamaker et al., 2015). RI-CLPM can be used to characterize fluctuations in symptoms over time but requires three or more time points to do so. In previous studies that have used only two time points, such as those described above, it was not possible to differentiate between subject (overall symptom level) from within subject change, a necessary step in determining whether bidirectional effects between mothers and children are estimated without statistical bias.

Based on previous literature we expected that mothers’ and children’s symptoms would be correlated at each time point. Further, we hypothesized that mother and child symptoms at each time point would predict their own symptoms at the next time point (the auto-regressive paths). However, because of the mixed, and limited, previous findings regarding parent-to-child and child-to-parent effects of PTSS, as well as the use of only two time points in previous studies, we did not have an a priori hypothesis regarding the cross-lagged associations.

Methods

Participants

Data from this project are drawn from the Multidimensional Assessment of Preschoolers (MAPS) Study. Families with a child between the ages of 3 and 5 years were recruited from pediatric primary care practices to establish a representative and diverse sample (N = 1,857). Study staff met with mothers to answer questions and obtain informed consent. All procedures were approved by the Northwestern Feinberg School of Medicine and UConn Health Institutional Review Boards. The analytic sample for the current study consists of a subsample of mothers and children who were invited to participate in more intensive in-person and longitudinal assessments. As the goal of the larger study was to investigate behavioral and environmental risk for psychopathology, families participating in the longitudinal study were oversampled based on past-year intimate partner violence (IPV) or elevated child disruptive behavior (Nichols et al., 2015). All of the mothers in the current study subsample (N = 331) reported that they had been exposed to at least one interpersonal trauma (e.g., physical, emotional, or sexual abuse in childhood, or intimate partner violence); 88% of the children also were reported to have been exposed to interpersonal and/or non-interpersonal trauma.

The current study utilized data from five assessment waves. Child symptoms were assessed at all five waves and maternal symptoms at four waves. At each time point, a wide range of ages were possible and the time lag between assessments also varied for participants. As a result, we utilized an approach suggested by (Singer & Willet, 2003) to most accurately reflect development effects and minimize the impact of variable age/assessment lag, all data was recoded into age bands rather than assessment wave (e.g., a hypothetical participant age 4 at the initial time point may have been as young as 8 by the final time point and therefore missing data at age 9 whereas another participant aged 7 at baseline would have missing data from ages 4, 5, and 6). The resulting sample size of available data from ages 4 to 9 are given in Table 2. The vast majority of participants (N = 274; 82.8%) included in the final analyses had at least five points of data.

Table 2.

Descriptive Statistics and Correlations.

	Child PTSS						Mother PTSS
	Age 4	Age 5	Age 6	Age 7	Age 8	Age 9	Age 4	Age 5	Age 6	Age 7	Age 8	Age 9
C Age 4	-	.64 ***	.53***	.68***	.41***	.35	.04	.58 ***	.37***	.29	.18	..09
C Age 5		-	.46 ***	.71***	.62***	.46***	.48 ***	.59***	.35 ***	.41***	.39***	.25
C Age 6			-	.60 ***	.36***	.47***	.27**	.32 ***	.38***	.43 **	.15	−.02
C Age 7				-	.69 ***	.62***	.34**	.34***	.32*	.29**	.23	.06
C Age 8					-	.50 ***	.23*	.21*	.30***	.23	.37***	.35 **
C Age 9						-	.15	.12	.29*	.44**	.28	.23*
P Age 4							-	.48 ***	.23**	.26	.24*	.24
P Age 5								-	.41 ***	.53***	.49***	.48***
P Age 6									-	.56 **	.30***	.35***
P Age 7										-	.56 ***	.51***
P Age 8											-	.56 ***
P Age 9												-
M	34.08	32.97	31.96	32.54	33.43	32.75	18.03	9.54	9.02	9.38	10.75	9.35
SD	8.56	7.84	6.46	8.61	8.08	6.90	11.65	4.50	4.18	4.56	5.16	4.06
N	164	180	197	134	169	105	197	172	183	111	156	108

^*p < .05.

^**p < .01.

^***p < .001

Note: Bolded coefficients indicate stability from T to T+1. Bolded and Italicized coefficients indicate lag predictive effects from Parent to Child and Child to Parent. Means and SD are given for the PCL and TSCYC without a log transformation. Correlations are given with the log transformation. Because Ns vary at each time point some correlation coefficients are significant whereas others are not simply because of the N (e.g., child symptoms at age 8 and mother symptoms at age 5 [r = .21; N=100] have a statistically significant, but smaller correlation coefficient compared to child symptoms at age 8 and mother symptoms at age 7 [r=.23; N =51]).

Upon enrollment in the study, children were on average 4.71 years of age (SD = 0.91) and their mothers where 30.90 years of age (SD = 6.15). The sample contained about as many girls as boys (52.0% female). The majority of children were identified as minorities, specifically, 163 (49.2%) identified as Black, 87 (26.3%) as Latinx, and 78 (23.6%) as White. At study entry 13 (3.9%) mothers had not completed high school, 63 (19%) had a high school diploma or GED, 113 (34.1%) had attended some college, 59 (17.8%) had an Associate degree, 50 (15.1%) had a Bachelor’s degree, and 26(7.9%) had a graduate degree. Two hundred and eleven (63.7%) mothers reported being married or living with a partner and 82 (24.8%) reported being single. Across a variety of income and socioeconomic indicators, 217 mothers (65.6%) were considered to be near poor or poor financially.

Measures

Trauma Symptom Checklist for Young Children (TSCYC; Briere et al., 2001) is a widely used, norm-referenced, adult report measure of trauma-related symptoms that is developmentally appropriate for use with 3 to 12 year-old children. The TSCYC can be used to assess traumatic stress symptoms in normative and clinical populations (Wherry & Dunlop, 2018) and discriminated between maltreated and non-maltreated children in previous studies (e.g., Augusti et al., 2018; Milot et al., 2010; Tierolf et al., 2018). The measure has demonstrated good internal consistency with clinical scale alphas ranging from .81 to .93 in a clinical sample of maltreated young children (Briere et al., 2001). In the current study, we used the TSCYC posttraumatic stress total symptom score, which consists of 27 items assessing the core PTSD symptoms of arousal (e.g., “Being easily startled”), avoidance (e.g., “Not wanting to talk about something that happened to him/her”), and intrusion (e.g., “Bad dreams or nightmares”). Mothers rated their child’s symptoms on a 4-point scale from “not at all” (1) to “very often” (4). These were summed to create a total posttraumatic stress symptom score and converted to a T-score using published norms. Internal consistency for the TSCYC had a median of .85 and ranged from .84 to .92 across assessments.

PTSD Checklist (PCL; Weathers et al., 1993) is a 17 item self-report measure of posttraumatic stress symptoms originally based on DSM criteria. The PCL is a widely used self-report measure used for assessing and monitoring posttraumatic stress symptoms. Across numerous samples of adults it has demonstrated internal consistency (alphas ranging from .80 – .96; Wilkins et al., 2011) and convergent validity with a structured clinical interview for PTSD (Keen et al., 2008). In the current study, we used a 6-item version that queries the core symptom components of PTSD (re-experiencing, reaction to reminders, avoidance of activities, distance from others, changes in emotionality, and concentration deficits; e.g., “Feeling very upset when something reminded you of a stressful experience from the past?”). Researchers have previously found the 6-item version used in the current study to correlate with the full version (.92) and demonstrate high sensitivity (.92) (Lang et al., 2012). Items were rated on a 5-point scale from “not at all” (1) to “extremely” (5). Internal consistency for our PCL-6 had a median of .88 and ranged from .85 to .89 across assessments.

Data analysis

We examined associations among symptom scores and demographic variables (Table 1). We also reported descriptive statistics and bivariate correlations among maternal and child symptoms (Table 2) which was done to establish stability of symptoms and provide unadjusted estimates of maternal symptoms predicting child symptoms (and vice versa). The RI-CLPMs were examined using Mplus version 8.3 (Muthén & Muthén, 2017) and specified according to previous articles (e.g., Hamaker et al., 2015; Mulder & Hamaker, 2020). The goal of the RI-CLPM is to demarcate between and within subject effects using a latent variable that reflects each individual’s average level of PTSS across time, which acts to capture between subject differences. The remaining variability is time-point specific and reflects within subject effects. Supplemental materials provide more details about the RI-CLPM and its advantages over traditional CLPM (see also Berry & Willoughby, 2017; Hamaker et al., 2015; Mulder & Hamaker, 2020; Mund & Nestler, 2019).

Table 1.

Demographics and symptoms at each age period

Symptoms by age band		Child sex		Poverty		Minority status		Education
		t(df)	p	t(df)	p	t(df)	p	t(df)	p
Child PTSS	Age 4	0.17 (162)	.866	−1.43 (162)	.155	−0.56 (162)	.579	1.98 (162)	.049
	Age 5	0.43 (178)	.667	−0.40 (178)	.688	−0.53 (178)	.594	1.78 (178)	.077
	Age 6	−0.29 (197)	.773	−1.61 (197)	.109	−1.23 (197)	.220	1.40 (197)	.163
	Age 7	3.02 (137)	.003	−1.34 (137)	.181	−0.16 (137)	.875	0.79 (137)	.429
	Age 8	0.28 (181)	.777	−0.31 (181)	.760	−0.34 (181)	.737	0.39 (181)	.696
	Age 9	−0.77 (105)	.441	−1.01 (105)	.317	0.16 (105)	.877	−0.99 (105)	.323
Mother PTSS	Age 4	0.34 (190)	.736	0.24 (190)	.814	−0.25 (190)	.804	0.60 (190)	.550
	Age 5	1.09 (167)	.279	0.24 (167)	.811	−0.04 (167)	.967	−0.48 (167)	.635
	Age 6	−2.11 (177)	.036	−0.37 (177)	.710	0.36 (177)	.722	−0.22 (177)	.828
	Age 7	−0.98 (102)	.329	−0.54 (102)	.592	−0.52 (102)	.601	0.84 (102)	.405
	Age 8	−1.05 (168)	.298	−1.12 (168)	.265	−0.73 (168)	.466	0.97 (168)	.333
	Age 9	−1.31 (107)	.192	−0.55 (107)	.585	−0.70 (107)	.483	0.11 (107)	.911

Note: Child sex compared boys (0) vs. girls (1); poverty compared those rated as poor or near poor (1) vs. not (0); minority status compared those identifying as any racial/ethnic minority (0) vs. White non-Hispanic (1); education compared those with less than a college degree (0) to those with a college degree (1). Degrees of freedom are in parentheses and differ by analysis due to missing data from some participants.

To address skewness, we log transformed the PCL and TSCYC and estimated the models using maximum likelihood with robust standard errors (MLR) to address missing data. Following Hamaker et al., (2015), prior to fitting the RI-CLPM we constrained the means of child and mother symptoms to be equal, while allowing all variables to correlate, in order to ease interpretation of the RI-CLPM. A fully constrained means-only model would not converge. A closer inspection of the symptom means suggested means were equivalent for children at ages 5, 6, 7, 8, and 9 (all time points except for age 4, which had higher symptom scores), but that mother symptoms means were never equivalent. Therefore, subsequent models constrained means at ages 5, 6, 7, 8, and 9 for children only. We next ran a series of 5 RI-CLPMs which varied whether autoregressive, cross-lagged, and within time point covariances (after the first time point, often termed innovations) were freed or constrained to be equal, following a model fitting strategy used by RI-CLPM experts (for an example see, Oh et al., 2020). The final model was selected by using the (Satorra & Bentler, 2010) chi-square test for nested models estimated using MLR. Mplus syntax for all models are available in Supplemental materials.

Results

Table 1 presents associations among maternal and child symptoms and child sex, poverty status, racial ethnic minority status, and maternal education. Only 2 of these associations were significant; given the large number of comparisons (48) and the lack of consistent findings across any age range these demographic variables were not included in our model. Table 2 contains the means, standard deviations, and correlations among mother and child PTSS. As can be seen in the Table, both mother and child symptoms exhibited significant stability over time. Moreover, the correlation coefficients suggest the possibility of bi-directional lagged effects as maternal symptoms were often significant predictors of child symptoms at the following time points and, similarly, child symptoms were often significant predictors of maternal symptoms at the subsequent time point.

RI-CLPM model fit

We examined 5 RI-CLPM in which we initially constrained all like paths to be equal and in subsequent models varied which paths were freed vs. constrained (Table 3). Model 1, with all like paths constrained, had acceptable fit according to CFI, TLI, and RMSEA. Furthermore, Model 1 fit better compared to all other models that freed additional paths, according to the Satorra-Bentler chi-square test. As a result, we interpret the more parsimonious, fully constrained model. A fully constrained model means that like autoregressive, covariances, and cross-lagged paths are held equal across time and freeing those paths (i.e. allowing them to vary across time) did not result in improved model fit.

Table 3.

RI-CLPM model fitting.

Model	AR Paths	Lagged Paths	Covariances	χ2	p	df	CFI	TLI	RMSEA	Comp Model	SB χ2	SB df	SB χ2 pval
1	Constrained	Constrained	Constrained	96.82	0.002	61	0.92	0.91	0.042
2	Constrained	Constrained	Free	90.22	0.003	57	0.92	0.91	0.042	1 v. 2	6.55	4	0.16
3	Constrained	Free	Free	79.98	0.003	49	0.93	0.90	0.044	1 v. 3	17.16	12	0.14
4	Free	Constrained	Free	81.77	0.002	49	0.92	0.90	0.045	1 v. 4	16.23	12	0.18
5	Free	Free	Free	67.29	0.006	41	0.94	0.90	0.044	1 v. 5	29.97	20	0.07

AR paths = autoregressive paths; Covariances = within age band covariances excluding age 4 (innovations); Comp Model = models being compared; SB = Satorra-Bentler. Model 1 was selected as most parsimonious as a result of non-significant SB test when compared to all other models.

Mother-Child PTSS across time

To formally examine the possibility of bi-directional effects, we examined the RI-CLPM (Figure 1). The relationship between the random intercept for mother and child symptoms was positive and significant, suggesting that mothers with higher symptoms also had a child with higher symptoms (unstandardized b = 0.003, SE = 0.001, z = 3.43, p < .001). Additionally, the cross-sectional time specific covariances (innovations) were also significant (unstandardized b = 0.002, SE < 0.001, z = 4.06, p < .001), which indicates that mothers and children tended to experience fluctuations in symptoms at the same time; i.e., at a given time point, both mother and child may have simultaneously had higher (or lower) levels of symptoms relative to their typical symptom levels. The autoregressive paths were significant for mothers, suggesting that within subject fluctuations in symptoms at age X predict fluctuations at age X+1 (unstandardized b = 0.27, SE = 0.06, z = 4.52, p < .001); i.e., mothers experiencing greater symptoms relative to their average symptom level tended to experience a “flare up” at the subsequent time point. However, within-subject autoregressive paths were not significant for children. Additionally, there were significant cross-lagged paths from mothers to children (unstandardized b = 0.10, SE = 0.027, z = 3.75, p < .001), this suggests that a mother’s symptom fluctuation at age X was associated with a significant fluctuation in their child’s symptoms at age X+1. Lastly, the cross-lagged path from children to mothers was not significant.

Figure 1. Overall Random Intercept Cross-Lagged Panel Model.

Significant substantive pathways are shown in black and labeled with unstandardized coefficients; N=331.

Discussion

Although bidirectional associations between parent and child mental health symptoms have been observed and documented (Bornstein, 2016), few studies have examined these patterns among posttraumatic stress symptoms (PTSS). Indeed, dyadic associations of parent and child PTSS have been almost exclusively tested with cross-sectional data. Therefore, the current study examined these patterns among a large, ethnically and racially heterogeneous sample of mothers and children over the course of preschool and elementary school. Study findings suggest a number of intrafamilial associations.

Importantly, the current data suggest significant co-variation of symptoms between mothers and their children over time. As noted above, it is not uncommon for PTSS to ebb and flow (Bryant et al., 2013). In this sample of mothers and children, these periods of exacerbation and remittance tend to be synchronized. It is possible that this synchrony could be due to shared external factors influencing fluctuations in both mother and child symptoms simultaneously (e.g., socioeconomic status, family social support). However, the results indicate that there is directionality to this co-variation, with the mother serving as the driving force of the subsequent symptom flare-ups for herself and her child. That is, an increase in maternal symptoms at one time point (as compared to her average level of symptoms) predicted both an increase in her own symptoms as well as an increase in her child’s symptoms at the subsequent time point. Conversely, children’s symptom fluctuations were not associated with either their own or their mother’s symptoms at subsequent time points. These findings extend previous cross-sectional examinations that have linked parental PTSS with children’s general distress, but found no association between children’s PTSS and their parents’ general distress (Juth et al., 2015).

It is possible, then, that this co-variation could be the result of emotional contagion, in which the mother is “driving” these patterns of fluctuation, and the child is mimicking and/or synchronizing with their mother’s emotional state (Hatfield et al., 2009). A similar pattern of physiological contagion from mothers to their children has been demonstrated during a lab-based emotional challenge task (Shih et al., 2018). These results are also consistent with theories of emotion socialization, which include parental modeling as a robust mechanism by which children learn to manage their emotions (Eisenberg et al., 1998).

It is interesting to consider possible explanations for why the pattern of effects are unidirectional from mother to child. First, this study examined these patterns among younger children who may be particularly vulnerable to the effects of parental PTSS. Disruptions in parental functioning due to psychopathology may interfere with mothers’ ability to effectively support their children’s development of adaptive emotion regulation skills and other key socioemotional tasks of childhood. Further, younger children spend more time with their parents, thus providing greater opportunity to be influenced by their emotions. Indeed, in studies of maternal depression, the association between maternal symptoms and children’s internalizing and externalizing behavior and negative affect is stronger in samples of younger children (Goodman et al., 2011). In contrast, children’s flare-ups did not predict mother’s flare-ups at the subsequent time point. It may be that, while children may become emotionally dysregulated in response to their mother’s symptom increase, for mothers, the shift in focus to supporting their child through a symptom flare-up may serve as an emotionally regulating form of attentional redeployment. This is consistent with (Kouros & Garber, 2010) who found that adolescent depressive symptoms were associated with subsequent decreases in their mother’s depressive symptoms.

Although the current study did not examine mechanisms that might explain the influence of mothers’ symptom fluctuations on their children, both positive and negative parenting behaviors, as well as the parent-child relationship, may be influenced by parent PTSS (Grasso et al., 2016; Greene, McCarthy, et al., 2020; van Ee et al., 2016). In turn, a growing body of research emphasizes the role of parenting behaviors in children’s response to and recovery from trauma (Williamson et al., 2017). In addition, children’s traumatic experiences may be painful reminders of mothers’ past experiences that trigger their own posttraumatic reactions, which in turn exacerbate their child’s symptoms (Scheeringa & Zeanah, 2001).

The current findings should be considered within the context of several limitations. It is well established that young children lack the verbal and cognitive capacities to reliably report about their own symptoms, and this is particularly evident in the preschool period when this longitudinal study began (Briggs-Gowan et al., 2016). For this reason, the data reported here were based on maternal report and thus potentially subject to shared method variance. In particular, mothers’ reports of their children’s symptoms may be influenced by their own mental health status. A substantial literature examines the question of psychopathology-related parental reporting bias (e.g., Ordway, 2011; Richters, 1992). Notably, a recent inquiry that makes use of a novel measurement invariance approach to distinguish between associations between maternal and child psychopathological symptoms and mother’s reporting biases found only small associations between numerous dimensions of maternal psychopathology and biases in reporting their children’s emotional and behavioral symptoms (Olino et al., 2021). This limitation is further tempered by the fact that because there is evidence of robust associations between trauma exposure and symptoms in young children (Briggs-Gowan et al., 2010) and evidence that early childhood trauma-related symptoms can predict school-age psychopathology (Briggs-Gowan et al., 2012), it is important to include these early years in longitudinal examinations of symptoms. Therefore, we have opted to rely on maternal report in order to examine the patterns of mother-child symptoms throughout the range of these early years. Finally, children’s symptoms may similarly be influenced by paternal mental health, which was not assessed in this study. Future research would benefit from multi-informant reports, clinical assessment of children’s symptoms, and the inclusion of fathers. Additionally, future research examining these reciprocal relationships would be improved by incorporating time varying covariates at each symptom assessment, including new and ongoing traumatic events, other environmental stressors, and changes in social support that may be associated with symptom fluctuations. Unfortunately, the current data did not include a single set of time varying covariates measured consistently at each assessment.

This is the first study to examine the fluctuation of PTSS among intergenerationally traumatized families over an extended period of time and provide evidence of emotional contagion from mother to child. The co-regulation of maternal and child symptoms and the predictive nature of maternal symptom flare-ups have important implications for both maternal and child mental health interventions and underscore the importance of attending to mothers’ symptomatology early in treatment. With regular monitoring of maternal symptoms, it may be possible to detect when mothers require additional support in order to prevent subsequent flare-ups in their own and their children’s symptoms. Further, because maternal mental health symptoms can interfere with children’s response to therapeutic interventions for PTSS (Martin, 2019), addressing maternal symptom flare-ups has the potential to enhance children’s recovery.