Children of depressed mothers suffer various emotional and behavioural difficulties, but little is known about the influence of the timing of children's initial exposure to maternal depression. Furthermore, previous research has not consistently addressed the question of specificity by considering whether the timing is differentially associated with internalising or externalising symptoms. These issues were addressed in a prospective study of kindergarten children who had been initially exposed to maternal depression in infancy, the toddler/pre-school period, or never, and who were rated by their teachers as evidencing high levels of ‘pure’ internalising, ‘pure’ externalising, or co-occurring symptoms.
It is well established that offspring of depressed mothers suffer a variety of adverse outcomes (reviewed by Reference Cummings and DaviesCummings & Davies, 1994; Reference Gotlib, Goodman, Silverman and OllendickGotlib & Goodman, 1999). Infants of depressed mothers have been found to be more withdrawn, to have more difficult temperaments, and to be less securely attached to their mothers (Reference FieldField, 1992; Reference Murray, Fiori-Cowley and HooperMurray et al, 1996; Reference Campbell, Cohn, Murray and CooperCampbell & Cohn, 1997). Toddlers and pre-school children of depressed mothers have been found to have more difficulties with self-regulation and in relationships with peers and adults (Reference Lyons-Ruth, Cicchetti and TothLyons-Ruth, 1992; Reference Cicchetti, Rogosch and TothCicchetti et al, 1998). School-age children of depressed mothers have been found to have even more difficulties (Reference Anderson and HammenAnderson & Hammen, 1993; Reference Fergusson, Horwood and LynskeyFergusson et al, 1995).
With few exceptions (for example, Reference Sinclair and MurraySinclair & Murray, 1998), prior studies have not employed a developmental perspective or prospective longitudinal designs of sufficient duration to investigate the association of earlier maternal depression with children's problems at later points, and none has specifically looked at the timing of depression. Timing is critical because it may interfere with mastery of developmental tasks such as attachment and emotion regulation during infancy, or self-regulation and social competence in the toddler/pre-school years (reviewed by Reference Goodman and GotlibGoodman & Gotlib, 1999). Also, to our knowledge, studies have not distinguished between child outcomes with co-occurring internalising and externalising symptoms, and those with ‘pure’ forms of these symptoms.
The study addressed these issues by investigating the association of the timing of maternal depression with child symptoms during the spring of the children's kindergarten year, an important developmental point when vulnerabilities established earlier may become risk factors as children negotiate new relationships and the increased structure of school (Reference Entwisle, Alexander, Namboodiri and CorwinEntwisle & Alexander, 1989). We also distinguished between ‘pure’ internalising, ‘pure’ externalising, and co-occurring symptoms. We addressed three questions. First, are the risks of internalising and externalising symptoms associated with differences in the timing of children's initial exposure to maternal depression? Second, is the effect of the timing different if ‘pure’ forms of symptoms are distinguished from co-occurring forms? And third, do symptom patterns vary according to child gender or family socio-economic status as indexed by maternal education?
Two major hypotheses were tested:
(a) Exposure to maternal depression in infancy would be associated with internalising symptoms, either alone or co-occurring with externalising symptoms, because this is a ‘sensitive’ period for attachment and the development of emotional regulation.
(b) Initial exposure to maternal depression in the toddler/pre-school years would be associated with externalising symptoms alone, because this is a period when children learn to self-regulate and gain social skills.
Women were recruited from prenatal clinics in two Midwestern cities in the USA. Participants were required to be over 18 years old, in the second trimester of pregnancy (12-21 weeks), and, because the original focus of the project was on issues of family and work, to be living with the child's father, and either employed or a homemaker — see Hyde et al (Reference Hyde, Klein and Essex1995) for details. Of those who met inclusion criteria, 570 (75%) consented to participate; 560 had live births. The analyses presented here are of 421 mothers (75% of the live births) whose children were attending formal kindergarten programmes and whose teachers provided ratings.
The sample of 421 mothers used in the study was composed of 6% ethnic minorities, 96% married, and 40% first-time mothers. The age range was 20-43 years, median 30 years. Two per cent had less than a high-school degree, 42% were high-school graduates, 35% were college graduates and 21% had some education beyond college. Annual family income ranged from $7500 to $200 000, median $47 000. There were no significant differences between the 421 families who are included in these analyses and the 139 who comprised the remainder of the original eligible sample in terms of ethnicity, family income, maternal education or parity, but the participating mothers were more frequently married (96% v. 91%) and older (30 years v. 28 years).
Maternal major depression
The depression section of the National Institute of Mental Health Diagnostic Interview Schedule (DIS) (Reference Robins, Heizer and CroughanRobins et al, 1981) was administered in home interviews to diagnose maternal major depression during the first year post-partum and when the children were aged 3½ and 4½ years. Responses were scored for the presence of major depression according to the DSM—III—R requirement of symptoms corresponding to five or more criteria, including either depressed feelings or loss of interest (American Psychiatric Association, 1987). Two indices of the timing of children's initial exposure were then derived:
(a) the presence of maternal major depression during infancy, defined as any time from the child's birth to age 12 months;
(b) only in the absence of depression during infancy, its presence during the toddler/pre-school period (age 2-4½ years; the DIS was not given between years 1 and 2).
For each index, 1 denotes initial exposure to maternal depression during that period and 0 denotes that the child was not first exposed during that period. Children who were never exposed to maternal major depression had scores of 0 on both indices.
Teacher ratings of child internalising and externalising symptoms
Teachers completed the Health and Behaviour Questionnaire (HBQ) in the spring of the kindergarten year. The HBQ is a multi-dimensional and multi-informant instrument for middle childhood assessment (described fully by Reference Essex, Boyce and GoldsteinEssex et al, 2001). The mental health symptom items used in the study were derived from the Ontario Child Health Study scales — revised (OCHS—R), well-established measures with known reliability and validity (Reference Boyle, Offord and RacineBoyle et al, 1993). The teacher form included a 14-item internalising scale (sub-scales of depression and overanxious disorder) and a 24-item externalising scale (sub-scales of oppositional defiant and conduct problems, and overt aggression). Internal consistencies for all scales and sub-scales exceeded 0.70 and test—retest reliabilities ranged from 0.71 to 0.94. A cut-off point of the upper 25% of the distributions on the internalising and externalising scales was used to define the children evidencing ‘high’ symptoms. This cut-off point was selected to ensure that the children's scores fell within 1 standard deviation of the mean for clinic samples in the multi-site case—control study that established the effectiveness of the HBQ in discriminating between clinic-referred and community groups of young children (Reference Ablow, Measelle and KraemerAblowet al, 1999). Effect sizes of 1.0 (internalising) and 1.3 (externalising) were found for discriminations between community and clinic-referred children. In addition, in a selected sample of children from the present study, teacher HBQ ratings of externalising symptoms discriminated between children who were ‘high’ and ‘low’ on a parent-rated composite score for temperamental anger, activity, low control and poor attention (effect sizes from 1.3 to 2.1). Teacher internalising ratings on the HBQ, however, did not identify children who were temperamentally fearful or sad.
Child gender and maternal educational level
Measures of child gender (boys, 0; girls, 1) and maternal education (high school or less, 0; more than high school, 1) were also included, based on prior findings demonstrating gender differences and the association of maternal education with child problems (Reference Lyons-Ruth, Cicchetti and TothLyons-Ruth, 1992; Reference Sinclair and MurraySinclair & Murray, 1998). The addition of income and paternal education, as components of socio-economic status, did not improve any models described below.
In addition to descriptive statistics and chisquared analyses, two sets of logistic regression analyses addressed the main questions. Predictors were the indices of the timing of children's initial exposure to maternal depression (‘never exposed’ was always the contrast group), child gender and maternal education. Interactions of the timing indices and the other two predictors were tested in separate models because of problems of multi-colinearity. One-tailed tests were used for tests of stated hypotheses; two-tailed tests of all other predictions. To reduce the risk of type II error, only interactions significant at a 0.01 level or less are reported.
Following an analytic strategy used in other studies that have not differentiated children with ‘pure’ symptoms from those with co-occurring symptoms, the first set of logistic regression analyses estimated separate models for internalising and externalising symptoms. These were followed by multinomial logistic regressions that included the distinction between the ‘pure’ and co-occurring symptom groups. The model simultaneously estimated three equations to evaluate the influence of the predictors on the children's membership in the groups ‘high internalising symptoms only’ v. ‘low symptoms’, ‘high externalising symptoms only’ v. ‘low symptoms’, and ‘high co-occurring symptoms’ v. ‘low symptoms’.
Because we used a clinically validated cut-off point of 25% to define high symptom groups, we also checked to see whether these results were robust. We were particularly sensitive to the common problem with this categorical approach that children whose symptoms were just above or below the threshold would not be distinguished from those at the more extreme ends of the distributions. When analyses were conducted using the upper 20% and the upper 30% of the distributions, no differences in results were obtained. Similarly, when analyses were conducted omitting the children in the third quartile, no differences in results were obtained. We therefore report only the results of the analyses using the 25% cut-off points.
In the spring of their kindergarten year, the average age of the children was 73 (s.d.=3.5) months; there were 207 boys and 214 girls. Twenty-seven per cent of their mothers had experienced at least one episode of major depression since the birth. Of these, 38% (n=44) had an episode within the first year of the child's life and 61% (n=69) experienced the initial episode in the child's toddler/pre-school years. Mothers of girls were significantly more likely than mothers of boys to have had major depression during the first year (14% v. 7%), χ2 (1,n=421)=5.17, P=0.023. There was no child gender difference for mothers' first episodes during the toddler/pre-school period.
Effects of child gender and maternal education on child symptoms
Both child gender and maternal education were associated with children's symptoms. When the ‘pure’ and co-occurring symptom groups werenot distinguished, boys were significantly more likely than girls to be in the high externalising group, χ2 (1,n=421)=5.89, P=0.015. However, when the group with ‘pure’ externalising symptoms was analysed separately from the group with co-occurring symptoms, this gender effect was seen in the ‘pure’ externalising group, χ2 (1,n=421)=12.79, P<0.001, but not in the ‘both’ group, χ2 (1, n=421)=0.52, P=0.470. There were no significant gender differences for internalising symptoms.
Maternal education was also related to child symptoms. Children whose mothers had a high-school education or less were significantly more likely to be rated by teachers as high in co-occurring symptoms, χ2 (1,n=421)=16.68, P=0.001. There were no significant associations between maternal education and teacher reports of ‘pure’ internalising or externalising symptoms (Table 1).
|Low maternal education1||High maternal education||Low maternal education||High maternal education|
|n (%)||n (%)||n (%)||n (%)|
|Low symptoms group||24 (51)||88 (55)||36 (58)||103 (68)|
|High ‘pure’ internalising group||8 (17)||22 (14)||6 (10)||25 (16)|
|High ‘pure’ externalising group||8 (17)||40 (25)||6 (10)||15 (10)|
|High ‘both’ group||7 (15)||10 (6)||14 (23)||9 (6)|
Timing of initial exposure to maternal depression and child symptoms
Exposure in the infancy period (0-12 months)
Results from the first set of logistic analyses showed that when children with ‘pure’ forms of symptoms were combined with those with co-occurring symptoms, exposure to maternal depression during infancy was associated with greater odds of being in the high internalising (odds ratio (OR) 3.01, 95% CI 1.55-5.83, P=0.001) and the high externalising symptom groups (OR 2.15, 95% CI 1.07-4.32, P=0.016) (Table 2). However, when children with high co-occurring symptoms were distinguished from those with ‘pure’ symptoms, the multinomial logistic regressions showed differences in risk. While children exposed to maternal depression during infancy were more than twice as likely to be rated by kindergarten teachers as high in ‘pure’ symptoms (OR 2.23, 95% CI 0.94-5.37,P=0.035), these odds were more than doubled for high levels of co-occurring symptoms (OR 5.00, 95% CI 2.05-12.30, P<0.001). Of particular importance is the finding that exposure to maternal depression in infancy was not associated with risk for ‘pure’ externalising symptoms. None of the interaction terms including exposure in the infancy period was significant (Table 3).
|Timing of initial exposure to maternal depression|
|Never||Infancy period||Toddler/pre-school period|
|n (%)||n (%)||n (%)||n (%)||n (%)||n (%)|
|Low symptoms group||81 (53)||110 (71)||5 (36)||14 (46)||26 (65)||15 (52)|
|High internalising group (‘pure’+‘both’)||32 (21)||34 (22)||7 (50)||13 (44)||8 (20)||7 (24)|
|High externalising group (‘pure’+‘both’)||51 (33)||23 (15)||5 (35)||11 (37)||9 (23)||10 (34)|
|Timing of initial exposure to maternal depression|
|Never||Infancy period||Toddler/pre-school period|
|n (%)||n (%)||n (%)||n (%)||n (%)||n (%)|
|Low symptoms group||81 (53)||110 (71)||5 (36)||14 (46)||26 (65)||15 (52)|
|High ‘pure’ internalising group||21 (14)||22 (14)||4 (29)||5 (17)||5 (12)||4 (14)|
|High ‘pure’ externalising group||40 (26)||11 (7)||2 (14)||3 (10)||6 (15)||7 (24)|
|High ‘both’ group||11 (7)||12 (8)||3 (21)||8 (27)||3 (8)||3 (10)|
Initial exposure in the toddler/pre-school period (age 2-4 years)
When children with ‘pure’ forms of symptoms were combined with those with co-occurring symptoms, initial exposure to maternal depression in the toddler/pre-school period was associated with greater odds of being rated by teachers as being in the high externalising group, but only for girls (OR 4.30, 95% CI 1.32-14.06, P=0.008) (see Table 2). However, when girls with high co-occurring symptoms were distinguished from those with ‘pure’ symptoms, initial exposure in the toddler/pre-school period posed a risk for high levels of ‘pure’ externalising symptoms (OR 9.03, 95% CI 2.16-37.71, P=0.002) but not for co-occurring symptoms (OR 1.84, 95% CI 0.27-12.43, P=0.267). Exposure to maternal depression in the toddler/pre-school period was not associated with risk for high internalising symptoms, in either ‘pure’ or co-occurring forms. The interaction term including exposure in the toddler/pre-school period and maternal education was not significant (see Table 3).
Exposure during the infancy period
Although the majority of studies of maternal depression during infancy have been limited to investigations of the concurrent or short-term effects on infants, there is evidence that exposure to maternal depression in this period is associated with more enduring problems (Reference Alpern and Lyons-RuthAlpern & Lyons-Ruth, 1993; Reference Sinclair and MurraySinclair & Murray, 1998; Reference Murray, Sinclair and CooperMurray et al, 1999). Our findings parallel these studies — that is, the children who were exposed to maternal depression during infancy were more likely to be rated by their kindergarten teachers as evidencing high internalising and high externalising symptoms. However, when the co-occurring symptom group was distinguished from the ‘pure’ symptom groups, it was clear that exposure to maternal depression in infancy increased the risk of externalising problems primarily as they occurred in the context of high internalising symptoms.
These findings support the idea that infancy may be a ‘sensitive period’ for the occurrence of internalising symptoms later in childhood because it is an important time for the development of a secure mother—infant attachment, which in turn provides a framework for the infant's regulation of emotion (Reference Campbell, Cohn, Murray and CooperCampbell & Cohn, 1997). Maternal depression may impede achievement of these stage-salient tasks and have a longer-range impact on the child's affective, behavioural or cognitive development.
Initial exposure during the toddler/pre-school period
Our findings indicated that initial exposure to maternal depression in the toddler/pre-school period increased the risk only of high ‘pure’ externalising symptoms in kindergarten, and then only among girls. This stands in contrast to other studies that do not distinguish initial exposure during this developmental period from episodes that began earlier, and which generally find increased concurrent internalising and externalising symptoms. However, caution must be exercised in interpreting this finding because the cell sizes are small. Moreover, reports of gender differences in prior studies on this subject have been inconsistent (reviewed by Reference Goodman and GotlibGoodman & Gotlib, 1999).
Nevertheless, there is evidence in prior studies to suggest several plausible explanations for the risk in girls. Each is predicated on the assumption that girls identify more strongly with their mothers than do boys. To the extent that maternal depression may arise in the context of family stressors (Reference Cicchetti, Rogosch and TothCicchetti et al, 1998) that are also known to be associated with externalising disorders (Reference Fendrich, Warner and WeissmanFendrich et al, 1990) and accompanying increased irritability and harsher parenting (reviewed by Reference Cummings and DaviesCummings & Davies, 1994; Reference Goodman and GotlibGoodman & Gotlib, 1999), girls may be especially likely to react with increased negativity. Further, it may be the change from maternal health and a secure relationship in infancy to maternal depression later that has a uniquely damaging effect on girls, who may be especially reactive to this shift.
The timing of maternal depression
Taken together, the findings suggest that the impact of maternal depression may be more severe with earlier exposure (Reference Goodman and GotlibGoodman & Gotlib, 1999). It is also possible that these effects are due to the persistence or recurrence of maternal depression into the toddler/pre-school period, so that it is the length of the exposure, rather than its timing, that is important (Reference Anderson and HammenAnderson & Hammen, 1993; Reference Campbell, Cohn, Murray and CooperCampbell & Cohn, 1997; NICHD Early Child Care Research Network, 1999). However, the results of the study did not change when we took account of children who were exposed to maternal depression in both periods. Although it is important to investigate this issue more fully, these findings suggest that exposure in the infancy period does have important and lasting effects on children's mental health symptoms, consistent with studies such as those by Murray and colleagues (Reference Murray, Sinclair and CooperMurray et al, 1999).
It is possible, of course, that children's difficult temperaments or emotional and behavioural problems precede the onset of maternal depression in the toddler/pre-school period. Although some studies have demonstrated that these are reciprocal and not one-way relationships (Reference Hammen, Burge and AdrianHammen et al, 1991; Reference Gotlib and WheatonGotlib & Wheaton, 1997), others have shown that maternal depression affects children's symptoms independent of any effects of children on mothers (Reference Murray, Fiori-Cowley and HooperMurray et al, 1996).
The distinction between pure and co-occurring symptoms
The findings from this study underscore the importance of distinguishing the co-occurring symptom group from those with ‘pure’ forms. To the extent that children exposed to maternal depression during infancy experience disruptions in attachment and security, they may be especially vulnerable to internalising problems as they negotiate the challenges of developing new relationships with peers and teachers during the first year of formal schooling. If disruptions in attachment also have interfered with children's learning to regulate emotions, they may also be vulnerable to externalising symptoms during this potentially stressful time. For girls especially, initial exposure to maternal depression in the toddler/pre-school period, when the stage-salient tasks are to develop autonomy and self-control, may pose a special risk for ‘pure’ externalising symptoms as new social relationships and the structured setting of the kindergarten classroom are encountered.
Unfortunately, in attempting to clearly define and measure child mental health symptoms, researchers have taken pains to distinguish internalising from externalising symptoms but have typically ignored their frequent co-occurrence (Reference ZoccolilloZoccolillo, 1992). While ‘pure’ forms of symptoms may often (but not always) have different aetiologies, their co-occurring forms may arise from problems of attachment and emotional dysregulation (Reference Cole, Zahn-Waxler, Cicchetti and TothCole & Zahn-Waxler, 1992) originating in infancy.
The importance of distinguishing the co-occurring symptom group from the ‘pure’ symptom groups is also emphasised when considering the direct influence of child gender and maternal education on children's symptom patterns. In this study boys were more likely to be in the high ‘pure’ externalising group, possibly because externalising behaviours in their ‘pure’ form are more normative for boys than for girls. In addition, children whose mothers had less education were more likely to have co-occurring symptoms. These results suggest that the associations between family socio-economic status and externalising symptoms found in other studies (Reference Sinclair and MurraySinclair & Murray, 1998) may actually reflect an association between socio-economic status and co-occurring symptoms, rather than an association with externalising symptoms per se.
Clinical Implications and Limitations
• Early postnatal detection and treatment of maternal depression is important for identifying children at risk for ‘pure’ internalising problems, but especially for children at risk for co-occurring problems.
• Identification and treatment of maternal depression that first occurs during the toddler/pre-school years may be especially important for mothers of girls.
• If children present with externalising problems, the possibility of co-occurring internalising problems and the role of early maternal depression in their development should be considered.
• Although the importance of the timing of initial exposure to maternal depression has been suggested, the influences of other aspects of parental psychopathology (severity and duration, potentially confounding comorbid psychopathologies, paternal psychopathologies), contextual stressors and individual child characteristics have not been considered.
• The focus on teacher-reported child symptoms during the kindergarten year limits the ability to generalise the findings to the more serious persistent (across several years) and pervasive (across school and home contexts) child problems. Teacher ratings were chosen to avoid bias from maternal depression, but such ratings may themselves be biased.
• Small cell sizes, especially girls in the high ‘pure’ externalising symptom group, argue for the importance of replication of these results.
This work was supported by National Institute of Mental Health Grant (RO1-MH44340) to Marilyn J. Essex and Janet S. Hyde, with Roseanne Clark, H. Hill Goldsmith, Marjorie H. Klein, Nancy A. Smider and Deborah Lowe Vandell; additional support was provided by a grant to Marilyn J. Essex from the John D. and Catherine T. MacArthur Foundation Research Network on Psychopathology and Development (David J. Kupfer, MD, chair) and the Wisconsin Center for Affective Science (P50-MH53524; Richard J. Davidson, director). Special thanks to project manager Jeffrey M. Armstrong, field supervisors Tracy Kryshak and Tracy Buechner, the interviewers and many students who coded interviews, and the participating families and teachers who have so generously committed their time to the project. We also thank our colleagues on the MacArthur Research Network for their advice and comments, and especially Helena Kraemer for her consultation on the statistical analyses.