Maternal Attachment and Adolescent Socioemotional Functioning

A second important potential source of influence on the adolescent's capacity for emotional intimacy and maturity in their friendships is the way in which parents represent their own close relationships. Such representations have been extensively investigated within the framework of attachment theory by means of the Adult Attachment Interview (AAI), where representations of childhood attachment experiences are examined (Main & Goldwyn, 1998; Main, Kaplan, & Cassidy, 1985). The dimensions that distinguish secure from insecure adult attachment representations bear directly on the issues of emotional sensitivity and maturity central to adolescent friendships: thus, the secure, or autonomous, state of mind regarding attachment relationships is manifest in emotional openness, and in the capacity to reflect on emotional responses and appropriately incorporate them into decisions and behavior. Insecure adult styles are, by contrast, expressed either in terms of emotional avoidance (as in the dismissing style), or else in increased emotional reactivity, with negative emotions, in particular, threatening to overwhelm the individual (as in the preoccupied style). In each case, emotional responses remain relatively unintegrated with respect to the individual's wider experience, reducing the possibility of arriving at mature resolutions to interpersonal problems.

A number of processes may underlie links between parental representations of attachment relationships and the development of emotional sensitivity and maturity in the offspring. With regard to emotional sensitivity, one route involves the parent's ability to make accurate attributions about their child's mental and emotional experience (Fonagy, Steele, Moran, Steele, & Higgitt, 1991; Pederson, Gleason, Moran, & Bento, 1998), an ability that has been found in a number of studies, to be compromised in the context of maternal insecurity of attachment (Cohn, Cowan, Cowan, & Pearson, 1992; Crowell, O'Connor, Wollmers, Sprafkin, & Rao, 1991; Haft & Slade, 1989; Zeanah et al., 1993). A related route concerns the extent to which spontaneous family discourse makes reference to emotions, something that has been found to be strongly and specifically predictive of children's own emotion understanding (Brown & Dunn, 1996; Denham, Zoller, & Couchoud, 1994; Dunn, Brown, & Beardsall, 1991) and talk about feelings (Jenkins, Turrell, Kogushi, Lollis, & Ross, 2003). Thus, if discourse about emotions shows an exaggerated concern with negative feelings, as would be predicted for those with preoccupied attachments, or is restricted, as can be expected in the case of a dismissing parental attachment style, the child can be expected to become more or less sensitized accordingly to emotional states in themselves and others (Steele, Steele, Croft, & Fonagy, 1999; Steele, Steele, & Johansson, 2002).

With regard to the influence of parental attachment style on child maturity, the affect regulation processes involved in the development of parents' representations of their early attachment relationships are likely to bear on the emotional regulation of relationships with their own children, and provide links to the child's capacity for mature resolution of difficulties. Furthermore, the parent's capacity to reflect on relationships in a coherent manner, characteristic of those with a secure attachment style (Fonagy & Target, 1997), is likely to influence the way difficulties in relationships are discussed and resolved in the family, and thereby affect the child's own ability to generate mature solutions to social problems. A number of studies have provided evidence to support such links in relation to young children, although their application to adolescent populations has not been established. Thus, Crowell and Feldman (1988) found mothers' adult attachment representations to be related to the way in which they interacted with their preschool children during challenging play tasks. Secure mothers were able to support their children through the challenge, whereas insecure mothers were either confusing (preoccupied group), or else overcontrolling and directive (dismissing group). In addition to these dysregulating effects of insecure parental attachment on parent–child interactions, there is evidence for corresponding adverse effects on child behavior. Thus, both DeKlyen (1996) and Cowan and colleagues (Cowan, Cowan, Cohn, & Pearson, 1996) found insecure parental attachment to predict adjustment problems in 4- to 5-year-old children. Of particular relevance to the development of social maturity, Steele and colleagues found that 11-year-olds' capacity to generate resolutions to cartoon depictions of distressing emotional and social interactions was significantly related to the mother's account of her attachment relationships on the AAI (Steele et al., 2002).

Relationship Between Maternal Depression and Attachment Style

Some of the interpersonal difficulties that characterize insecure adult attachments clearly have elements in common with those expressed by depressed individuals. Indeed, a recent meta-analysis (van IJzendoorn & Bakermans-Kranenburg, 1996) showed a strong overrepresentation of insecure attachments among clinical populations in general. There is also evidence for a specific link with affective disorder, preoccupied attachments being overrepresented in depressed populations (Bifulco, Moran, Ball, & Bernazzani, 2002; Rosenstein & Horowitz, 1996). The question of whether insecure adult attachment and depression provide independent sources of influence on parenting and child outcome has received little attention, although one recent study found depression to outweigh the influence of adult attachment relationships on parental perceptions of infant negativity (Pesonen, Raikkonen, Strandberg, Keltikangas-Jarvinen, & Jarvenpaa, 2004). The relative contributions of maternal depression and insecure attachment to the socioemotional functioning of the offspring in the longer term require investigation.

Developmental Processes Linking Maternal Depression and Attachment to Adolescent Functioning

The Current Study

We simultaneously examined these issues in the context of a prospective longitudinal study of the offspring of postnatally depressed and well mothers. We have previously reported on the children's development in infancy and through the early school years. In common with other studies, we found infants of PND mothers to be more frequently insecurely attached than those of well mothers (Murray et al., 1996), and to have difficulties in their social relationships (Murray, Sinclair, et al., 1999). We also assessed the way in which the children represented their family relationships via a doll play procedure: compared to nonexposed children, children exposed to PND showed increased awareness of the emotional experience of family members in their depictions of caretaking interactions, as well as a reduced ability to reflect on their family experiences in a coherent way (Murray, Woolgar, et al., 1999). Both these dimensions of 5-year representations are of clear relevance for the emotional sensitivity and maturity that are central aspects of adolescent social development. In the current study we examined the children's socioemotional development as they reached adolescence at the age of 13 years, when their sensitivity to emotional experiences and social maturity was assessed. As a significant body of work has suggested that schema relating to the self and others are best elicited in relatively unconstrained or spontaneous conditions (Main & Goldwyn, 1998; Murray, Woolgar, Cooper, & Hipwell, 2001; Teasdale & Barnard, 1993), we assessed adolescent socioemotional functioning by examining narrative descriptions of their own friendship experiences. Furthermore, following research indicating that cognitions relevant to depression are primarily manifest under conditions of low mood (Miranda & Persons, 1988; Teasdale, 1988), we set two conditions in which the adolescents were asked to talk about their friendships: a positive condition, in which they described a good experience, and also a negative condition in which they recounted a bad or difficult friendship experience. Indices of emotional sensitivity and maturity were derived from the resulting accounts. We examined these outcomes in relation to adolescent exposure to PND, and also in relation to the nature of maternal attachment status. We also addressed the question of whether any links between maternal PND and attachment and adolescent functioning were associated with earlier aspects of the children's development, namely, their attachment status in infancy, and the nature of their family representations at 5 years.

The following specific predictions were tested.

1. In the context of accounts of negative friendship experiences, exposure to PND would predict high levels of adolescent emotional sensitivity and reduced maturity; we predicted similar effects of maternal attachment insecurity which, on the basis of previous research, we expected to be predominantly of the preoccupied form in this sample. As previous reports have suggested that girls may be particularly sensitive to maternal distress, and develop a caregiving role in response to such maternal difficulties, whereas boys are more likely to develop oppositional behaviors, we anticipated that gender would be a moderating influence, with high rates of emotional sensitivity being primarily observed in exposed girls, and low rates of maturity being predominant in exposed boys.
2. Any such influences of maternal depression or insecure attachment on adolescent development would be part of a developmental pathway from earlier aspects of the child's functioning, that is, the nature of infant attachment as assessed at 18 months, and the childhood representations of family relationships assessed at 5 years. With regard to infant attachment, in line with previous research (e.g., Cohn, 1990; Renken, Egeland, Marvinney, Mangelsdorf, & Sroufe, 1989; Turner, 1991), we anticipated that the effects of insecurity would differ for boys and girls, and specifically, that insecure boys would show poorer functioning (i.e., less maturity) than insecure girls; with regard to 5-year representations, we predicted sensitivity to the emotional experience of family members would be directly associated with later emotional sensitivity, and the ability to give a coherent narrative account of family experience at age 5 would predict adolescent social maturity.

In addition to examining the precursors of emotional sensitivity and maturity in adolescent friendships, as described above, a secondary aim of the study was to address the question of the significance of these two developments for wider adolescent functioning. The prediction for social maturity was straightforward, that is, that low levels of maturity would be associated with general poor adjustment. However, with regard to emotional sensitivity, the question of whether high levels are adaptive (i.e., facilitating prosocial behavior), or maladaptive (i.e., predisposing to depression), is more complex; in view of the literature on the origins of gender differences in the occurrence of depression in this age group, we predicted that risk for adolescent depression in association with heightened emotional sensitivity would apply principally to girls (Gjerde, 1995; Nolen-Hoeksema & Girgus, 1994; Zahn-Waxler, 2000).

Participants

Assessments were carried out as part of a prospective longitudinal study of the impact of PND on child development. Primiparous mothers of healthy infants presenting between February 1986 and 1988 on the postnatal wards of the Cambridge (UK) maternity hospital, who intended to be the infant's principal caretaker, were identified with a view to recruitment in the study. Because we wished to focus on the effects of maternal depression, rather than those of general adversity, teenage and single mothers were not approached. Of the 702 women meeting study criteria who were invited to participate, all but nine (1.3%) were willing to be sent a questionnaire concerning their mood. This questionnaire, the Edinburgh Postnatal Depression Scale (EPDS; Cox, Holden, & Sagovsky, 1987), was administered by post at 6 weeks postpartum; the return rate was 97%. Women scoring over 12 on the EPDS were interviewed at 8 weeks with the Schedule for Affective Disorders and Schizophrenia (Endicott & Spitzer, 1978); 61 women who met Research Diagnostic Criteria (Spitzer, Endicott, & Robins, 1978) for depressive disorder were identified, 58 of whom were recruited for the study. Forty-two nondepressed mothers with low EPDS scores were randomly selected from the same postnatal population, potentially to serve as controls; these women were similarly interviewed to confirm they were not depressed, and were invited to participate: all agreed. Follow-up assessments were carried out when the child was 18 months and 5, 8, and 13 years old. The current research focuses on child development at 13 years, using information from prior assessments to explain variance in the dimensions examined.¹

Fifty-three (91.4%) PND group and 41 (97.6%) comparison group families were retained at 13 years. Attrition was as follows: two dyads (both PND group) were lost to follow-up at 18 months: one family emigrated and one could not be traced; two mothers (one PND group, one control) were unwilling to continue in the study at 5 years; at 13 years, one further mother was unwilling to continue, and one adolescent was lost to follow-up when he emigrated to live with his father (both PND group).

Procedure

Adolescents and their mothers were assessed in their home. In the course of a range of assessments, the friendship interview was administered and assessments were made of adolescent mental state (interview and questionnaires). Maternal mental state was also assessed. Permission was sought to approach the school to obtain teacher reports concerning adolescent functioning.

Maternal Measures

Adolescent/Child Measures

Data Analyses

Inspection of the ratings of the adolescent emotional sensitivity and maturity scales in the context of a negative experience showed them to be suitable for parametric analyses. Therefore, in addressing the first two study hypotheses, namely that these two adolescent outcomes would be predicted by exposure to each of PND and maternal attachment security, in interaction with child gender, we used two-way analyses of variance (ANOVAs). The third study hypothesis, that any influence of exposure to PND or maternal insecure attachment on adolescent functioning would be preceded by particular child developmental trajectories, was investigated in two stages: first, the prediction of adolescent socioemotional functioning was examined with respect to (a) infant attachment using two-way ANOVAs (with gender as the second predictor) and (b) 5-year child representations of the family using partial correlation (controlling for gender). Second, we used linear regression techniques, in which the influence of the two maternal variables, along with child gender, were considered together with the relevant earlier child development measures. Third, we examined the relationship between the friendship interview measures of emotional sensitivity and social maturity in relation to wider aspects of adolescent functioning, namely the occurrence of depressive symptoms, and overall psychological adjustment, respectively, by means of regression and correlational analyses, as appropriate.

Effects of Maternal PND

Two-way ANOVAs examined the prediction of adolescent emotional sensitivity and maturity (as assessed in the friendship interview) by maternal PND and adolescent gender. The results are illustrated in Figure 1a. For emotional sensitivity, there was a significant main effect of gender, F (1, 87) = 10.73, p = .002, with girls being more sensitive overall than boys. There was no main effect of PND, F (1, 87) = 1.53, ns, but there was a significant PND × Gender interaction, F (1, 87) = 5.64, p = .02. Post hoc t tests showed that PND group girls had significantly higher levels of emotional sensitivity than control group girls, t (46) = 2.62, p = .012; although boys exposed to maternal PND showed slightly lower emotional sensitivity than control group boys, this effect was not significant, t (41) = −0.79, ns.³

The effects of postnatal depression and maternal and infant attachment on adolescent socioemotional responses in relation to gender: (a) maternal depression, (b) maternal attachment, and (c) infant attachment; ([utrif ]) girls and ([bull ]) boys.

For social maturity, there was again a significant main effect of gender, F (1, 87) = 11.58, p = .001, due to girls showing higher levels of maturity than boys. There was no main effect of PND, F (1, 87) = 0.82, ns, but a significant PND × Gender interaction was again observed, F (1, 87) = 7.61, p = .007. Post hoc tests showed that exposure to PND was associated with significantly higher maturity in girls, t (46) = 2.84, p = .007; the marginally lower levels of maturity in PND group boys did not differ significantly from those in control group boys, t (41) = 1.20, ns (see Figure 1a).

Mothers who developed PND were more likely than mothers who were not postnatally depressed to experience one or more episodes of depression following the postnatal episode, with the postnatal period being defined as the first 6 months postpartum (see Table 1). We therefore examined the extent to which such subsequent exposure influenced adolescent socioemotional development. Thus, we repeated the ANOVAs described above, including the presence or absence of maternal depression outside the postnatal period as a third predictor. This third factor did not prove to be significantly related to adolescent emotional sensitivity or maturity, either by itself, or in interaction with gender and PND (all ps >.16). In addition, the previously reported significant effects of PND were retained. We also used linear regression to examine total months of exposure to maternal depression during the course of the study in relation to adolescent outcomes; in no instance did this factor prove to be significant (either by itself or in interaction with gender) once the effects of PND had been taken into account (all ps > .33).

Effects of Maternal Attachment

Maternal attachment classifications are reported in Table 1. Mothers who were postnatally depressed were significantly more likely to show insecure attachments on the AAI than control group mothers, particularly due to a high frequency of preoccupied classifications. Two-way ANOVAs examined adolescent friendship outcomes with respect to maternal attachment and child gender. For emotional sensitivity, there was a significant main effect of gender, as already described, F (1, 83) = 8.42, p = .005, but there was no main effect of maternal attachment, F (1, 83) = 0.21, ns, nor was there a significant Attachment × Gender interaction, F (1, 83) = 3.02, ns. For maturity, in addition to the main effect of gender previously described, F (1, 83) = 12.96, p = .001, there was a significant main effect of maternal attachment, F (1, 83) = 4.20, p = .044, with adolescents of insecure mothers showing lower levels of maturity; there was also a significant Attachment × Gender interaction, F (1, 83) = 5.00, p = .028. Post hoc t tests examining the interaction showed significantly lower levels of maturity in boys whose mothers had insecure as opposed to secure adult attachments, t (39) = −3.05, p = .004, but no differences in relation to maternal attachment in girls, t (44) = 0.13, ns, as illustrated in Figure 1b.

Intervening Variables

Both infant 18-month attachment and child representations at 5 years were considered as possible precursors of socioemotional representations at 13 years. Infant attachment classifications are reported in Table 2, by maternal PND and maternal attachment. As previously described for this group (Murray et al., 1996), significantly more offspring of PND mothers versus comparison mothers had insecure attachments at 18 months, particularly insecure–avoidant attachments. However, there was no association between maternal and child security of attachment (see Table 2).

Infant attachment at 18 months reported in relation to maternal postnatal depression and maternal adult attachment security

Dolls' house representations at 5 years are presented in Table 3. We have previously reported that girls exposed to maternal PND showed higher levels of awareness of emotional experience in their depictions of caretaking interactions, and gave more coherent accounts than control group girls, whereas boys showed the opposite pattern (Murray, Woolgar, et al., 1999), effects that were maintained in the current subset of the sample.⁴

Mean (SD) values for emotional sensitivity in depictions of caretaking and narrative coherence in dolls' house play at 5 years reported with respect to maternal postnatal depression, maternal attachment, and child gender

Overall Model

For both adolescent emotional sensitivity and social maturity variables, we carried out regression analyses to examine the simultaneous prediction of these outcomes by those variables previously identified as being significant, including both maternal factors (PND and adult attachment security) and preceding child characteristics (18-month attachment security and doll-play variables at 5 years), as appropriate. The relatively small sample size coupled with the use of multiple, sometimes overlapping, predictors limit the power available for such an analysis; as such, we adopted a significance level of p < .10 for this set of analyses. Consecutive chronological steps were examined. Thus, for emotional sensitivity we carried out linear regression with variables entered in four logically consecutive blocks, as follows: (a) child gender, (b) maternal PND status and the PND × Gender interaction term, (c) 18-month infant attachment security and the Attachment × Gender interaction term, and (d) 5-year emotional depictions in doll play. The results are presented in Table 4. Consecutive steps in the model improved the fit when the significance of the change in R² was considered, with the exception of Step 3, for which p > .10. The failure of the infant attachment variables to significantly improve the fit of the model may have been due to their strong association with maternal PND status, which was entered into the model in Step 2. Indeed, in the final model, only the 18-month infant attachment by gender interaction and the dolls' house variable proved to be independently significant predictors of adolescent emotional sensitivity.

Linear regression examining the cumulative prediction of 13-year emotional sensitivity by gender, maternal postnatal depression, 18-month attachment security, and 5-year emotional sensitivity in representations of maternal caretaking

For adolescent maturity, we similarly carried out linear regression analyses with the previously identified predictor variables entered in three logically consecutive blocks: (a) child gender; (b) maternal variables: PND status and PND × Gender interaction term, adult attachment security, and Attachment × Gender interaction term; (c) 5-year narrative coherence in doll play (Table 5). Once again, consecutive stages in the model proved progressively to improve the fit. In the final model, the PND × Gender interaction term, maternal attachment security, and the 5-year narrative coherence score were all significant, independent predictors of adolescent maturity.

Linear regression examining the cumulative prediction of 13-year maturity by gender, maternal postnatal depression, maternal adult attachment security, and narrative coherence in 5-year doll play

Although the total variance explained by the two final regression models was modest (R² = .26 for emotional sensitivity, R² = .33 for maturity), this must be considered in the light of the fact that, with the exception of gender, the most proximal predictor of the 13-year outcomes was measured 8 years previously, and in consideration of the conceptual overlap between predictors examined. Notably, for both emotional sensitivity and maturity the striking apparent differences between adolescent girls and boys, evident in the highly significant effects of gender when considered alone (Step 1 of each model), were wholly accounted for by the distinctive impact on each of maternal depression and/or maternal attachment, and their ensuing divergent developmental trajectories; the overall pattern of these links is illustrated in Figure 2.

The overall patterns of pathways to emotional sensitivity and maturity in adolescents in relation to gender: the pattern of trajectories to (a) high and low emotional sensitivity and (b) high and low maturity in adolescent girls and boys.

Relationship to Wider Psychological Functioning

Finally, we addressed the issue of the significance of the adolescents' accounts of their friendships for their wider psychological functioning; we examined both their experience of depression, assessed using the MFQ, and their overall adjustment, as assessed independently in teacher reports on the SDQ (total problem scores).

The Role of Maternal Attachment

We observed a higher number of insecure maternal attachments, as assessed by the AAI at 18 months postpartum, in mothers who became depressed in the postnatal period compared to nondepressed mothers, an association consistent with previous research on attachment in clinical populations (van IJzendoorn & Bakermans-Kranenburg, 1996). Notably, the elevated rates of insecure maternal attachment observed in association with PND were entirely attributable to a higher number of preoccupied attachments; insecure–dismissing attachments were no more common among PND mothers than in the control group. This specific link between depression and a preoccupied attachment style has also been noted in previous research (Bifulco et al., 2002; Rosenstein & Horowitz, 1996).

Although there was a significant association between insecure maternal attachment and PND, there were independent effects of maternal attachment and depression on adolescent functioning. Thus, maternal attachment insecurity, unlike depression, was not associated with adolescent emotional sensitivity; however, it was associated with lower maturity in adolescent offspring, independent of maternal PND status.

The failure to identify an effect of maternal attachment on adolescent emotional sensitivity was somewhat unexpected in view of the evidence of insecure parents' poor capacities to identify their children's mental and emotional states accurately (Cohn et al., 1992; Crowell et al., 1991; Haft & Slade, 1989; Zeanah et al., 1993). It is important to note, however, that our measure of emotional sensitivity was not concerned with the accuracy of adolescent perceptions of emotion, but with their tendency to focus on emotional aspects of their relationships, a dimension that is markedly different in insecure preoccupied, as opposed to dismissing, adult attachments. Thus, sensitivity to, and expression of, emotional experiences is primarily a characteristic of preoccupied attachments, but is notably absent in dismissing attachments. The absence of a simple association between insecure maternal attachment and child emotional sensitivity in the current sample may have been due to the fact that, overall, preoccupied and dismissing styles were present in approximately equal proportions. Our sample lacked power to properly address the question of whether different insecure maternal attachment patterns were associated with different adolescent outcomes. Nevertheless, a preliminary examination suggests that this is likely to be the case, emotional sensitivity scores being highest for girls whose mothers had preoccupied attachments and lowest for boys with dismissing mothers.

The effect of maternal attachment on adolescent maturity, over and above that of maternal depression, was notable. A number of processes are likely to be compromised in the context of adult insecurity that may have an adverse impact on the development of social maturity in the offspring. One possible route concerns the parent's ability to regulate their own and their child's emotional states during parent–child interactions; adult insecurity has been found to be associated with difficulties in this domain, as well as with disturbances in child behavior (Cowan et al., 1996; Crowell & Feldman, 1988; DeKlyen, 1996). A second route concerns the promotion of child capacities to understand their own and others' mental states, a process that may also be impaired in insecure parents. Thus, as noted above, there is an increased likelihood that the adult's identification of the child's emotions will be inaccurate where the parent is insecure. In addition, it is likely that family discourse concerning personal relationships will be affected: dismissing mothers may be prone to avoid personal topics, and restrict emotional discourse, whereas in those with preoccupied attachments, balanced discussion of the causes and consequences of difficult feelings may be missing (Steele et al., 1999, 2002). In the context of adult security, by contrast, family discourse concerning the mental states and motivations underlying behavior is likely to be both more considered, as well as accurate, thus fostering the child's capacity to understand different social perspectives (Fonagy & Target, 1997). The fact that we found maternal attachment security to predict 5-year-olds' capacity to give coherent narrative accounts of their family relationships in the current study (at least in boys), and that this, in turn, was predictive of the degree of maturity shown in adolescent relationships, supports this view. These findings extend the evidence concerning the kinds of family relationships that promote the understanding of others' mental states (Denham et al., 1994; Dunn, Brown, & Beardsall, 1991; Dunn, Brown, Slomkowski, Tesla, & Youngblade, 1991; Jenkins & Astington, 1996; Meins et al., 2002; Perner, Ruffman, & Leekman, 1994; Ruffman, Perner, Naito, Parkin, & Clements, 1998) to include parental attachment security as an important source of variation. Finally, our finding that 5-year-olds' representations were predictive of adolescent accounts of relationships extends previous research showing concurrent and short-term associations between early narrative style and wider child functioning (Bretherton & Munholland, 1999; Woolgar, 1999).

The Significance of Heightened Emotional Sensitivity: Sign of Resilience or Risk Factor for Depression?

In the current study we found high levels of emotional sensitivity in girls to be associated with increased depressive symptoms. Previous research has similarly suggested that heightened emotional sensitivity may raise the risk for depression (Block, Gjerde, & Block, 1991; Gjerde, 1995; Goodyer, Herbert, Tamplin, & Altham, 2000; Gore et al., 1993; Zahn-Waxler, 2000). However, other perspectives have emerged in the research literature concerning the implications of a high level of emotional sensitivity for wider psychological functioning. Thus, such sensitivity has also been linked to certain psychological strengths, including responsible, empathic and well-regulated behaviors that may serve as protective factors in relation to the development of externalizing problems (Carlo et al., 1996; Eisenberg et al., 1995, 1996; Fabes et al., 1994; Zahn-Waxler, Klimes-Dougan, & Slattery, 2000). Emotional sensitivity has also been seen as a core component of mature self-conceptualization, and of meaningful relationships with others (Harter, 1986).

An important question arising from previous research concerns, therefore, the conditions under which emotional sensitivity becomes associated with either good adjustment, or else renders the individual vulnerable to depression. At present, these conditions are not fully understood. Longitudinal research stands to elucidate this question, but has so far largely failed to identify clear prospective links between individual differences in early psychological functioning, at least in girls, and the later emergence of depression (Zahn-Waxler et al., 2000). Consistent links between emotional sensitivity and depression have, however, emerged in adolescence (Gjerde, 1995; Gore et al., 1993), and this has contributed to the view that emotional sensitivity may act as a risk factor for depression only when occurring in combination with other developments that emerge at this time, and particularly those that may threaten emotional regulation. These include the increasing tendency in adolescence to become involved in interpersonal situations that may be stressful (Gore et al., 1993), as well as the development of higher level cognitive functions, such as rumination (Nolen-Hoeksema & Girgus, 1994; Park, Goodyer, & Teasdale, 2004; Zahn-Waxler, 2000). Of particular interest in the light of the current study findings is the fact that the vulnerability to depression conferred by emotional sensitivity, in interaction with both a ruminative cognitive style and exposure to family stress, has been found in previous research to be more prevalent in girls than in boys (Gore et al., 1993; Nolen-Hoeksema, Girgus, & Seligman, 1992; Silk, Steinberg, & Morris, 2003).

Although links between emotional sensitivity and the experience of depression may only become fully evident in the adolescent years, from a developmental psychopathology perspective it remains important to understand the pathways by which problems in regulating emotional sensitivity arise, and it may be fruitful to consider whether particular dispositional characteristics or environmental conditions determine when high emotional sensitivity is adaptive or else maladaptive. In the current study, we found that heightened emotional sensitivity in adolescent girls occurred in the context of the development of an insecure attachment relationship in infancy, an association consistent with previous research (Turner, 1991). In turn, it can be hypothesized that whether or not such sensitivity poses a risk for depression may be fundamentally affected by the nature of the individual's attachment status. Thus, in the context of security, a high degree of sensitivity to the emotional components of relationships may be advantageous, enabling the adolescent to respond adaptively while maintaining a well-regulated state; by contrast, in cases where high emotional sensitivity occurs in the context of insecurity, the individual may find themselves easily overwhelmed when difficult interpersonal situations occur, resulting in a depressive state. Family environment is also likely to be a significant determinant of the adaptiveness or otherwise of high emotional sensitivity. Given early exposure to family adversity or maternal depression, the development of high emotional sensitivity may carry negative connotations of worry and concern for others (Zahn-Waxler, 2000; Zahn-Waxler et al., 1991).

An exploratory inspection of our data supports a role for both early attachment and maternal depression in determining the psychological outcomes associated with high emotional sensitivity. Thus, high emotional sensitivity occurring in the context of a secure child attachment was unrelated to adolescent depressive symptoms, whereas in those who were insecure the association was positive. In addition, we noted a significant link between emotional sensitivity and adolescent depressive symptoms in daughters of depressed mothers, whereas no such relationship was evident in those whose mothers had been well. In sum, and consistent with the views of Zahn-Waxler and colleagues (Zahn-Waxler, 2000; Zahn-Waxler et al., 1991), a preliminary analysis suggests that both insecurity of attachment, as well as exposure to maternal depression, may be early sources of individual differences in the propensity for emotional sensitivity to be associated with either resilience or vulnerability in the face of normative adolescent developmental challenges. Systematic investigation of interactions between such early vulnerability and the consequences of emotional sensitivity represent an important area for future research concerning the causes of depression.

Further Considerations

A number of aspects of our study require further comment. First, it was notable that we failed to find an association between maternal attachment status and infant attachment security. This was surprising in the light of the extensive literature documenting the correspondence between maternal AAI and infant Strange Situation attachment status (van IJzendoorn, 1995). The nature of our sample, however, needs to be borne in mind; thus, a higher proportion of mothers coded autonomous–secure on the AAI had securely attached infants in the control group, compared to the rate in the PND group (72 vs. 20%). Indeed, mothers in the control group tended to have securely attached infants although mothers in the PND group tended to have insecure infants, regardless of maternal attachment status. Speculatively, therefore, the pattern of association between maternal and infant attachment we observed may have been a function of the use of postnatal depressive symptomatology to select women for the study. In particular, women who are insecurely attached, but who do not develop depressive symptoms following birth, as in our control group, may be relatively good at separating their own attachment experiences from their early relationship with their infant, enabling secure infant attachment to develop. In addition, where depression does occur, its negative effects may outweigh the potential benefit of maternal attachment security. This interpretation is, in fact, consistent with the findings of a recent study showing that the normal links between adult attachment status and the quality of other family relationships were obscured in the presence of depression (Dickstein, Seifer, Albus, & Magee, 2004).

Second, the nature of the effects of PND requires comment. Thus, with regard to its timing, our research has emphasized the role of early exposure in subsequent child socioemotional development: adolescent emotional sensitivity was associated with maternal depression occurring in the first few months, but it was not linked to the overall duration of exposure to depression, nor to depression that occurred after the postnatal period. Although these results are consistent with those of other reports of children of postnatally depressed mothers that have found persistent impairments in child functioning to be specifically associated with postpartum exposure (Hay et al., 2001; Sharp et al., 1995), it should be noted that, in general, research on the effects of maternal depression on child development has found that chronic, rather than brief, exposure is pernicious. In addition, the relatively small sample size of the current study places limitations on the extent to which the effects of different timings of exposure can be investigated. Caution is therefore required when considering this aspect of our results. Furthermore, although we have focused on the role of maternal behavior in the development of adolescent socioemotional functioning, there are other ways in which the children of postnatally depressed mothers may acquire vulnerability (see reviews by Goodman & Gotlib, 1999; Murray & Cooper, 2003). In particular, any difficulties in emotional responsiveness and its regulation may, in part, be linked to the heritability of temperamental traits such as neuroticism (Goldsmith, Buss, & Lemery, 1997; Plomin et al., 1993; Tellegen et al., 1988), as well as to neuroendocrine effects on the fetus that influence the capacity to regulate state and emotions (Abrams, Field, Scafidi, & Prodromidis, 1995; Field, 2002; Zuckerman, Bauchner, Parker, & Cabral, 1990).

Regardless of the precise source of the disturbances in adolescent functioning that we observed in relation to early maternal factors, the observation that these factors predict aspects of adolescent socioemotional development 13 years later is a striking and significant result. Our data suggest that early exposure sets in train a pattern of responding that has significant continuity throughout development. Furthermore, the outcomes identified have implications for our understanding of risk for depressive disorder, particularly as it occurs in girls. Future research should further consider these issues, examining how differences in socioemotional functioning might relate to earlier developments that may confer risk, as well as underlying biological and higher level cognitive processes, and the new social demands of adolescence that have been implicated in the development of disorder.