Are mothers’ protective attachment strategies related to their children’s strategies?

Abstract

Aims:

We explored the relation between mothers’ protective attachment strategies and those of their school-age children.

Methods:

In total, 49 child–mother dyads participated in a short longitudinal study when the children were 5.5 and 6.0 years old. Their strategies were first assessed with the Preschool Assessment of Attachment (PAA) and then with the School-age Assessment of Attachment (SAA). Mothers were assessed with the Adult Attachment Interview (AAI). The Dynamic-Maturational Model of Attachment and Adaptation (DMM) was used to classify the assessments.

Results:

The validity and precision of the DMM-AAI were supported: Mothers’ AAI classifications were related to their referral group (normative or clinical) and measures of stress and distress. The DMM categories were more associated with risk than the Ainsworth categories. Types A, C and A/C were differentiated by trauma, triangulation and depression. Mothers’ and children’s protective attachment strategies were related, with B mothers having B children and A or C mothers having children using the same or opposite strategy. Children whose classification changed from the PAA to the SAA had mothers with complex traumas.

Conclusion:

When psychosocial treatment is needed, knowing whether mother and child use the same or different strategies and whether mothers have complex trauma can affect treatment success.

Keywords

Attachment AAI DMM-AAI school-age children SAA transgenerational attachment

It is widely recognized that maternal sensitivity and attachment strategies influence the well-being of infants and that this relation requires early intervention directed towards risk mothers. Much less attention has been paid to school-age children, and yet the school years are when most children with psychological or behavioural problems are identified. Furthermore, such children are often given individual diagnoses (e.g. attention deficit hyperactivity disorder (ADHD) and school phobia) and treated individually (e.g. with medication and cognitive behavioural treatment (CBT)). We explored the relation between mothers’ attachment strategies and their children’s attachment and well-being. We wanted to know whether children’s problems, like infants’ problems, still reflected important contributions from their mothers. Specifically, we wanted to know whether or not school-age children were adversely affected by their mothers’ past or current endangerment. If they were and, particularly, if the mothers’ experience of danger was complex and expressed inexplicitly, then intervention for the children might need to address mothers’ endangerment.

Below, we first define attachment in the school years in terms of how attachment reflects mothers’ and children’s understanding of what is dangerous and how to protect oneself from danger. We then review what is known about the relation between mothers’ and children’s attachment. We conclude by offering three hypotheses, two of which our data support strongly and one of which produced complex results. In the discussion, we consider how our findings can be used by social workers, educators, psychologists and others who work with troubled school-age children.

Attachment in the school years

Attachment is an enduring affectional bond that connects mothers and children for the evolutionary function of promoting survival (Ainsworth, 1979). With Bowlby, Ainsworth emphasized not only the importance of attachment for protecting babies but also its impact on older children and even spousal relationships (Bowlby, 1979). We take danger as our starting point (Crittenden, 1999) and define attachment as ‘strategies’ (Main & Weston, 1981) for protecting and comforting the self, one’s partner and one’s children from the dangers one has experienced (Crittenden, 2008/2015). Although most attachment research is limited to Ainsworth’s infant patterns (A1-2, B1-4 and C1-2), with or without Main and Solomon’s addition of D (disorganization), we take the perspective that attachment develops as children develop and is organized protectively, especially when protection is most needed (Crittenden & Ainsworth, 1989). This idea is central to the Dynamic-Maturational Model of Attachment and Adaptation (DMM; Crittenden, 1995/2015, 2008/2015). The crucial aspects of the DMM, as it affects school-age children’s adaptation, are that it describes a wider range of strategies, for both children and adults, than the ABC+D model (see Figures 1 and 2) and that it ties the strategies to information processing (Bowlby, 1980, ch. 4). Information processing, we think, is the key to both understanding maladaptation and guiding treatment.

Figure 1.

The DMM classifications in the school years (used with permission).

Figure 2.

The DMM classifications in adulthood (used with permission).

The DMM and a gradient of severity of risk

The original Ainsworth patterns of attachment (i.e. A1-2, B1-4 and C1-2) were first observed in Ainsworth’s middle-class Baltimore sample of infants (Ainsworth, Blehar, Waters, & Wall, 1978) and have been replicated in many other safe samples and at ages beyond infancy. At all ages, people using the ‘Ainsworth’ strategies have experienced few dangers and use information with clarity and transparency. Risk samples of both infants and older people indicate that endangered people transform information and behave in ways that can be misleading to others (but are often protective to the individual). For example, some abused children use false positive affect and caregiving or compliant behaviour to appease distressed or angry parents (A3-4; Crittenden & DiLalla, 1988). Conversely, children with unpredictable parents often heighten the display of negative affect and behaviour to increase the predictability and supportiveness of their parents (C3-4; Crittenden, 1992). By the school years, children can deny perception of distress and pain and behave as if they did not need protection or comfort (A6), or conversely, they can deceive others regarding their needs such that children who feel vulnerable appear aggressively invulnerable (C5) and others, who feel intense anger, can appear seductively helpless (C6). These are examples of the ‘DMM’ strategies (A3-8, C3-8 and A/C combinations); each behavioural strategy is protective in certain contexts but maladaptive in others. Children who have experienced danger more often use DMM strategies than Ainsworth strategies (Crittenden, Claussen, & Kozlowska, 2007; Crittenden, Robson, & Tooby, 2015; Spieker & Crittenden, 2010). Similarly, adults’ Adult Attachment Interviews (AAI; George, Kaplan, & Main, 1985/1996) coded with the DMM (i.e. the DMM-AAI) usually have higher numbered Type A and C strategies when the adults have been endangered (Farnfield, Hautamäki, Nørbech, & Sahhar, 2010).

Going to school marks a major change of context with school, and not family, claiming the majority of children’s waking time. The question is whether a child’s protective strategy – that he or she organized around conditions at home – is adaptive at school. The inappropriateness of home strategies for school conditions may be one reason that some children are first identified as maladaptive when they go to school. However, school may offer some children the opportunity to reorganize their strategies in the direction of greater balance. In particular, school-age children are in the midst of a transition from person-specific strategies to a consolidated generalized strategy. It is not clear yet when this integrative step is usually complete, and it is certain that some individuals do not accomplish this integration at all.

The DMM and type of risk

Although Ainsworth identified Types A and C as having different styles of maternal caregiving and showing different behaviours in the Strange Situation, she did not apply Bowlby’s ideas about information processing to Types A and C. The DMM describes Types A and C as using opposite forms of information processing. Individuals using Type A protective attachment strategies tend to rely on logical ‘cognitive’ information to regulate their behaviour, whereas individuals using the Type C strategies rely more on emotional ‘affective’ information. These differences in behaviour and underlying information processing may be relevant to treatment such that treatment strategies that are helpful to individuals using a Type A strategy might augment the problems of someone using a Type C strategy, and vice versa.

The DMM and exposure to danger

Self-protective strategies indicate individuals’ probable past exposure to danger, but by themselves, the DMM strategies are not maladaptive. This is because strategies are elicited by danger; if the context has little danger, the strategies will not be elicited. Moreover, the strategies function flexibly in reciprocal exchanges with other people. Thus, when the context changes, yielding unexpected responses to one’s strategy, most people modify either their strategy (i.e. reorganize the strategy) or do not use the strategy under the new conditions. The point is that being classified as using a DMM strategy may not, by itself, signal maladaptation.

Instead, when past endangerment has been beyond children’s developmental capacity to adapt, children may take ‘psychological short-cuts’ that by-pass crucial aspects of the context of the threat. The function of these shortcuts is to reduce the feeling of anxiety or, conversely, heighten one’s state of vigilance. Psychological shortcuts include dismissing or denying information, thus leaving oneself vulnerable to similar danger in the future or, conversely, becoming hyper-vigilant to signals of possible danger, thus risking over-identification of threat and overuse of one’s protective strategy. For example, raised voices can signal impending violence. Some children ignore the sounds of their parents’ shouting, thus failing to predict or protect themselves from spousal violence; in the DMM, they might display psychological trauma, in a ‘dismissed’ or ‘denied’ form, for parental violence. Other children are hyper-vigilant to any suggestion of parental conflict and respond with distress even when parents’ raised voices are in agreement; these children display psychological trauma in a ‘preoccupying’ or even ‘anticipated’ form. Conditions that are beyond children’s ability to understand include parents’ childhood traumas, parental violence and parents’ appropriation of the children into the parents’ conflict, that is, ‘triangulation’ (Dallos, Lakus, Cahart, & McKenzie, 2016). Psychological trauma is most frequent when children were endangered in ways that exceeded their ability to protect themselves, and in addition, their parents neither protected nor comforted them (Crittenden & Heller, 2017).

When children’s strategies persistently fail to protect them, they may give up, becoming ‘depressed’, or activate further to all kinds of signals and situations, becoming ‘disoriented’ regarding the source of the danger. These pervasive changes in arousal are denoted in the DMM as ‘modifying’ the strategy with a depression or disorientation marker. A third modifier reflects ‘reorganization’; mothers’ state of reorganizing from an A or C strategy in the direction of B has been associated with infant security (Iyengar, Kim, Martinez, Fonagy, & Strathearn, 2014). If replicated, this finding is crucially important for clinicians who seek to ameliorate risk relationships. We propose that psychological trauma, including loss, and pervasive states of low arousal (depression) or high arousal (disorientation) will predict maladaptation better than protective attachment strategies alone.

Needless to say, capturing so much detail in strategy, psychological trauma and unresolved loss, and modifiers makes the tasks of coding and achieving reliability more difficult than using a dichotomous (secure/anxious) or four-category (ABC+D) system. We wanted to know whether the DMM model of attachment and adaptation had sufficient clinical utility to justify the coding effort.

The relation of mothers’ attachment strategy to infants’ and children’s strategies

Many studies have found that mothers’ security or insecurity matched their infants’ security (Type B) or insecurity (non-B). However, in a meta-analysis of 206 studies of transgenerational concordance, the effect was weaker than expected (Verhage et al., 2015). Specifically, the early report of 22% matching on secure/insecure was reduced to only 9% when all studies, including unpublished studies, were considered. Put another way, 91% of the variance in children’s attachment was not explained by the statistically significant effect of infants’ matching their mothers’ attachment. Risk samples, newer studies and unpublished studies had lower effect sizes than middle-class samples in older published studies. Strikingly, the studies did not differentiate Types A, C and D. This may be an important omission because if mothers and infants differ in the form of their anxious attachment, then treatment might need to be different than if they used matching strategies. Because all but two of the studies in Verhage et al.’s meta-analysis used the ABC+D classificatory system, we wondered whether the wider array of attachment strategies in the DMM would permit both greater specificity of degree of risk and greater precision of the type of risk. Moreover, in other studies, when D is used as a category, normally developing children are over-identified as being at risk, that is, insecurely attached.

Almost all of the existing data refer to first-born infants – for whom danger is invariably tied to caregivers’ behaviour. Because by 5 or 6 years of age children have their own experiences, the relation of their strategies to their mothers’ strategies is less obvious. Indeed, the implicit assumption behind diagnosing and treating children individually (e.g. with play therapy and CBT) is that the disorder is independent of parental contribution. We looked at both mothers’ protective attachment strategies and their stress and distress to ask whether or not there was a relation between mothers’ and children’s protective strategies. We also asked whether birth order had an effect, with later-born children having a more complex developmental context than first-born children. Three studies, using the DMM, found that first-born children of mothers using an A or C strategy tended to organize the opposite A or C strategy (Crittenden, Partridge, & Claussen, 1991; Hautamäki, 2010; Shah, Fonagy, & Strathearn, 2010). A family systems perspective suggests that, except for Type B, later-born children would be unlikely to duplicate the first-born child’s strategy; instead, they would discover the strategy that functioned most protectively in the context of their older sibling(s).

We used the DMM because it (1) offered a wider and more nuanced range of protective strategies than the ABC+D model; (2) used information processing to differentiate Types A, B and C; (3) had detailed notation of psychological traumas; and (4) permitted the designation of pervasive changes in arousal. In addition, two DMM studies have found that troubled children had mothers with psychiatric diagnoses (Crittenden, Kozlowska, & Landini, 2010) and extreme AAI classifications (Landini, Crittenden, & Landi, 2016). We hoped that the more elaborated DMM organizations would better fit the range of older children’s and mothers’ behaviour, particularly in risk samples, than a secure/insecure dichotomy.

Hypotheses

Because the DMM-AAI was central to our study, we first considered evidence of its validity in our sample. We then considered the central issue of the relation between mothers’ and children’s attachment strategies in a low-income sample from a disadvantaged community in England:

DMM-AAI distributions: We expected the DMM-AAIs of mothers of children drawn from the normative population to be classified in ‘Ainsworth’ classifications, whereas mothers of clinically referred children were expected to receive ‘DMM’ classifications. Similarly, mothers of clinically referred children were expected to have more psychological trauma and loss and more depression and disorientation than mothers of children in the normative group.

DMM-AAI construct validity: We expected mothers’ DMM-AAI classifications to be related to indices of stress and distress, with mothers assigned to a Type A strategy having experienced the most exposure to danger and mothers assigned to a Type B strategy the least.

Matching and meshing: We expected mothers’ DMM-AAI classifications to be related to their children’s Preschool Assessment of Attachment (PAA) and School-age Assessment of Attachment (SAA) classifications, but in a complex manner that reflected both referral status and birth order. Matching referred to mothers and children using the same strategy, whereas meshing referred to A versus C reversal between mother and child.

Design

This was a short-term longitudinal study comparing normative and clinically referred children and their mothers at two time points. Multiple constructs and multiple methods of generating data were used. The constructs were stressors and distress, attachment, trauma/loss, depression, triangulation and reorganization; the data were generated through self-report, parent report, clinical history and standardized coded observational procedures and interviews.

Method

Participants

The participants were 49 English mothers with a child between 5.5 and 5.9 years old at Time 1 and 6.0–6.5 years old at Time 2. In total, 24 of the families were from the normative community and 25 were clinical referrals from social services or the school; one mother from the normative referral group declined the AAI. The mothers signed an informed consent as approved by the Research Ethics Committee of their local authority after which the children participated in a PAA at Time 1 and an SAA at Time 2; the mothers were given an AAI at Time 2. Of the children, 22 were first born, and 17 were second to fifth born.

All of the families lived in an economically depressed small city in England; thus, the normative group reflected mild demographic risk. The recruitment procedure is described in Crittenden, Landini, and Kozlowska (2015). The children were evenly split between boys and girls, with the mothers having an average age of 34 years; there were no referral group differences. Most of the mothers were non-working or held blue-collar jobs, but a few in the normative group had white-collar jobs. More than half of the mothers were married, with more married mothers (16 vs 12) and almost all single mothers being in the clinical group and most cohabiting mothers (8 vs 1) in the normative group (p = .048, Fisher’s exact test).

Measures of stress and distress

Both stress and distress were measured by self-report which is known to be vulnerable to social desirability and may also be effected by individuals’ attachment strategies. The Life Events Inventory (Cochrane & Robertson, 1973) provided information on stressors occurring within the previous year. The Parenting Stress Index (Abidin, 1990) and Daily Hassles Questionnaire (Department of Health, 2000) assessed mothers’ perceived distress. The Beck Depression Inventory (BDI) assessed mothers’ depression (Beck, Steer, & Brown, 1996; Beck, Ward, Mendelson, Mock, & Erbauch, 1961). The purpose of the BDI was to test the relevance of maternal low arousal for Type A. These measures are described in detail in Crittenden et al. (2010).

Assessments of attachment

Attachment was assessed at 5.5 years by the PAA (Crittenden, 1992). The PAA uses a Strange Situation with age-appropriate classificatory procedures. Like all separation-based procedures, it loses power as children approach school age. The PAA yields a dyad-specific classification of the child’s enacted protective strategy. Many studies indicate the validity of the PAA (Farnfield et al., 2010).

At 6 years of age, we used the SAA (Crittenden, 1997/2005). SAA classifications consist of three parts: a self-protective attachment strategy, psychological traumas and modifiers. To date, eight published studies demonstrate that the SAA has discriminant validity for normative and clinical groups (Crittenden, 2015; Farnfield et al., 2010; Kozlowska & Elliott, 2014).

Mothers’ attachment was assessed with the AAI (George et al., 1985/1996) classified with the DMM method (Crittenden & Landini, 2011). The DMM-AAI classifications contained the same three parts as for the SAA, but with a wider range of strategies (see Figure 1). Traumas and losses were coded in 12 categories that were reduced for analysis to four theory-defined groups: (1) none, (2) Type A related (dismissed, denied, blocked, depressed and disorganized), (3) Type C related (preoccupied, vicarious, imagined and displaced) and (4) complex. The modifiers (depression, disorientation, triangulation and reorganization) were coded as absent, partially present or present.

Coding and classification

The details of coding the PAA and SAA are given in Crittenden et al. (2015). Here, we only repeat that the PAA and SAA coders were reliable at Level II (coder reliability), blind to all information about the participants and achieved adequate agreement on these data: κ = .56, p < .001 for the PAA and κ = .59, p < .001 for the SAA. The AAIs were transcribed verbatim, including parenthetical notes for non-verbal behaviour. The transcriptions were classified by two coders who had reached Level I reliability (forensic reliability) and were blind to all information about the mothers. The six-way (A3-8, A1-2, B, C1-2, C3-8 and A/C) inter-coder agreement on 18 of the 49 AAIs was 84% (κ = .79).

Results

We first describe the distributions of the attachment variables, then show the evidence in this sample for the validity of the DMM-AAI variables and finally test the transgenerational hypothesis. We clustered the attachment classifications into several groupings: (1) secure/insecure (i.e. B vs all other categories) for direct comparison with ABC+D studies; (2) Ainsworth/DMM (i.e. A1-2, B and C1-2 vs A3-8, C3-8 and A/C) for potentially greater differentiation of risk in a dichotomous clustering; (3) a four-way clustering of Ainsworth, A+(3-8), C+(C3-8) and A/C to capture the differences between the A and C strategies; and (4) a six-way clustering of A+, A1-2, B, C1-2, C+ and A/C to capture as much differentiation in risk as possible in our relatively small sample. Due to the small cell sizes for contingency tables, the statistical significance of associations between categorical variables was assessed with Fisher’s exact tests. Associations between categorical and continuous variables were assessed with Kruskal–Wallis tests. All analyses were done with Stata 12.1 (STATA Corporation, 2011). We treated reorganizing classifications as non-B classifications.

Hypothesis 1: distributions

In our previous report on these families (Crittenden et al., 2015), we found that 40% and 42% of the children’s PAAs and SAAs, respectively, were assigned to the DMM classifications (A+, C+ and A/C). We pointed out that given that attachment classifications are not equivalent to psychiatric diagnoses and that the families came from a disadvantaged area, this distribution seemed appropriate. As shown in Figure 3, the proportion of mothers assigned to DMM classifications on the AAI was substantially higher at 67.3% (22.4% A+, 18.3% C+ and 26.6% A/C), with only 16.2% classified as B. Notably, there were no mothers assigned to A4, A8 or C7-8 classifications in this sample; given that the mothers had not been in numerous foster placements or institutionalized (mental hospital, institutions of children or prison), these strategies were not expected.

Figure 3.

The distributions of mothers’ AAI classifications by referral group.

As shown in Figure 3, the distribution of AAI classifications differed by referral group (p = .001). The normative group had 62.7% Ainsworth classifications (25% A2, 33.5% B and 4.2% C2) and 37.3% DMM classifications. Six mothers from the normative group, compared to only one mother from the clinical group, were assigned to A2. Overall, this fit our expectation of a bit more than the expected 25–30% DMM classifications in this disadvantaged community (Hagnell, Öjesjö, Otterbeck, & Rorsman, 1994; Kessler, 1994; McLaughlin et al., 2012; Roberts, Attkisson, & Rosenblatt, 1998; Rutter & Rutter, 1993; Schepank, 1987). In the clinical group, all but one mother (96%) received a DMM classification (A+, C+ and A/C).

The association between referral group and the presence or absence of psychological trauma/loss was almost significant (p = .093). When tested dichotomously as none/preoccupied/dismissing versus more complex forms, clinically referred mothers had significantly more complex traumas (p = .031).

None of the mothers in the normative group showed any depression on the DMM-AAI as compared to six in the clinical group. Four of the mothers in the clinical group met criteria for partial depression and two for full depression. The three mothers whose DMM-AAI showed partial disorientation were in the normative group. There were nine DMM-AAIs coded as showing triangulation; three were in the normative group and six in the clinical group; the difference was not a statistically significant. Eight AAIs were given a reorganizing modifier, four from each referral group.

Hypothesis 2: construct validity of the DMM-AAI

We looked at the relations of the mothers’ DMM-AAI classifications to stressors and feelings of distress; we were particularly interested to know whether the DMM classifications added anything to the secure/insecure analyses used in most AAI studies.

Comparisons of DMM-AAI classifications with negative life events, parenting stress and depression are shown in Table 1. The differences were in the expected direction for all three clusters of attachment classifications with those for Ainsworth versus DMM being somewhat stronger than secure versus insecure. Differences by the four-way clusters showed a pattern of the Ainsworth categories having the lowest scores and Type A the highest scores on the stress and distress variables.

Table 1.

Stress and distress: negative life events, parenting stress and depression by secure versus insecure, Ainsworth classifications versus DMM classifications, and a four-way clustering of classifications.

		Secure (B)	Insecure (not B)		χ² (1)	p value
		Mean rank	Mean rank		χ² (1)	p value
Negative life events		13.38	27.27		6.49	.011
Total parenting stress		6.75	14.19		3.44	.064
Beck Depression		13.25	18.41		1.34	.248
		Ainsworth: A1-2, B, C1-2	DMM: A+, C+, A/C		χ² (1)	p value
		Mean rank	Mean rank		χ² (1)	p value
Negative life events		14.13	30.27		14.11	<.001
Total parenting stress		6.57	15.50		7.42	.006
Beck Depression		12.75	19.48		3.24	.072
	A1-2, B, C1-2	A+(3-8)	C+(3-8)	A/C	χ² (3)	p value
	Mean rank	Mean rank	Mean rank	Mean rank	χ² (3)	p value
Negative life events	14.13	33.59	29.30	28.04	15.06	.002
Total parenting stress	6.57	17.10	15.83	14.07	7.93	.047
Beck Depression	12.75	22.43	18.56	18.00	4.13	.248

DMM: Dynamic-Maturational Model of Attachment and Adaptation.

Sample size for analyses involving negative life events was 49. Due to missing data, sample sizes were 25 and 34 for analyses involving total parenting stress and Beck Depression Inventory, respectively. χ² is based on a Kruskal–Wallis test.

Comparisons across AAI classifications for trauma and loss, depression modifier and triangulation are shown in Table 2. Again, differences were in the expected direction for all three ways of clustering the attachment classifications. Differences by secure/insecure were not statistically significant, whereas the Ainsworth versus DMM differences were. There were also differences in the four-way clusters, with the mothers with Ainsworth classifications having less depression and fewer instances of complex trauma and loss than other mothers. All nine cases of triangulation were in the C+ and A/C groups.

Table 2.

Distress: depression modifier, the presence of trauma and loss and triangulation by secure versus insecure, Ainsworth classifications versus DMM classifications, and a four-way clustering of classifications.

		Secure (B)	Insecure (not B)		Fisher’s exact test
		n (%)	n (%)		p value
Depression modifier		1 (14.3)	19 (48.7)		.119
Trauma and loss					.097
None		4 (57.1)	7 (17.5)
Preoccupied or dismissed		1 (14.3)	11 (27.5)
Complex		2 (28.6)	22 (55.0)
Triangulation		0 (0.0)	9 (22.0)		.322
		Ainsworth: A1-2, B, C1-2	DMM: A+, C+, A/C		p value
		n (%)	n (%)		p value
Depression modifier		2 (13.3)	18 (58.1)		.005
Trauma and loss					.013
None		6 (40.0)	5 (15.6)
Preoccupied or dismissed		6 (40.0)	6 (18.8)
Complex		3 (20.0)	21 (65.6)
Triangulation		0	9 (100)		.022
	A1-2, B, C1-2	A+(3-8)	C+(3-8)	A/C	p value
	n (%)	n (%)	n (%)	n (%)	p value
Depression modifier	2 (13.3)	6 (54.5)	4 (50.4)	8 (66.7)	.031
Trauma and loss					.059
None	6 (40.0)	3 (27.3)	0 (0.0)	2 (16.7)
Preoccupied or dismissed	6 (40.0)	1 (9.1)	3 (33.3)	2 (16.7)
Complex	3 (20.0)	7 (63.6)	6 (66.7)	8 (66.7)
Triangulation	0 (0.0)	0 (0.0)	7 (70.0)	2 (16.7)	<.001

DMM: Dynamic-Maturational Model of Attachment and Adaptation.

Sample size for analyses involving triangulation was 49. Due to missing data, sample sizes were 46 and 47 for analyses involving depression modifier and trauma and loss, respectively.

There were eight mothers classified as partially reorganizing towards B (none were given the full reorganizing marker); their children were scattered across the PAA and SAA categories with no tendency for them to be classified as Secure (B), nor even in the Ainsworth categories.

We also tested whether marital status (i.e. married, living together, divorced or single) differed by AAI classification. When marital status was compared by mothers’ Ainsworth/DMM classification, the result was marginally significant (p = .065). The effect was complex because DMM risk mothers were more likely to be married, but not living together, than mothers in the Ainsworth cluster (see Table 3). Differences by both the secure/insecure and four-way grouping were not statistically significant (p = .228 and p = 200, respectively).

Table 3.

Mothers’ marital status by Ainsworth versus DMM-AAI classifications.

	Ainsworth (A1-2, B, C1-2)	DMM (A+, C+, A/C)
	n (%)	n (%)
Single	2 (12.5%)	6 (18.8%)
Divorced	1 (6.3%)	3 (9.4%)
Cohabiting, not married	6 (37.5%)	2 (6.3%)
Married	7 (43.8%)	21 (65.6%)

DMM: Dynamic-Maturational Model of Attachment and Adaptation; AAI: Adult Attachment Interview.

Fisher’s exact test: p = .065.

Hypothesis 3: transgenerational matching and meshing

To compare mother and child classifications as precisely as our sample size would permit, we clustered the classifications into six categories: A+, A1-2, B, C1-2, C+ and A/C.

Concordance on the PAA

For the PAA and AAI, there was a significant association between mothers’ protective strategy and their children’s strategy (p = .024; see Table 4). Notably, the majority of cases fell in the predicted cells. Overall, discounting B, the matching and meshing effects were about equal (12 of 39 cases of matching versus 11 of 39 cases of meshing).

Table 4.

The distribution of children’s six-way PAA classifications by mothers’ six-way AAI classifications.

Six-way AAI classifications	Six-way PAA classifications						Total
Six-way AAI classifications	A+	A1-2	B	C1-2	C+	A/C	Total
A+	3	1	0	1	4	2	11
A1-2	3	0	3	0	1	0	7
B	1	1	4	0	1	0	7
C1-2	0	0	0	1	0	0	1
C+	5	1	0	0	2	1	9
A/C	3	2	1	0	1	6	13
Total	15	5	8	2	9	9	48

PAA: Preschool Assessment of Attachment; AAI: Adult Attachment Interview.

Matches in bold; meshes in italics.

p = .024.

Concordance on the SAA

For the SAA and AAI, there was a significant relation between mothers’ protective strategy and their children’s strategy (p = .016; see Table 5). Again, the majority of cases fell within the predicted cells. Overall, discounting B, the matching and meshing effects were equal (10 of 39 cases each).

Table 5.

The distribution of children’s six-way SAA classifications by mothers’ six-way AAI classifications.

Six-way AAI classifications	Six-way SAA classifications						Total
Six-way AAI classifications	A+	A1-2	B	C1-2	C+	A/C	Total
A+	5	0	0	1	3	2	11
A1-2	1	1	1	2	2	0	7
B	1	0	5	0	0	0	6
C1-2	0	0	0	1	0	0	1
C+	5	0	0	0	2	1	8
A/C	6	2	1	0	2	2	13
Total	18	3	7	4	9	5	46

SAA: School-age Assessment of Attachment; AAI: Adult Attachment Interview.

Matches in bold; meshes in italics.

p = .016.

Although we tested for differences in the relation of mothers’ strategies to their children’s strategies between the normative and clinical groups, there was no suggestion of differences for either the PAA or the SAA, with proportions for matching and meshing being roughly equal across referral groups.

Matching and meshing as a function of birth order

There were seven first-born children whose PAA matched the mother’s AAI (including B and A/C) and nine cases of meshing. For later-born children, the numbers were eight matching and four meshing. On the SAA, nine first-borns matched the mother’s AAI (including B and A/C) and five meshed. For later-born children, the numbers were seven matches and five meshes. There were too few cases to analyse statistically, but the groups looked fairly similar.

Discussion

We address four topics regarding our study: (1) the validity of the DMM-AAI, on which the findings of this study depend, (2) findings about concordance of mothers’ and children’s protective attachment strategies, (3) findings regarding family composition and (4) understanding cases of children whose strategy changed. We close by considering the meaning of the findings for psychological treatment of troubled school-age children and their families.

The validity of the DMM-AAI

Referral group

There has been an accumulation of studies whose results support the validity of the DMM classificatory system and coding method as applied to the AAI. This study is consistent with previous studies, particularly the finding that parents of child psychiatric patients uniformly used DMM strategies that were marked by trauma, depression and disorientation (Landini et al., 2016) Nevertheless, we offer three caveats. First, there was a bias towards A2 in the normative group of this British sample; because our numbers are small, replication is needed to support or reject a cultural bias. Second, the families in our sample experienced greater disadvantage than the British population as a whole which skewed the normative distribution away from the Ainsworth classifications and towards the DMM classifications. Third, the normative volunteers may have over-represented mothers who were worried about their children and under-represented more adequate mothers who might have been disinclined to reveal intimate aspects of their family life. Even so, the DMM-AAI classifications fit our hypothesis about referral group.

Danger

The DMM classifications were validated by measures of stress and distress such that both were associated with the DMM classifications. Notably, fewer mothers used a Type B strategy than children. We think this occurred because (1) the mothers had lived longer and, thus, been exposed to more danger than their children; (2) the mothers buffered their children’s experience by providing protection and comfort; and (3) mothers in this low-income, risk community may have had limited opportunity for, and examples of, integrative reorganization. Put another way, both the community context of risk and individual histories of risk likely contributed to the lower proportion of secure classifications among the mothers than their children.

Severity of risk

The Ainsworth versus DMM dichotomy accounted for more variance than the secure versus insecure dichotomy that is used in most ABC+D studies. This suggests that A1-2 and C1-2 are not indicative of risk, and the DMM classifications offer more specificity than ‘disorganization’. Moreover, unlike the ABC+D studies that lost power when three-way and four-way analyses were used, the DMM analyses become stronger in the four-way and six-way comparisons. This suggests that adding precisely and theoretically defined classifications to the model increased the power of DMM attachment strategies (as compared to ABC+D) to discriminate among troubled dyads.

Specificity of risk: Type A versus Type C

Most studies using the ABC+D model have combined Types A and C. DMM theory proposes that these reflect opposite patterns of processing information that would place individuals using the strategies at risk for different problems. Our analyses found that Type A+ was associated with greater stress and distress, whereas Type C+ was associated with triangulation (see Table 2). Type A/C was associated with the most complex traumas and depression. These findings were not strong and our sample was small, but they suggest that future research be designed to explore A versus C versus A/C differences.

Psychological trauma and extremes of arousal

Our analyses make clear that it is the disruptions of strategic response, more than the extreme quality of some strategies, that signal maladaptation. Such disruptions can be brief and punctuate more strategic functioning (i.e. psychological trauma and unresolved loss) or ongoing (i.e. the modifiers of depression and disorientation). There appeared to be a gradient of risk from no traumas to mild traumas, complex traumas and modifiers with more traumas and traumas, with modifiers being associated with higher numbered A and C strategies (but not in every case). This suggests that high numbered Type A and C strategies can be functional and adaptive. It is also true, however, that the probability of there being psychological trauma or modifiers increases with the increase in the number of the strategy. Whether psychological trauma and unresolved loss and modifiers can be adaptive is unclear; Bowlby made a sharp differentiation between adaptation at the individual and species levels, with the notion being that individual casualties did not nullify the idea of a benefit to the species. In a number of studies using the DMM-AAI, the contribution of different forms of trauma, loss and modifiers has been demonstrated empirically for specific diagnostic groups (Crittenden & Heller, 2017; Crittenden & Newman, 2010; Ringer & Crittenden, 2007). For both post-traumatic stress disorder (PTSD; Crittenden & Heller, 2017) and borderline personality disorder (BPD; Crittenden & Newman, 2010), the signifiers of trauma functioned adaptively; for PTSD, information was retained in ways that led to both protective arousal and protective deactivation, whereas for BPD, intrusive arousal elicited protection from outside the family (i.e. the professional system). This study demonstrates the principle of adaptive management of danger in general, without fully differentiating among individual adaptation, family/community level adaptation, species level adaptation and maladaptation.

Findings regarding the concordance of mothers’ and their school-age children’s protective attachment strategies

Although we found evidence of concordance between mothers’ and children’s attachment, our findings differed from those using the ABC+D model in (1) being more precise (six-way comparisons, rather than secure/insecure), (2) extending the relation from infancy to children turning 6 years old and (3) showing a relation of discourse-based AAI classifications to both the enacted PAA and discourse-based SAA classifications. We considered birth order, but our data were too few to feel confident of supporting or refuting an association of first-born status with meshing mother–child classifications. Nevertheless, our findings suggest that children’s well-being in the school years still reflected important contributions from their mothers, particularly mothers’ ongoing psychological traumas.

Findings regarding family composition

Although it is generally assumed that having a father in the home and having married parents is desirable, our data suggest greater complexity. That is, having a father-figure in the home can be either protective or risky, depending on the parents’ relationship. When the relationship is supportive (i.e. Ainsworth A1-2, B and C1-2 strategies), marriage appears to benefit the children. But when parents quarrel, fight or triangulate their children, the effects can be harmful. Our data suggest that visible disagreement led to less psychological trauma in children than obscured struggles (e.g. parental denial of conflict, dichotomization of victim/aggressor roles and triangulation). This often appeared in both mothers’ and children’s assessments of psychological trauma around parental discord and triangulation, even when the events occurred in different generations. For an example from this study of the transgenerational effects of triangulation, see Brewerton, Robson and Crittenden (this issue); for other examples of triangulation, see ‘Denise’ (Crittenden et al., 2015) and ‘Melissa’ (Crittenden et al., 2015). In all the cases, the importance of fathers stands out sharply, albeit differently in each case.

Children’s change in classification from the PAA to the SAA

Ainsworth believed that it was important to examine the cases that did not fit the hypothesis because new understandings might lie in the errors. We hypothesized stability from the PAA to the SAA. Nevertheless, there were 12 cases of change.

Table 6 indicates that, despite our optimistic rationale for positive change, only one of the 12 children changed in the direction of B. In this case, the school context was unlikely to be the reason. Instead, the mother’s AAI suggests that the problem was a brief, but powerful, life event that disrupted the mother’s reorganizing process from an A+/C+ strategy towards B. The event was the recent death of her baby. At the time of the mother’s AAI, the loss was not only visible as an unresolved loss, but it had also expanded to irrational anticipation of other deaths. Notably, this loss occurred in the complex context of the mother’s ‘loss’ of her father through his gender change. The mother’s daughter was reorganizing from compulsive caregiving (A3) on the PAA to B4, with unresolved loss of the baby, on the SAA. We would expect that as the mother recovered (a process that can take several years), so will her daughter, yielding a future B/B match of mother to child.

Table 6.

Change in classification from the PAA at 5.5 years to the SAA at 6.0 years, with mothers’ AAI.

ID	Clinical referral	No. negative life events	PAA classification	SAA classification	AAI classification
2	Yes	1	C4	Dp Utr(dx or dp)_{PA, PN, CSA} A, [ina]i	Utr(dx)_dv tr(b)_csa? tr(p)_PEN tr(p,ds)_{F leaving} [A] C5-6Δ
3	Yes	3	Dp A+	Utr (p)_bullying C4	Dp Utr(dp)_EAN (b)_CSA l(dp)_{F, ch} A4(5) [ina]h
16	No	2	B1-2	C4(3)	Utr(p&ds)_{boarding school} A2(3)
21	Yes	2	~~B4/B3/A3/C3~~	Dp A+ (A4 or A4-) [ina]	Utr(p)_bully (p)_{M’s suicide attempt} (ds, dpl)_PAN l(p)_MGM (a)_ch R(A3_M 4_F 7_MGP/C3-4 → BO)
28	No	0	~~A3/B4~~	C2(4)	Ul(p)_F A1-2
33	No	1	A3	Ul(p)_baby B4	R(Utr(p)_{F sex change} l(dpl)_baby (a)_all A+C+ → BO)
35	Yes	0	~~B3/A3/A4~~-	(dp) Utr(p)_F?C4(3)	Utr(p)_PEA C4(5)
40	Yes	8	C3/C3	(dp) Utr(dx)_{F leaving} A+(4,5) [ina]i	Utr(p)_{SF aban} (b)_{CSA Uncle Phil} (dn)_PAN l(dx) _{bio F/SF/PGF} (dx)_{Uncle Phil, 3 friends} A7
46	Yes	2	C3	Dp A3-4	Utr(dx)_sex A7_MGP R(Utr(p)_PAN l(dx)_son A3_M6 → BO)

PAA: Preschool Assessment of Attachment; SAA: School-age Assessment of Attachment; AAI: Adult Attachment Interview.

Classifications – ~~strikeout~~: changed classifications by coders; [A]: false A; BO: secure other (quasi-B that does not fit B1-2, B3 or B4-5 criteria; common during reorganization towards B). Modifiers – Dp: depressed; (dp): partial depression; R: reorganizing; Δ: triangulated between parents; [ina]: intrusions of forbidden negative affect/behaviour; [ina]h: [ina] in the history (not occurring during the AAI); [ina]I: incipient [ina]s during the AAI (usually prevented by adroit interviewers). U types – U: unresolved; tr: trauma; l: loss; dx: disorganized; dp: depressed; b: blocked from recall; p: preoccupying; ds: dismissed; dpl: displaced; a: anticipated; dn: denied. U events – PA: physical abuse; PN: physical neglect; PAN: physical abuse and neglect; PEN: physical and emotional neglect; PEA: physical and emotional abuse; EAN: emotional abuse and neglect; CSA: child sexual abuse; DV: domestic violence; aban: abandonment. U persons – F: father; M: mother; ch: child; MGM: maternal grandmother; MGP: maternal grandparents; all: all family members; SF: stepfather; PGF: paternal grandfather.

The remaining 11 cases pose the dilemma of why the PAA and SAA classifications differed. Notably, the PAA has no mechanism for representing psychological trauma. If psychological traumas were present, the PAA classification would not reflect them. Table 6 shows that three of the PAA classifications were disputed; all three had complex SAA classifications. In each case, once the mother’s AAI classification was known, the complexity of the child’s situation became clear. This suggests that the PAA lacked constructs to capture the complex situation. The SAA, however, offered constructs that permitted children to represent danger in various forms of psychological traumas and modifiers.

Alternatively, the assessments might accurately have identified change. Our data cannot disentangle these alternatives, but the mothers’ AAIs suggest that these 11 mothers brought obscure problems and complex strategies (A/C, false A, delusional idealization, etc.) to their child-rearing. Indeed, the classifications of their AAIs suggest that even the mothers could not fully articulate their problems. In nine cases, the child had been referred for diagnosis. If, as appears to be the case, the children’s increasing maturity yielded more complex protective strategies, this would bode well for both children’s spontaneous adaptation and the possibility of intervention to shape the process.

The cases of change highlight the importance of understanding the psychological functioning of mothers. Without the mothers’ AAIs, it was not possible to understand the discontinuity in children’s the PAAs and SAAs. We did not assess fathers, but the mothers’ AAIs suggest the important contributions of their fathers to their protective organization. There is every reason to think that the fathers of the children in our study – and in every family – are equally important (see Crittenden, Dallos, Landini, & Kozlowska, 2014).

In conclusion, it seems to us that development is dynamic, unique and comprehensible. It may not always be simple for either family members or professionals to understand, but it is not ‘disorganized’ in the sense of lacking meaning and structure (intra- or interpersonally) To the contrary, it is exquisitely organized to fit the particular conditions of each child’s experience. Moreover, children’s protective strategies change as they develop new capacities for understanding and as their experience changes. We think that theory and assessments need to be attuned to this process if they are to be helpful to children and families.

Limitations and future directions

Many of our findings were significant, but the effect sizes were not great. This is a reminder that statistical significance is not equivalent to the predominant effect in a sample. Instead, it only means that there is an effect, even if small. This expansion of our previous report (Crittenden et al., 2010) suffers from the same limitations: small sample size, the cultural homogeneity of the sample, missing data on the self-report measures and questions about the validity of the self-report measures. Moreover, coding requires highly skilled coders, and most clinicians do not achieve Level I or II reliability. We note the substantial difference between Level II coders’ agreement and that of Level I coders, with only Level I codings being suitable for court use. In the light of the studies of siblings (Farnfield, this issue; Kozlowska & Elliott, 2016), replication should include siblings and, of course, fathers. As Farnfield suggested, longitudinal follow-up and information from school personnel can reveal developmental pathways. For example, see the longitudinal case study of a boy ultimately diagnosed with autism (Brewerton et al., this issue). Such studies increase our ability to intervene before developmental processes solidify and limit children’s futures – and those of their children.

Implications for treatment of troubled school-age children

School-age children can appear quite independent and able to respond to treatment as individuals. Nevertheless, our finding of a relation between children’s and mothers’ protective strategies suggests the need to work with at least mothers and probably also families, including fathers. Moreover, simple main effect hypotheses were insufficient to explain the real-life diversity and complexity of this sample of mild and high-risk families. Although identifying family members’ protective strategies and history of past endangerment is important, it seemed insufficient to guide treatment. Instead, both the psychological processing of information, especially the clarity of processing, around the dangerous events and the inter-twining of parents’ and children’s strategies seem crucial. Accounting for these moves the evaluative process from assessment to Family Functional Formulation (Crittenden & Dallos, 2014). With sufficiently nuanced theory and sound assessment structured in a Family Functional Formulation, clinicians should be able to guide children and their families to safer developmental pathways. Farnfield (this issue) called for family case studies, but we would like to point out that when treatment is needed, every family is different. With a structured Family Functional Formulation, each can be understood for its unique presentation of universal principles. Planned sets of structured family case studies could guide clinicians to use a defined approach to case work while concurrently addressing some of the complex and unanswered questions about how threatened family adapt.

Footnotes

Acknowledgements

The authors wish to thank the families and children who participated in this reasearch. In addition, we are grateful to the following individuals and agencies for their assistance: Margaret Randall, Consultant Child Psychotherapist; Nicky Brewerton, Head, Ramsden Infant School, Barrow-in-Furness; Attachment and Play Therapy; Cumbria Partnership NHS Foundation Trust, Cumbria County Council; and Walter Tooby and Tooby’s Electrical for technological assistance and equipment. In addition, they thank Susan Spieker and Simon Wilkinson for their comments on early drafts of this paper.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship and/or publication of this article.

Author biographies

Patricia M Crittenden, MEd, PhD, is a developmental psychopathologist who studied under Mary D. S. Ainsworth. Her specializations are attachment and child protection. She has developed a model of attachment strategies across the lifespan (the Dynamic-Maturational Model of Attachment and Adaptation) together with assessments of attachment across the age range. She has been on several faculties internationally and published more than 100 empirical papers and chapters and several books on attachment, child protection, and parenting.

Katrina Robson, MSW, is a social worker and child therapist who has over thirty years’ experience working with children and families in statutory public services and the third sector. Katrina developed and leads the Love Barrow families (LBF) project. LBF uses the DMM to inform intervention with multi-problem families; the outcome has been to transform how public services are delivered. She has published a chapter about Love Barrow Families and several papers on DMM applications.

Alison Tooby, MSW, has over 20 years’ experience of working in local authority children’s services and is currently the lead social worker for Love Barrow families. She has also worked as a child therapist for the NSPCC and independently for the courts. She is a founding member of IASA and uses DMM assessments with families to inform her own practice and that of her team.

Charles Fleming, MA, is a research scientist at the Department of Psychiatry and Behavioral Sciences and the School of Nursing at the University of Washington. He has worked for over 20 years as a data manager and analyst on studies of preventive and treatment interventions and the etiology of prosocial and antisocial behaviors among children and adolescents.

References

Abidin

R. R.

(1990). Parenting stress index (3rd ed.). Psychological Assessment Resources. doi: https://dx-doi-org.web.bisu.edu.cn/10.1037/t02445-000

Ainsworth

M. D. S.

(1979). Attachment as related to mother-infant interaction. Advances in the Study of Behavior, 9, 2–52.

Ainsworth

M. D. S.

Blehar

Waters

Wall

(1978). Patterns of attachment: A psychological study of the strange situation. Hillsdale, NJ: Lawrence Erlbaum.

Beck

A. T.

Steer

R. A.

Brown

G. K.

(1996). Beck Depression Inventory (BDI-II). San Antonio, TX: The Psychological Corporation.

Beck

A. T.

Ward

C. H.

Mendelson

Mock

Erbauch

(1961). An inventory for measuring depression. Beck Depression Inventory, 4, 561–571.

Bowlby

(1979). The making and breaking of affectional bonds. London, England: Tavistock Press.

Bowlby

(1980). Attachment and loss. Vol. III: Loss. New York, NY: Basic Books.

Brewerton

Robson

Crittenden

P. M.

(2017). Formulating autism systemically: Part 2: A 12-year case study. Clinical Child Psychology and Psychiatry, 22(3), 390–401

Cochrane

Robertson

(1973). The Life Events Inventory: A measure of the relative severity of psycho-social stressors. Journal of Psychosomatic Research, 17, 135–139.

10.

Crittenden

P. M.

(1992). Preschool Assessment of Attachment Manual (Unpublished manuscript). Miami, FL.

11.

Crittenden

P. M.

(1999). Danger and development: The organization of self-protective strategies. In Vondra

J. I.

Barnett

(Eds.), Atypical attachment in infancy and early childhood among children at developmental risk. Monographs of the Society for Research on Child Development (Vol. 64, 258, pp. 145–171). Malden, MA: Wiley-Blackwell.

12.

Crittenden

P. M.

(2005). School-age assessment of attachment coding manual (Unpublished manuscript). Miami, FL. (Original work published 1997)

13.

Crittenden

P. M.

(2015). Attachment and psychopathology. In Crittenden

P. M.

Landini

Baim

Hart

Landa

(Eds.), Danger, development, and adaptation: Seminal papers on the dynamic-maturational model of attachment & adaptation (pp. 297–340). Hook, UK: Waterside Press. (Original work published 1995)

14.

Crittenden

P. M.

(2015). Raising parents: Attachment, representation, and treatment (2nd ed.). London, England: Routledge. (Original work published 2008)

15.

Crittenden

P. M.

(2015). Understanding children: Assessing school-aged children’s self-protective strategies. Clinical Child Psychology and Psychiatry, 20, 341–347.

16.

Crittenden

P. M.

Ainsworth

M. D. S.

(1989). Child maltreatment and attachment theory. In Cicchetti

Carlson

(Eds.), Handbook of child maltreatment (pp. 432–463). New York, NY: Cambridge University Press.

17.

Crittenden

P. M.

Claussen

A. H.

Kozlowska

(2007). Choosing a valid assessment of attachment for clinical use: A comparative study. Australia New Zealand Journal of Family Therapy, 28, 78–87.

18.

Crittenden

P. M.

Dallos

R. D.

(2014). An attachment approach to treatment: DMM-FST integrative treatment. Clinical Child and Family Psychology Review, 2, 53–61.

19.

Crittenden

P. M.

Dallos

R. D.

Landini

Kozlowska

(2014). Attachment and family therapy. London, England: McGraw Hill.]

20.

Crittenden

P. M.

DiLalla

D. L.

(1988). Compulsive compliance: The development of an inhibitory coping strategy in infancy. Journal of Abnormal Child Psychology, 16, 585–599.

21.

Crittenden

P. M.

Heller

M. B.

(2017). The roots of chronic posttraumatic stress disorder: Childhood trauma, information processing, and self-protective strategies. Chronic Stress. Advance online publication. doi:10.1177/2470547016682965

22.

Crittenden

P. M.

Kozlowska

Landini

(2010). Assessing attachment in school-age children. Clinical Child Psychology and Psychiatry, 14, 185–208.

23.

Crittenden

P. M.

Landini

(2011). Assessing adult attachment: A dynamic-maturational approach to discourse analysis. New York, NY: W. W. Norton.

24.

Crittenden

P. M.

Landini

Kozlowska

(2015). The effect of parents’ psychiatric disorder on children’s attachment: Theory and cases. In Reupert

Maybery

Nicholson

Seeman

Göpfert

(Eds.), Parental psychiatric disorder: Distressed parents and their families (3rd ed., pp. 29–41). Cambridge, UK: Cambridge University Press.

25.

Crittenden

P. M.

Newman

(2010). Comparing models of borderline personality disorder: Mothers’ experience, self-protective strategies, and dispositional representations. Clinical Child Psychology and Psychiatry, 15, 433–452.

26.

Crittenden

P. M.

Partridge

M. F.

Claussen

A. H.

(1991). Family patterns of relationship in normative and dysfunctional families. Development and Psychopathology, 3, 491–512.

27.

Crittenden

P. M.

Robson

Tooby

(2015). Longitudinal, concurrent, and construct validation of the School-age Assessment of Attachment. Clinical Child Psychology and Psychiatry, 20, 348–365.

28.

Dallos

Lakus

Cahart

M.-S.

McKenzie

(2016). Becoming invisible: The effect of triangulation on children’s well-being. Clinical Child Psychology and Psychiatry, 21, 461–476.

29.

Department of Health. (2000). Daily parenting hassles: Framework for the assessment of children in need and their families: The family pack of questions and scales. London, England: Author.

30.

Farnfield

(2017). Fix my child: The importance of including siblings in clinical assessments. Clinical Child Psychology and Psychiatry, 22(3), 421–435

31.

Farnfield

Hautamäki

Nørbech

Sahhar

(2010). Dynamic-maturational model methods for assessing attachment. Clinical Child Psychology and Psychiatry, 15, 313–328.

32.

George

Kaplan

Main

(1996) The adult attachment interview: Interview protocol (Unpublished manuscript). University of California, Berkeley. (Original work published 1985)

33.

Hagnell

Öjesjö

Otterbeck

Rorsman

(1994). Prevalence of mental disorders, personality traits and mental complaints in the Lundby Study. Scandanavia Journal of Social Medicine, Supplementum, 50, 1–77.

34.

Hautamäki

(2010). Transmission of attachment across three generations: Continuity and reversal. Clinical Child Psychology and Psychiatry, 15, 347–354.

35.

Iyengar

Kim

Martinez

Fonagy

Strathearn

(2014). Unresolved trauma in mothers: intergenerational effects and the role of reorganization. Frontiers in Psychology, 5, 9661–9669. doi:10.3389/fpsyg.2014.00966

36.

Kessler

R. C.

(1994). The national comorbidity survey of the United States. International Review of Psychiatry, 6, 365–376.

37.

Kozlowska

Elliott

(2014). School-age assessment of attachment. In Holmes

Farnfield

(Eds.), Attachment: The guidebook to attachment (Vol. 2, pp. 102–118). London, England: Routledge.

38.

Kozlowska

Elliot

(2016). Don’t forget the siblings: School-aged siblings of children presenting to mental health services show at-risk patterns of attachment. Clinical Child Psychology and Psychiatry, 1–15. doi: 10.1177/1359104516653993

39.

Landini

Crittenden

Landi

(2016). The parents of child psychiatric patients. Annals of Psychiatry and Mental Health, 4, 1087.

40.

Main

Weston

D. R.

(1981). The quality of toddlers’ relationship to mother and father: Related to conflict behavior and readiness to establish new relationships. Child Development, 52, 932–940.

41.

McLaughlin

K. A.

Greif Green

Gruber

M. J.

Sampson

N. A.

Zaslavsky

A. M.

Kessler

R. C.

(2012). Childhood adversities and first onset of psychiatric disorders in a national sample of US adolescents. Archives of General Psychiatry, 69, 1151–1160.

42.

Ringer

Crittenden

(2007). Eating disorders and attachment: The effects of hidden processes on eating disorders. European Eating Disorders Review, 15, 119–130.

43.

Roberts

R. E.

Attkisson

C. C.

Rosenblatt

(1998). Prevalence of psychopathology among children and adolescents (special article). American Journal of Psychiatry, 155, 715–725.

44.

Rutter

(1993). Developing minds: Challenge and continuity across the lifespan. New York, NY: Basic Books.

45.

Schepank

(Ed.) (1987). Epidemiology of psychogenic disorders. The Mannheim study: Results of a field survey. Berlin, Germany: Springer.

46.

Shah

P. E.

Fonagy

Strathearn

(2010). Exploring the mechanism of intergenerational transmission of attachment: The plot thickens. Clinical Child Psychology and Psychiatry, 15, 329–346.

47.

Spieker

S. J.

Crittenden

P. M.

(2010). Comparing two classification methods applied to preschool strange situations. Clinical Child Psychology and Psychiatry, 15, 97–120.

48.

STATA Corporation. (2011). Intercooled STATA 12.1 for Windows. College Station, TX: Author.

49.

Verhage

M. L.

Schuengel

Madigan

Fearon

R. M. P.

Oosterman

Cassibba

. . .van IJzendoorn

M. H.

(2015). Narrowing the transmission gap: A synthesis of three decades of research on intergenerational transmission of attachment. Psychological Bulletin, 142, 337–366. doi:10.1037/bul0000038