Introduction
This paper broadens the scope of earlier analyses linking parental maltreatment with adult depression (Brown et al., Reference Brown, Craig, Harris, Handley and Harvey2007a, Reference Brown, Craig, Harris, Handley and Harvey2007b, Reference Brown, Craig, Harris, Handley and Harvey2007c, Reference Brown, Craig and Harris2008a, Reference Brown, Craig and Harris2008b) by exploring the contribution of insecure attachment style both in general and in terms of enmeshed, dismissive, fearful and withdrawn insecure types. Finally, it considers a link with whether an episode took a chronic course.
Relating styles such as dependence (Birtchnell et al., Reference Birtchnell, Deahl and Falkowski1991), shyness (Simpson and Stevenson-Hinde, Reference Simpson and Stevenson-Hinde1985) and hostility (Philip, Reference Philip1973; Arrindell et al., Reference Arrindell, Hafkenscheid and Emmelkamp1984) have been included in aetiological models of depression with one in the late 1970s covering ‘dependence’, ‘nurturance’ and ‘hostility’, along with ‘helplessness’ (Harris et al., Reference Harris, Brown and Bifulco1990). However, John Bowlby's trilogy (Bowlby, Reference Bowlby1980, Reference Bowlby1982, Reference Bowlby1985), with its emphasis on the differing impact of loss in terms of attachment patterns, conveyed the importance of combining such personality characteristics in ways which reflect overall styles of relating to others. This, and Mary Ainsworth's research with children, resulted in Mary Main's Adult Attachment Interview which explores how individuals responded to childhood relationships in terms of secure/autonomous, insecure ambivalent/enmeshed and insecure avoidant/dismissive categories (George et al., Reference George, Kaplan and Main1984). One of our colleagues (Antonia Bifulco) proposed that combinations of high v. low dependency, nurturance and hostility approximated this scheme in terms of adulthood relationships, with this found associated with clinical depression (Harris and Bifulco, Reference Harris, Bifulco, Parkes and Hinde1991). The index concerning insecure attachment reflects current adult experience and has been found to relate prospectively to new onsets of depression (Bifulco et al., Reference Bifulco, Moran, Ball and Bernazzani2002). More recently, other measures have been introduced and confirmed this link (Mikulincer and Shaver, Reference Mikulincer and Shaver2016, Table 13.3).
The present analysis is concerned to locate such attachment within a broad psychosocial aetiological model of depression and to explore its link with earlier childhood experience, particularly parental maltreatment. Three core contributions have been shown to play proximal roles. First, studies using the Life Events and Difficulties Schedule (LEDS) have found that severely threatening life events provoke most depressive episodes with these most often involving loss when defined broadly, but with humiliation markedly increasing risk irrespective of any involvement with loss (Brown and Harris, Reference Brown and Harris1978). Second, given a severely threatening event of any kind, the presence of psychosocial vulnerability associated with low self-esteem or a severe difficulty in a core relationship further increases risk (op cit.). Third, once an onset has occurred, perpetuating factors, such as an ongoing interpersonal difficulty, increase the chance of an episode taking a chronic course of at least 12 months (Brown et al.,Reference Brown, Bifulco and Andrews1990; Nanni et al., Reference Nanni, Uher and Danese2012). In addition, it has proved important to consider distal risk factors – for example, that parental maltreatment is associated with all three proximal types of risk (Brown et al., Reference Brown, Craig and Harris2008b), including chances of a depressive episode taking a chronic course (Fig. 1 provides a summary). This paper will explore the role of attachment style as another distal risk factor that may link with adult depression by mediating between types of parental maltreatment and psychosocial vulnerability.
Fig. 1. Four aetiological mechanisms.
Methods
Samples and data collection
The study uses pairs of adult sisters with a two-phase sampling strategy (Brown et al., Reference Brown, Craig, Harris, Handley and Harvey2007a). First, women registered at two Islington general practitioner Health Centres in North London were contacted by postal questionnaire to identify those under 50 who had a sister within 5 years of age with whom they had been brought up. This initial approach included the Childhood Experience of Care and Abuse Questionnaire (CECA-Q, Bifulco et al., Reference Bifulco, Bernazzani, Moran and Jacobs2005) as a screen to identify those likely to have had an adverse childhood. Possible positive responses were followed up by telephone or brief face-to-face interviews. Those meeting a formal definition of severe maltreatment were invited to participate and subsequently asked for permission to contact a sister to carry out a similar interview with another interviewer. Somewhat more than half the pairs of sisters in the whole study were selected in this way because of the likelihood of the first sister's experience of maltreatment (the high-risk series). In addition, we included an unselected or comparison series of 80 women, again in terms of pairs, with the first recruited from the same general practice register, but now not taking account of the likelihood of maltreatment, and therefore providing a provisional estimate of the population frequency of such risk factors as insecure attachment, and of the generalisability of the findings of the high-risk series (see Bifulco et al., Reference Bifulco, Brown, Moran, Ball and Campbell1998; Brown et al., Reference Brown, Craig, Harris, Handley and Harvey2007a for details). The age of the two samples ranged from 17 to 55 with an average of 35 years.
The research interview, usually based on two visits, was tape-recorded and relevant extracts transcribed with measures of insecure attachment developed for the enquiry (Bifulco et al., Reference Bifulco, Moran, Ball and Bernazzani2002). All ratings were investigator-based with team members, blind to the psychiatric material, checking key social ratings at consensus meetings. Extensive manuals, covering the various rating scales, were used during training and the research itself. All measures (see below) had achieved an inter-rater reliability of around a minimum of 0.80 (Kw; Cohen, Reference Cohen1968).
Measures
Insecure attachment was measured using the Attachment Style Interview (ASI; Bifulco et al., Reference Bifulco, Moran, Ball and Bernazzani2002) based on detailed questioning about ongoing close relationships using rating scales concerning typical adulthood behaviour and attitudes with a particular attention to those that were potentially dysfunctional covering ‘constraints on closeness’, ‘fear of rejection’, ‘mistrust in relationships’, ‘fear of separation’, ‘desire for engagement’, ‘self-reliance’, ‘anger in relationships’ and ‘overall ability to make and maintain close supportive relationships’. Although dealing with current relationships, the interview could cover incidents and behaviour at any point in a relationship as a whole. If a woman currently reported some irritability or social withdrawal, along with clinical depression, interviewers were instructed to question about how typical these tendencies were and whether pre-onset behaviour had been different. Frequent probes encouraged comments with, for example, for the rating of ‘attitudinal constraints on closeness’ asking whether she had ‘difficulty in getting close’, ‘confiding’ or ‘asking anyone for help’, and ‘whether she was too proud to approach others’, or ‘there were persons she would not approach’ or ‘experienced feelings of nervousness about doing so’. The presence of insecure attachment itself was rated in terms of an ‘overall ability to make and maintain close supportive relationships’. Further scales were used to allocate women to the four types of insecure attachment: enmeshed, angry-dismissive (henceforth simply called dismissive), fearful and withdrawn. Those rated only mildly insecure were placed with the securely attached. Figure 2 provides brief definitions together with the scales considered in making the basic allocations.
Fig. 2. Four types of insecure attachment with relevant subscales.
Parental maltreatment involves three scales (rated 2 – marked, 1 – moderate and 0 – none) from the CECA instrument (Bifulco et al., Reference Bifulco, Brown and Harris1994) concerning experience before age 17, using mother's lack of affection, mother's rejection and father's physical abuse. Surrogate mothers and fathers were included. A high threshold was taken with, for example, a mother's rejection rated ‘moderate’ based on comments such as ‘She was cold and distant’ and ‘made it clear she would prefer not to be bothered’, and ‘who wished me away’. The scales give a score for parental maltreatment ranging from 0 to 6 with 2 or more defining severe threat (Brown et al., Reference Brown, Craig, Harris, Handley and Harvey2007a, Reference Brown, Craig, Harris, Handley and Harvey2007b). In the present analysis, we have added to the 113 women selected in this way an additional 10 who, although scoring ‘moderate’ on only one of these three scales, had undergone particularly intrusive or shameful experiences concerning either: (i) sexual abuse involving age incongruent physical contact that would be deemed frightening or distasteful by most children, or (ii) a mother's peak physical abuse in childhood involving hitting with an implement or around the head with a hand in a way that could cause damage.
Schedule for Clinical Assessment in Neuropsychiatry (SCAN) (Wing et al., Reference Wing, Barbor and Brugha1990) was used to assess major depression throughout adulthood targeting the 3 and 10 years before interview. The threshold had an almost total overlap with DSM-3-R criteria. A depressive episode defined as chronic had to have lasted at least 12 months after the age of 17 (Brown et al., Reference Brown, Craig, Harris, Handley and Harvey2007a, Reference Brown, Craig, Harris, Handley and Harvey2007b).
Chaotic life style covers behaviour after the age of 17 lasting at least 2 years and involving at least three of ‘heavy drinking’, ‘substance use’, ‘promiscuity’, ‘anti-social behaviour’, ‘criminal behaviour’ and ‘several highly troubled relationships’, with the style usually present at the time of leaving home. The behaviour was not necessarily persistent with some showing an overall improvement by early adulthood. In instances where a partner led such a life style, a woman was rated positive only if she took part.
Violence exhibited in childhood. Extensive questioning dealt with eight kinds of conduct problems, such as attacking property, cruelty to pets and bullying other children before the age of 17 (Brown et al., Reference Brown, Craig, Harris, Handley and Harvey2007c).
Anger shown in adult relationships is part of the measure of insecure attachment covering the extent of feeling ‘hostile’, ‘resentful’, ‘jealous towards others’, including parents, siblings, partners, children, confidants, colleagues, neighbours and other friends. A positive rating could be made even if the anger was not directly expressed, with particular weight given to apparent over-reactions to minor ‘slights’.
Quality of relationship of cohabiting partners covers partners of either sex using the Camberwell Family Interview (Brown and Rutter, Reference Brown and Rutter1966; Rutter and Brown, Reference Rutter and Brown1966; Quinton et al., Reference Quinton, Rutter and Rowlands1976; Bifulco et al., Reference Bifulco, Bernazzani, Moran and Ball2000) with ratings based on questions about actual behaviour as well as comments made spontaneously. Important changes during a relationship were recorded with ‘good’ involving mutual concern and affection and the remaining part of the scale having two sections. For the first dealing with ‘discord’, a poor relationship involved important episodes of open disruption or hostility, but with periods of harmony or neutrality, and very poor open antagonism, frequent quarrelling and recrimination. The second section dealt with ‘apathy’ with poor typified by indifference, dislike or avoidance, despite some one-sided affection or co-operation in mutual activities and very poor by marked apathy or dislike/avoidance.
Low self-esteem at interview uses all negative comments made spontaneously about self or in response to standard questions and defined by the presence of either negative evaluation about: (i) personal attributes, (ii) role performance or (iii) low self-acceptance (Brown et al., Reference Brown, Craig, Harris, Handley and Harvey2007c for details).
Shyness in adolescence based on questions from CECA about social life, confidence with peers and popularity in teenage years.
Severe life events using the LEDS. It was impractical to use the full LEDS schedule to cover a woman's life in terms of threatening life events and difficulties. However, material was collected systematically by the Adult Life Phase Interview (ALPHI; Bifulco et al., Reference Bifulco, Bernazzani, Moran and Ball2000) about the period preceding any onset of adult episodes of depression. Attention was given to events and difficulties occurring in the context of core relationships, most often involving a partner or lover such as rows, separations, violence, infidelity, miscarriages and deaths, although other core relationships were occasionally involved. These were rated on contextual grounds in terms of the likelihood of marked threat or unpleasantness. For example, a severe rating of a separation from a boyfriend would typically only be justified to the extent it was unexpected and the extent of the threat likely to have been involved (Brown and Harris, Reference Brown and Harris1978 for details). A rating of ‘severe humiliating event’ was made to the extent it markedly devalued a woman in terms of: (i) a separation involving failure and rejection, (ii) delinquency of someone close, in most instances involving a child, or (iii) a ‘put down’ of a central aspect of self-identity with this often-involving violence or threats of violence (Brown et al., Reference Brown, Harris and Hepworth1995). A supplementary index was also employed recording whether such a humiliating crisis had occurred during each partnership of a woman irrespective of whether it had been associated with depression.
Analysis plan
We deal first with the link of insecure attachment with depression, then with four types of insecure attachment, and finally with such experience and depression taking a chronic course. The Statistical Package for the Social Sciences (SPSS22) is used, including odds ratios (ORs) and logistic regression. Tests for statistical significance of per cent differences use χ2, with degrees of freedom given if these are two or more.
Results
Insecure attachment, parental maltreatment and depression during a 3-year period
Of the total 198 women, 61 experienced depression in a 3-year period. Parental maltreatment and insecure attachment were highly correlated (OR of 7.7, p < 0.0001) with both independently associated with depression with partial ORs of 3.8 (p < 0.003) for maltreatment and 4.0 (p < 0.0001) for attachment with this holding for the high-risk and general population samples (Table 1). As many as 28% (17/61) of the depressive episodes were still present at the research interview, but the link with insecure attachment remained essentially unchanged when these were excluded.
Table 1. Insecure attachment, parental maltreatment and depression in 3-year period for all women and comparison series
After taking parental maltreatment into account, insecure attachment explained 37% (22.5/61) of the depressive episodes with this holding for both high-risk and general population women (see Appendix for details).
The four types of insecure attachment and depression in the 3-year period
The four types of insecure attachment were associated with broadly the same marked risk of depression in the 3-year period with 50% (9/18) holding for the dismissive, 88% (7/8) enmeshed, 55% (17/31) fearful and 40% (6/15) withdrawn. However, key aetiological differences emerge in terms of the types of psychosocial risk factors involved when the dismissive and enmeshed are compared with the fearful and withdrawn.
The dismissive/enmeshed and childhood and adult risk factors
Earlier research has shown that severely threatening events involving humiliation double risk of depression compared with those concerning loss (Brown et al., Reference Brown, Harris and Hepworth1995). In terms of women reporting an episode in the 3-year period as many as 75% (12/16) of those of dismissive/enmeshed were linked with humiliation compared with 38% (17/45) of the remaining women with depression (χ2 = 6.56, p = 0.01) – with for the latter 39% (9/23) holding for fearful/withdrawn, 41% (7/17) for securely attached with parental maltreatment and 20% (1/5) for other women. At this point in the analysis, given this link of the dismissive/enmeshed with humiliation, we will focus on the role of such events. Here the experience of persistent anger as an adult was of importance with as many as 80% (21/26) of dismissive/enmeshed positive compared with just 13% (22/172) of other women (χ2 = 61.39, p < 0.00001) (Table 3, column 1). Such anger, but only for the dismissive/enmeshed, was highly associated with a mother's physical abuse as a child (Table 2, column 3). To obtain an estimate of the overall aetiological contribution of such adult anger, the number of women experiencing at least one episode of depression will be increased by taking account of its presence in a 10-year period (Table 3). When restricting analysis to episodes associated with humiliation as many as 88% (15/17) of the dismissive/enmeshed with such a provoked onset in the 10 years had shown persistent anger as an adult compared with just 17% (6/36) of other women (χ2 = 24.72, p < 0.00001).
Table 2. Type of insecure attachment and two key distal risk factors
Table 3. Humiliation-provoked depressive episodes by parental maltreatment and type of insecure attachment
The link of adult anger, often associated with a mother's earlier physical abuse, with the experience of humiliation raises the possibility that it was a more common experience among the dismissive/enmeshed irrespective of any link with adult depression. Here women were also questioned about the presence of at least one humiliating crisis during each marital-type partnership with this occurring on average in 0.75 (36/48) of the partnerships of the dismissive/enmeshed irrespective of any link with depression compared with a lower rate of 0.34 (83/246) for the fearful/withdrawn and secure maltreated with a similar experience and an overall relative risk of 2.21 (p < 0.01).
A chaotic life style involving conduct disorder, promiscuity and excessive alcohol or drug use originating by early adulthood was present for 39% (10/26) of dismissive/enmeshed, 22% (10/46) of fearful/withdrawn and 3% (4/126) of other women (χ2 = 30.40, 2df, p < 0.00001). However, since for the dismissive/enmeshed, the life style largely involved those reporting a mother's physical abuse together with her own anger, it did not add further to the risk of humiliating provoked depression.
The insecure fearful/withdrawn and ongoing vulnerability
A different aetiological pathway played a key role for the fearful/withdrawn, involving either low self-esteem or an ongoing ‘very poor’ core relationship both established vulnerability factors increasing the risk of a depressive onset in the presence of a severely threatening life event (Brown et al., Reference Brown, Bifulco and Andrews1990). As many as 91% (42/46) of the fearful/withdrawn had at least one such risk factor compared with 62% (16/26) of dismissive/enmeshed, 50% (30/60) of secure maltreated and 35% (23/66) of those securely attached (χ2 = 36.6, 3df, p < 0.0001) with this difference sufficient to explain their greater experience of depression (Table 4).
Table 4. The insecure fearful/withdrawn and depression in 3-year period
Consistent with a life course perspective, 61% (28/46) of the fearful/withdrawn reported shyness as an adolescent compared with 37% (32/86) of the dismissive/enmeshed and secure maltreated (χ2 = 5.86, p < 0.02). This in turn was linked with the experience of current low self-esteem with 86% (24/28) of such women among the fearful/withdrawn reporting shyness compared with 37% (13/35) of other women (χ2 = 15.14, p < 0.0001).
Chronic depression and the role of a chaotic life style
During a 10-year period, a chronic depressive episode defined as lasting at least 12 months held for 35% (41/118) of high-risk and 14% (11/80) of the comparison women (χ2 11.61, p < 0.001). Among those reporting parental maltreatment, the insecurely attached were at approaching double the risk of such an episode (Table 5a, column 3), although this risk was not associated with the type of insecure attachment. The link, however, was limited to those not only with maltreatment, but who also reported chaotic-type behaviour, which was present for 19 of the 22 insecurely attached women involved (Table 5a, column 1). However, among those without parental maltreatment too, few experienced such a life style for such an analysis to be possible (Table 5b).
Table 5. Chronic depressive episode in 10 years by parental maltreatment, type attachment and chaotic life style
A question of the time order
So far it has been assumed that insecure attachment was usually present by the time a woman left home, although questions about it were largely restricted to the period around the time of our research contact. It is, however, possible to explore the likelihood of an early dating by taking account of two risk factors that were dated – parental maltreatment and the presence at some point of a ‘very poor’ partnership, dealing with the 175 women with at least one partner. Using an analytic procedure meeting criterion for establishing the mediating role of a variable (Baron and Kenny, Reference Baron and Kenny1986), parental maltreatment no longer correlates with such a partnership once insecure attachment is controlled, supporting the early dating of the maltreatment (Fig. 3a) with this consistent with maltreatment's link with insecure attachment remaining unchanged when the quality of partnership is assumed to link the two (Fig. 3b).
Fig. 3. Effect of changing the time order of insecure attachment of the assumed causal model.
Discussion
Previous research had established that insecure attachment, parental maltreatment and depression are highly associated (Mikulincer and Shaver, Reference Mikulincer and Shaver2016, Table 13.3) with this ensuring that, given somewhat more than half the women in the study were selected as likely to have experienced parental maltreatment, there would be a sufficient number of insecurely attached women to explore its link with depression.
Our initial analysis by correcting for the substantial link of parental maltreatment with insecure attachment estimated that a third of depressive episodes were explained by such attachment with this also holding for the general population sample of 80 women when considered alone (Table 1 and Appendix 1).
As in earlier studies, types of insecure attachment (Mickelson et al., Reference Mickelson, Kessler and Shaver1997; Bifulco and Thomas, Reference Bifulco and Thomas2013) were unrelated to the risk experience of depression. However, such types did differ in the psychosocial risk factors involved. For the dismissive/enmeshed as much as two-thirds of depression in a 10-year period was provoked by a humiliating event compared with around a quarter holding for the fearful/withdrawn and maltreated securely attached (Table 3). This for the dismissive/enmeshed was linked with a woman's persistent anger as an adult with as many as 80% (21/26) compared with just 13% (22/172) among all other women (Table 2). This difference was underlined by a link of such adult anger among the dismissive/enmeshed with a mother's earlier physical abuse to the woman as a child, which did not hold for other women (Table 2). An additional more exploratory analysis suggested that there was also a greater general experience on their part of such humiliation irrespective of that linked with depression with the possibility that their heightened experience of anger as an adult played some part in bringing this about.
In contrast, the depression of the fearful/withdrawn was highly associated with the well-established risk factors of low self-esteem and ‘very poor’ partnership, which given the presence of a threatening life event was sufficient to explain their increased experience of depression (Table 4). Here there was also evidence that their more frequent experience of adolescent shyness was a common early precursor of this risk with it linked to their later greater experience of low self-esteem.
The findings concerning attachment are consistent with the definition of the insecure dismissive in terms of ‘emotional barriers to becoming close, low need for company and frequent anger’, and that of the insecure enmeshed in terms of the common presence of ‘ambivalence about core relationships involving resentment of any dependence’ if their ‘need for closeness from a core tie’ is unmet (Bifulco and Thomas, Reference Bifulco and Thomas2013; see also Feeney, Reference Feeney2007).
Finally, a chaotic life style, involving behaviour such as persistent drug abuse usually present at least by early adulthood was closely associated with a chronic depressive episode lasting at least 12 months. However, such highly disturbed behaviour was uncommon in population terms and almost entirely restricted to the insecurely attached. But despite this the link with chronic depression is obviously relevant in terms of public health terms. Here its somewhat greater frequency among the dismissive/enmeshed, despite not reaching statistical significance, may be of relevance given that it was highly associated with their experience of a mother's physical abuse.
Our findings concerning the types of insecure attachment are tentative in the sense of involving a combination of risk factors identified for the first time (Bifulco et al., Reference Bifulco, Moran, Ball and Bernazzani2002). For example, an earlier publication using half the present sister pairs reported a link of the insecure dismissive with severe abuse without taking account of the parent involved, nor physical from sexual or psychological abuse, and with the role of adult anger and humiliating events unexamined (Bifulco and Thomas, Reference Bifulco and Thomas2013, p. 120). It should be added that it is possible that our analysis, largely using sister pairs, has placed too much weight on childhood maltreatment and underplayed broader family-wide effects, although given the extent sister pairs differed in the type of insecure attachment this would seem perhaps unlikely. Taking pairs where at least one sister was insecurely attached, 29 involved a secure and an insecure sister, 14 had different types of insecure attachment, with only eight pairs similar in such terms, with all others both securely attached.
Conclusions
Finally, there are potential clinical implications: knowledge that insecure attachment types have different links with depression may help clinicians focus attention on relevant automatic behaviour of their patients – on the one hand on anger among the dismissive/enmeshed and its potential to produce humiliating situations among those whose childhood exposure to physical abuse from a mother, the very person supposed to protect them, has contributed to this anger; and on the other hand, on the less angry self-paralysis of low self-esteem, which can hold back the fearful/withdrawn from exploring other possibilities and moving on to a new way of life and perhaps lesser experience of depression.
Acknowledgements
Funding was from the Medical Research Council. The authors would like to acknowledge the support and advice from Antonia Bifulco, and the help from Ann Brackenridge, Amanda Lillie, Julie Jarvis, Ruth Robinson, Hedy Wax and Anna Westlake. We are indebted to Islington general practitioners and health visitors for permitting the selection of subjects from their registration lists, and, of course, the women themselves who gave up hours of their time to help with the research.
Conflict of interest
None.
Appendix 1. The aetiological contribution of insecure attachment
To assess the aetiological contribution of insecure attachment, two corrections have been made to the data shown in Table 1. The first with the fact that 8% (5/66) of those with neither core risk factor of parental maltreatment or insecure attachment experienced depression in the 3 years with it is therefore necessary to deduct this proportion from the depression reported by either the insecurely attached or parental maltreated but securely attached. This resulted in the 39 onsets of the insecurely attached being reduced to 33.9 (i.e. 2.7 for those only insecurely attached, and 31.2 for those insecurely attached together with maltreatment), and those with only parental maltreatment with 16 onsets reduced to 11.4. The second correction deals with those just made by subtracting the women with an onset among those with only parental maltreatment from the onsets of the insecurely attached giving 22.5 (33.9–11.4).
Taking both corrections into account, 36.9% (22.5/61) of onsets are explained by insecure attachment with much the same holding for onsets among the 80 women of the comparison series [32.3% (4.2/13)].
