It is just as necessary for analysts to study the way a child is really treated by his parents as it is to study the internal representations he has of them, indeed the principal form of our studies should be the interaction of the one with the other, of the internal with the external.
(Bowlby, 1988a)
One of Bowlby's main reasons for recasting psychoanalysis in the language of Attachment Theory was the hope that this would make it more open to empirical testing. This hope has been fully justified. The past half Century has seen an explosion of research into infant and child development, much of which was stimulated by the work of Bowlby and Mary Ainsworth. The aim of this chapter is to show how these findings point to a remarkably consistent story about the emergence of personality, or ‘attachment style’, out of the matrix of interactions between infant and caregivers in the early months and years of life. The issue of how what starts as interaction becomes internalised as personality is a key question for developmental psychology. ObjectRelations Theory rests on the assumption that early relationships are a formative influence on character. My hope is to demonstrate how Bowlby's movement away from psychoanalysis has come full circle and produced ideas that are highly relevant to, and provide strong support and enrichment for the psychoanalytic perspective.
As a scientific discipline, Attachment Theory has two great advantages over psychoanalysis. First, it rests on direct observation of parent-child interaction, rather than on retrospective reconstructions of what may or may not have gone on in a person's past. Second, it starts from empirical measures of normal development which can then be used as to understand psychopathology, rather than building developmental models from inferences made in the consulting room. It is perhaps no accident that the psychoanalyst with whom Bowlby has most in common, Winnicott, was also keenly aware of normal developmental processes through his earlier work as a paediatrician. Winnicott and Bowlby both believed that their observations of normal development were relevant to psychotherapy, and that by getting a picture of what makes a good parent, we are likely to be in a better position to know what makes a good psychotherapist.
There is an intimate relationship between technology and scientific advance. Galileo's discovery of the heliocentric solar system depended on the expertise of the sixteenthcentury Italian lens grinders; Darwin's Beagle voyage would have been impossible without the navigational and cartographic skills of Victorian maritime imperialism. Science depends, ultimately, on measurement. Ainsworth's ‘Strange Situation’, ‘a miniature drama in eight parts’ (Bretherton, 1991a) with a cast of parent, oneyearold infant and experimenter, enables the vicissitudes of parentinfant relationships to be recorded, rated, and researched. Without Ainsworth's invention, it is unlikely that Attachment Theory would occupy the central position in developmental psychology it does today, remaining in the wings as yet one more speculative psychodynamic theory.
Ainsworth devised the Strange Situation in the late 1960s as part of her studies of mother-child interaction in the first year of life. She had worked with Bowlby in the 1950s, then moved with her husband to Uganda where she had made naturalistic studies of mothers and their babies, and finally settled in Baltimore, Maryland. Influenced by Attachment Theory, although at first wary of its ethological bias, she was interested in the relationship between attachment and exploratory behaviour in infants, and wanted to devise a standardised assessment procedure for human mothers and their children which would be both naturalistic and could be reliably rated, comparable to the methods used by animal experimenters like Harlow (1958) and Hinde (1982b).
The Strange Situation (Ainsworth et al., 1971; 1978) consists of a twentyminute session in which a mother and her oneyearold child are first introduced into a playroom with a ‘stranger’, a friendly participant-experimenter. The mother is then asked to leave the room for three minutes and to return, leaving the child with the experimenter. After her return and the reunion with the child, both mother and experimenter go out of the room for three minutes, leaving the child on its own. Mother and child are then once more reunited. The whole procedure is videotaped and rated, focusing particularly on the response of the child to separation and reunion. The aim is to elicit individual differences in coping with the stress of separation (three minutes alone in an unfamiliar room can seem like an eternity to a oneyear old). Three, and later four, major patterns of response have been identified. In the early studies these were referred to by the arbitrary letters A, B and C. The letter D, while next in the sequence, is nonarbitrary in that it stands for ‘Disorganised’ Attachment.
1 Secure attachment (‘B’) These infants are usually (but not invariably) distressed by the separation. On reunion with their caregiver they greet their parent, sometimes angrily, receive appropriate comfort, distraction and/or soothing, and then re turn to excited or contented play.
2 Insecure-avoidant (‘A’) children show few overt signs of distress on separation, and tend to ignore their mother on reunion, especially on the second occasion when presumably the stress is greater. They remain watchful of her and inhibited in their play. ‘A’ individuals can also be described as ‘hypoactivating’ (of affect).
3 Insecure-ambivalent (insecure-resistant) (‘C’) individuals are highly distressed by separation and cannot easily be pacified on re union. They seek contact, but then resist by kicking, turning away, squirming or batting away offered toys. They continue to alternate between anger and clinging to the mother, and their exploratory play is inhibited. These can be seen as ‘hyperactivating’.
4 Insecure-disorganised (‘D’). After studying tapes of a small group of children who did not fit into any of the above categories (CC or ‘cannot classify’) Main and Solomon (1990) identified this fourth category. These children show a diverse range of confused and confusing behaviours including ‘freezing’, seeming dissociation, or stereotyped movements such as rocking back and forwards, or curling up in a ball when reunited with their parent.
In Ainsworth's original middleclass nonclinic Baltimore sample the proportions were ‘B’ (secure) 66 per cent, ‘A’ (avoidant) 20 per cent, and ‘C’ (ambivalent) 12 per cent. ‘D’ had not been identified at that stage. Since her original publication, the Strange Situation has become widely accepted as a reliable and valid instrument (Van Ijzendoorn and Kroonenberg, 1988). Cross cultural studies have been important in validating the ‘universality hypothesis’, that is, that the Strange Situation taps into meaningful psychobiological phenomena rather than being a narrow manifestation of Western childrearing procedures (Van Ijzendoorn et al., 2006). There are significant crosscultural variations, in that ‘A’ (avoidant) classifications tend to be commoner in Western Europe and the United States (Grossman et al., 1981), while ‘C’ (ambivalent) is commoner in Israel and Japan (Posada et al., 1995); but intracultural variation between different socioeconomic groups and between disturbed and nondisturbed families is greater than intercultural variance, again supporting the ‘universality hypothesis’.
Numerous research and theoretical questions followed from the establishment of this robust research tool, whose ramifications continue to the present day. How do we understand these different patterns of response? Can they be related to patterns of maternal-infant interaction in the early months of life? Are they stable over time, and if so for how long? Do they predict disturbed behaviour later in childhood? Do they, as psychoanalytic theory might assume, persist into adult life? If maternal handling is relevant to classification pattern, what is the relationship between this and the mother's own experience of being mothered? If so, what are the psychological mechanisms by which attachment patterns are transmitted from one generation to the next? Can these patterns be altered by therapeutic intervention? The remainder of this chapter is devoted to a discussion of these and related questions.
Bowlby and his immediate successors saw personality development primarily in terms of environmental influence: relationships determine relationships. The overwhelming evidence is that differing patterns of attachment result from differing patterns of interaction, rather than reflecting variations in infant temperament (Sroufe, 1985; although for a contrary view see Kagan, 1995). Sroufe (1979) makes an important distinction between ‘emergent patterns of personality organisation’ as revealed in the Strange Situation and temperament. The latter represent quasiphysiological styles of behaviour, while the former reflect much more complex habitual relationship patterns. Thus babies may be sluggish, or active, ‘cuddly’ or noncuddly, slow or fast, but still be classified as secure. Even more telling is the finding that children have different, but characteristic, attachment patterns with their two parents, and may be classified as secure with one and insecure with the other. This argues strongly that attachment patterns are a feature of the parent-child relationship, as yet not ‘internalised’ at one year, although by 18 months patterns have become more stable, with maternal patterns tending to dominate over paternal.
The Strange Situation research was part of a much larger study, in which Ainsworth and her colleagues visited mothers and their infants regularly for periods of observation and rating during the first year of life. She found that attachment status at one year correlated strongly with the maternal relationship in the preceding twelve months (Ainsworth, 1979), and this finding has been replicated in several other centres (Main and Weston, 1982; Grossman et al., 1986; Sroufe, 1979). In summary, these prospective studies show that mothers of secure oneyearolds are responsive to their babies, mothers of insecure-avoidant children tend to be unresponsive and mildly rejecting, and mothers of those who are insecure-ambivalent are inconsistently responsive.
The key to secure attachment is active, reciprocal interaction (Rutter, 1981), and it seems that it is quality of interaction more than quantity that matters – a finding that contradicts Bowlby's earlier view on the causes of maternal deprivation. Passive contact alone does not necessarily promote attachment. Many babies are strongly attached to their fathers even though they spend relatively little time with them, and kibbutzimreared children are more strongly attached to their mothers than to the nurses who feed them and look after them during the day, but often without much active interaction. In the first three months of life, mothers of secure infants respond more promptly when they cry; look, smile at and talk to their babies more; and offer them more affectionate and joyful holding. Mothers of avoidant children tend to interact less, and in a more functional way in the first three months, while mothers of the ambivalently attached tend either to be somewhat intrusive even if the baby appears quite happy, or to ignore their babies’ signals for attention when they clearly need it, and are generally unpredictable in their responsiveness. By the second half of the first year, clear differences can be detected in the babies, and those who will be classified as secure at one year cry less than the insecure group, enjoy body contact more, and appear to demand it less (Bretherton, 1991b).
The factor which mothers of insecurely attached children have in common can be understood in terms of Stern's (1985) concept of maternal attunement. He shows how sensitive mothers interacting with their children modulate their infant's rhythms so that when activity levels fall and the infant appears bored the mother will stimulate them, while with an overstimulated child the mother will hold back a little so as to restore equilibrium. In crossmodal attunement the mother follows the baby's babbling, kicking, bouncing and so on with sounds or movements of her own that match and harmonise with those of the baby, although they may be in a different sensory mode. As he bounces up and down she may go ‘Oooooh … Aaaaah …’ matching the tempo and amplitude of her responses to the baby's movements. Stern argues this helps develop the infant's sense of integrated selfhood. Attunement may be impaired in mothers of insecurely attached infants, leading to ‘derailment’ or mismatching in maternal response (Beebe and Lachmann, 1988): thus mothers of ambivalently attached children can be observed to force themselves on their children when they are playing happily, and ignore them when they are in distress.
Brazelton and Cramer (1991) break down the components of secure mother-child interactive patterns into four aspects: synchrony (temporal attunement), symmetry (matching of actions), contingency (mutual cueing), and entrainment (the capturing of each other's responses into a sequence of mutual activity). On the basis of this, play, and later infant autonomy, begins to emerge. Insecure attachments result from intrusiveness or underresponsiveness. While the Strange Situation measure is specifically designed for 12–18 month infants, the ‘Still Face’ procedure picks up the rupture-repair cycle in much younger children. Fourmonth-yearold infants are for one minute confronted with a mother who has been prohibited from smiling, picking up the baby, or responding to it (Tronick, 1989). ‘Still face’ predicts later classification in the Strange Situation. Secure babies can tolerate these brief periods of ‘rupture’ better than the insecure, and with less tendency to regress and become chaotic in their behaviour. They are more able to ‘repair’ the rupture and resume pleasurable interaction with their parent. The latertobeclassified insecure children show disappointment, gaze aversion and selfsoothing strategies, comparable to those seen in the children of clinically depressed mothers (see Beebe et al., 2012).
As we have seen, expressing anger is a major component of the initial response to threatened separation. Secure babies are justifiably cross with their mothers/fathers for ‘abandoning’ them, are able to show it, and the parents equally are able to accept, tolerate and soothe the child's angry distress (Strathearn et al., 2009). It is this that leads Schore (Schore and Schore, 2008) and others to think of attachment in terms of ‘affect regulation’.
The two common patterns of insecure attachment can be understood in terms of the interplay between the need to maintain attachment at all costs, and the builtin aggressive response to the threat of separation. If the secure child's overt and accepted display of angry emption is a ‘primary’ attachment response to threatened or actual separation, both avoidant and ambivalent children employ what has been described as ‘secondary’ attachment strategies (Mikulincer and Shaver, 2007). The ambivalently attached child shows overt aggression towards the inconsistent mother who, in the Strange Situation, has just left him for two successive periods (albeit only for 3 minutes – but how was the one-year-old to know that?). It is as if he is saying, ‘Don't you dare do that again!’, but clinging on at the same time since he knows from experience that the chances are that she will. The avoidant child shows little overt aggression in the Strange Situation, although such children do show outbursts of unprovoked aggression at home. Avoidance is a way of dampening aggression and so appeasing the mother to whom the child needs desperately to feel close, but whom he fears will rebuff him if he reveals his needs too openly, or shows her how angry he feels about being abandoned (Main and Weston, 1982).
A clinical example of avoidant attachment in adulthood comes from a patient who, although faithfully reliable in her attendance at therapy, found it difficult to enter affectively into sessions. Her narrative consisted mostly of a dull catalogue of the preceding week's events. She was meticulous about timekeeping, and kept a close eye on her watch throughout the sessions, because, she said, she was terrified to overrun by a single second. As a solicitor she knew how annoying it was when clients outstayed their allotted time. The effect of this clock-watching was quite irritating to her therapist, who commented that timekeeping was his responsibility and tried to persuade her to remove her watch for sessions. It then emerged that she was frightened that without her watch she would get ‘lost’ in the session, lose control of her feelings and at that very moment the therapist would announce that it was time to stop. In this imagined narrative she would then get so angry she would ‘disgrace’ herself, the therapist would not tolerate this and would break off the treatment altogether.
She had been a rather ‘good’ if distant child who had spent a lot of time on her own, in contrast to her older sister, renowned for her tantrums and angry outbursts. By keeping her distance in a typically avoidantly-anxious pattern, she had maintained contact with her therapist (and presumably as a child, her parents), while avoiding the imagined danger of threatening to destroy her tenuous attachments with her rage. She also kept some sort of coherence in the face of fear of disintegration. The price she paid for this adaptation was affective distancing, low self-esteem (‘He would not tolerate me if he knew what I was really like’), and the lack of a sense of movement and growth in her life. When eventually she was able to get cross with her therapist – for failing to show appropriate sympathy when her beloved cat died – her vitality and self-esteem began to improve, and she was able to make appropriate and productive demands on her previously rather held-at-a-distance boyfriend.
Ainsworth's early work stimulated a host of studies looking at the long-term outcomes for children classified in the Strange Situation at one year. Sroufe and Waters (1977) undertook the first reliability and validity studies in the mid-1970s, recruiting a low-risk sample followed for a number of years, beginning with the assessment of attachment status at 12 and 18 months. In another early study, Matas, Arend, and Sroufe (1978) examined the link between infant security as measured in the Strange Situation and later persistence, enthusiasm, cooperation, and positive affect in play at three years. Here was the first evidence that the infant's attachment classification had effects on a range of adaptations and competencies beyond infancy.
Next, attention switched to economically disadvantaged mothers and their children (Vaughn et al., 1979). This sample was ultimately to grow to over 200 families, and was followed for over 30 years (Waters et al., 2000; Sroufe, 2005). The findings indicated that with higher levels of instability in the care-giving environment, children are more likely to be insecurely attached, and their attachment organization more likely to shift from secure to insecure in the face of stress and environmental upheaval. This research was critical in pointing out the risks to children living in high stress, disadvantaged, environments of developing insecure or unstable attachments.
These studies showed that at two years, securely attached children have a longer attention span, show more positive affect in free play, show more confidence in using tools, and are more likely to elicit their mother's help in difficult tasks compared with anxiously attached children. Their nursery teachers (blind to attachment status) rated them as more empathic and compliant and higher on positive affect. In peer interaction avoidant children are hostile or distant, while those classified as ambivalent tend to be ‘inept’ and to show chronic low-level dependency on the teacher, and to be less able to engage in free play by themselves or with peers.
In sum, long-term follow-up studies show that the patterns of behaviour defined by the Strange Situation behaviour at one year carry forward into childhood and beyond (Grossman, Grossman and Waters, 2005). Grossman and Grossman (1991) found strong correlation between patterns of behaviour on re-union after separation at one year and six years. They also showed that six-year-olds, classified at one year as secure, played concentratedly and for longer, were more socially skilful in handling conflict with their peers, and had more positive social perceptions, compared with children who had been rated as insecure as infants.
Sroufe (1979) sees secure-rated children as having greater ego control and ego resiliency than the insecure. Secure children were rated by their teachers as balanced in their control, while their avoidant and ambivalent counterparts were overcontrolled and undercontrolled respectively. Resiliency was inferred from such statements as ‘curious and exploring’, ‘self-reliant, confident’. Stroufe concludes:
What began as a competent caregiver-infant pair led to a flexible resourceful child.… Such predictability is not due to the inherently higher IQ of the securely attached infant, or, apparently, to inborn differences in temperament.
Sroufe's perspective can be encapsulated in the aphorism: relationships determine relationships. Attachment research tells us that psychological health is not an inherent property of an individual; it arises out of the early parent–child matrix, and manifests itself, and is reinforced, or undermined, by subsequent and current relationships. The implications of this, both at the level of the individual experience and of the family and wider society have, as we shall see, still to be fully worked out.
If Bowlby and Ainsworth are the founding father and mother of Attachment Theory, its first child, and the person who has moved the field on more than any other researcher since, is unquestionably Mary Main. Main revolutionised the field in two ways. First, through devising the Adult Attachment Interview she extended Ainsworth's research on mother-infant attachment to the study of attachment in adults, and with it moved the field from behaviour to ‘the level of representation’. Second, as described above, she delineated a fourth category of attachment – ‘insecure disorganised’. Both discoveries radically altered the landscape of attachment, and expanded its reach into the domain of clinical theory and practice.
Main's theory of Disorganised Attachment is summarised in her now classic paper (Main and Hesse, 1990) suggesting that the caregivers of infants who are disorganised in relation to attachment are – as a result of their own unresolved loss or trauma – either frightened by, or frightening to their infants. Consequently, such infants face an insoluble paradox, in which the caregiver is both ‘[the] source of and the solution to its alarm’ (1990, p. 163). This viewpoint is supported by subsequent meta-analysis (Schuengel et al., 1999).
The early attachment studies were done on non-clinical, middle-class samples. The discovery of the disorganised/disoriented pattern of infant attachment, and its links to frightened or frightening parenting, opened out a new seam of research into the relationship between environmental risk, especially child maltreatment, and disrupted attachment. A key deficit when thinking about the psychopathology and treatment strategies for maltreated infants is their lack of a coherent strategy for seeking safety (Carlson, et al., 1989; Carlson, 1998). Disorganised Attachment is predicted by environmental antecedents (including the quality of the relationship and history of maltreatment), and predictive of a range of behavioural difficulties and psychopathology through childhood and adolescence. The seeds of these have an impact from the very early stages of a child's life. At 4 months Beebe (Beebe, et al., 2012) found that inter- and intra-personal conflict in the face of distress leaves infants with early indicators of disorganised attachment: the experience of not being ‘known’ or ‘found’ by their mothers.
Moving beyond the Strange Situation to studying older children, Main et al. (1985), Bretherton (1991b) and Cassidy (1988) tapped into the child's experience of attachment (rather than its behavioural manifestations) through the use of play techniques such as picture completion tasks, story completion, and puppet interview and story. All rely on depicting episodes of separation and re-union. Children reveal their attachment histories through their play and imaginative activity in response to these stimuli. Secure children in response to a picture-story task give coherent, elaborated responses, including references to their own separation experiences, and are able to suggest positive ways in which separated figures could resolve their difficulties.
Avoidant children, by contrast, describe separated children as sad, but cannot think of ways to help them. At six years they tend to draw figures with blank faces and no hands, suggesting a lack of responsiveness and holding (physical and metaphorical) in their lives. Secure children at six were more able to give a balanced view of themselves as good, but not perfect, while insecure children saw themselves as either faultless or bad. The disorganised group (‘D’) show bizarre or disorganised responses to these picture- and story-completion tasks.
Main and Cassidy (1988) also found that children classified as disorganised in infancy, when briefly separated from their mothers at age 6, were more likely to respond to their parents upon reunion with ‘controlling/punitive’ or ‘controlling/caretaking’ strategies. Main (1991) looked at ten- and eleven-year-olds, classified in the Strange Situation at one year, who were now asked for a spoken autobiography. There was a 75 per cent correlation between attachment classification at one year and a decade later. Compared with the insecurely attached, secure children's stories were consistently more coherent, had greater access to memories, especially of their pre-school years, and showed more self-awareness and ability to focus in on their own thought processes. Main calls the latter ‘metacognitive monitoring’, the ability to think about thinking, a phenomenon to which we shall return.
Lyons-Ruth (Lyons-Ruth et al., 1999) has been another major contributor to the understanding of Disorganised Attachment, linking infant disorganization to parental psychopathology, social risk, and disrupted mother–infant communication. A recent study revealed a relationship between the development of the symptoms of Borderline Personality Disorder in adolescents with maternal withdrawal measured at age 18 months (Lyons-Ruth et al., 2005; see Holmes, 2010).
Taken as a whole, these findings can be summarised as follows: relationship patterns established in the first year of life continue to have a powerful influence on children's subsequent behaviour, social adjustment, self-concept and autobiographical capacity. These effects last well into adolescence and early adulthood. Mother–infant relationships characterised by secure holding (both physically and emotionally), responsiveness and attunement are associated with children who are themselves secure, can tolerate and overcome the pain of separation, and have the capacity for self-reflection.
These results support the view that the early years of life play a crucial part in character formation, and demonstrate the continuity between the pre-verbal infant self and the social self as we commonly conceive it. A number important qualifications need to be noted. First, since the parent–child relationship operates continuously as development proceeds, what we are seeing is not so much the result of some irreversible early events (as might be conceptualised in a simplistic psychoanalytic model), as an ongoing relationship with its own ‘epigenetic’ stability. Attachment patterns are not immutable and can move in positive or less favourable directions depending on the impact of traumatic or favourable environmental events. Thus, second, if a mother's circumstances change – for example, a previously single parent now establishes a stable relationship with a new partner – her child's attachment status may change, in this case from insecure to secure. Similar changes can occur, as we shall see later in the chapter, when mother and infant are both treated with psychotherapy (Murray et al., 2011).
Third, the emphasis in the discussion so far has been on the contribution of the parent, especially the mother, with the infant's role being relatively passive. Clearly this is an oversimplification; temperamental, and neurobiological factors in the child will play their part in the relationship with the parent, and subsequent social adjustment. The ‘differential susceptibility’ hypothesis suggests that children's genetic makeup strongly affects the extent to which they are susceptible to environmental influence (Belsky et al., 2007; Bakermans-Krankenburg and Van Ijzendoorn, 2011). Some children do very well in sensitive and rich environments and very badly in bleak insensitive ones; others seem relatively impervious. Nevertheless, the evidence suggests that the parent is the determining factor and that a ‘good’ mother even with a ‘difficult’ baby will, by one year, be likely to have a securely rather than insecurely attached child. Thus the amount a child cries at one year seems to depend more on the mother than the child. There is a strong correlation between prompt and sensitive maternal responsiveness to infant crying in the first three months of life and reduced crying at one year, whereas there is no correlation between the extent of infant crying in itself in the first three months, and the amount of crying at one year.
Fourth, attachment status is a categorical (that is, A, B, C, or D), rather than a dimensional (that is, on a continuum) classification, yet clearly there will be a spectrum of subtle characterological features within it. Indeed some argue that severity of insecurity is a more important feature for predicting psychopathology, than the specific insecure category assigned (Maunder and Hunter, 2012).
Finally there is the question of intergenerational transmission of attachment. If relationships are in some way internalised by the growing child as ‘character’, what happens when that child grows up and him- or herself becomes a parent? We know from Harlow (1958) and Suomi's (2008) that infant monkeys separated from their mother, when they become sexually mature, show gross abnormalities in mating and parenting behaviour. In the infinitely more complex language- and experience-based world of the human primate, are there connections between a mother's capacity to provide secure attachment for her child, and her experiences with her own mother when she was a child? A crucial tool for exploring this issue has been the Adult Attachment Interview.
Main's second great contribution to the field has been The Adult Attachment Interview (AAI). This was initially developed as a specific tool for assessing the working models or inner world of parents with respect to attachment, using a classification paralleling that of the Strange Situation. Later its use has been extended to assess attachment ‘states of mind’ in adolescents and adults.
The AAI is a semi-structured interview conducted along the lines of a psychotherapy assessment aiming to ‘surprise the unconscious’ into revealing its inner contours (Main, 1991). Subjects are asked to choose five adjectives which best describe the relationship with each parent during childhood, and to illustrate these with specific memories; to describe what they did when upset in childhood; to which parent they felt closer and why; whether they ever felt rejected or threatened by their parents; why they think their parents behaved as they did (this question is especially important in relation to the capacity to ‘mentalise’); how their relationship with their parents has changed over time; and how the imagine that early experiences may have affected their present functioning.
The interviews are audio-taped and then rated along eight scales: loving relationship with mother; loving relationship with father; role reversal with parents; quality of recall; anger with parents; idealisation of relationships; derogation of relationships; and coherence of narrative. A crucial aspect of AAI rating is that categories are assigned not so much on content as the process of the interview. It is the narrative style of the interviewee that is key, not the specific details of their experience, whether favourable or traumatic. The ‘state of mind with respect to attachment’ of the interviewees is then reliably assigned to one of four categories, each of which is comparable with one in the Strange Situation: Autonomous–secure, Dismissing–detached, Preoccupied– entangled and Unresolved–disorganised.
Autonomous–secure parents give accounts of secure childhoods, described in an open, coherent and internally consistent way. Attachments are valued, and even where their experiences have been negative, there is a sense of pain felt and overcome. The Dismissing–detached group give brief, incomplete accounts, professing to having few childhood memories and tending to idealise the past with such remarks as ‘I had a perfect childhood’, ‘My Mum? Just a normal Mum’. Preoccupied– entangled parents give inconsistent, rambling accounts in which they appear to be over-involved with past conflicts and difficulties with which they are still struggling. The Unresolved–disorganised category is identified through logical gaps and lacunae in the narrative flow, is rated separately and refers specifically to traumatic events such as child abuse or neglect which have not been resolved emotionally.
Main saw these differences as reflecting underlying differences in internal working models (IWMs) of attachment, that is, representational models built up as the result of actual experience. This delineation of adult categories of attachment allowed Main and Goldwyn (1984) to examine the relationship between parents’ attachment classification and those of their children. Unsurprisingly, they found high rates of correspondence: infants judged secure at one year were more likely to have mothers who were secure in their internal working models of attachment; those infants judged avoidant at one year were more likely to have mothers who dismissed the impact of early attachment experiences; those judged resistant were more likely to have mothers whose state of mind in relation to attachment was preoccupied.
Even more striking were the findings of Fonagy and his co-workers (Fonagy et al., 1991a), who administered the AAI to prospective parents during pregnancy and found that the results predicted infant attachment status in the Strange Situation at one year with 70 per cent accuracy. Of insecure infants 73 per cent had insecure mothers, and only 20 per cent of secure infants had insecure mothers, while 80 per cent of secure mothers had secure infants. Fathers’ influence appeared less striking: 82 per cent of secure fathers had secure infants, but 50 per cent of insecure fathers still had secure infants. This supports the view that attachment status is a function of the infant–parent relationship, rather than of temperament, and also suggests that, at least in Western cultural conditions, maternal, rather than paternal, insecurity is the more potent transmitter of insecure attachment across the generations.
The AAI enabled these researchers to confirm the hypothesis, implicit in Attachment Theory from the start, that there was intergenerational transmission of attachment. But how does this come about? Van Ijzendoorn's (1995) meta-analysis found that only 12 per cent of the variance in a child's attachment is accounted for by parental ‘sensitivity’, a lacuna they dub the ‘transmission gap’, leaving open the question of what other mechanisms might explain patterns of intergenerational transmission. Enter ‘mentalising’ or ‘mentalisation’ (the terms are interchangeable).
The emergence of mentalisation theory marked, like Main's ‘move to the level of representation’, another quantum leap in attachment research. While studying the AAIs that they had collected as part of their longitudinal study (Fonagy et al., 1991), Fonagy and colleagues noted that adults who were secure in relation to attachment seemed more able to appreciate and reason about mental states relating to their early childhood experiences and relationships. In this initial research Fonagy et al. (1995) discovered that the capacity to envision mental states (thoughts, feelings, and intentions) in the self or the other, and to understand behaviour in light of mental states, was more highly correlated with adult attachment classification than was narrative coherence by itself. In addition, this meta-representational process, in which an individual reflects upon his or another's psychic experience, was highly predictive of infant attachment. Initially they dubbed this ‘reflexive function’, a term now incorporated into the notion of mentalising.
Fonagy and Target (1997) argued that a secure attachment relationship provides a context for the emergence of a reflective self and a theory of mind. Mentalising refers to the ability to see others as having thoughts, feelings, desires and projects of their own. In addition, it encompasses the perception that one's perspective on the world, and especially the interpersonal world of others, is filtered through the mind, and is inevitably subject to error and the need for constant updating and correction. Holmes’ (2010) portmanteau definition of mentalising is ‘the ability to see oneself from the outside and others from the inside’.
Fonagy and his colleagues’ research suggested that parents’ capacity to make sense of their children's minds is a crucial element of maternal or paternal sensitivity. Thus mentalising provides a possible mechanism for the intergenerational transmission of attachment. This led several research groups to study the link between parental conceptions of mind and a range of attachment outcomes. Slade and her colleagues coded parental representations of the child for parental reflective functioning (Slade et al., 2005) using the Parent Development Interview (PDI). Parents high in reflective functioning were more likely to have secure children and be secure in relation to attachment themselves; in addition, parental reflective functioning was found to mediate the relationship between adult and child attachment. The same group (Grienenberger et al., 2005) linked low parental reflective functioning with disrupted affective communication and disorganised infant attachment.
Meins (Meins et al., 2001) coined the rather attractive phrase ‘mind-mindedness’ in preference to the latinate ‘mentalising’. Whereas Fonagy and Slade had coded mentalisation using semi-structured narrative interview methods, Meins coded the spontaneous language mothers used to describe their children's experiences, and found that to the extent which mothers used words to convey mental experience, reflecting their ability to recognise that the child has a mind, the more likely their children were to be secure. Koren-Karie et al. (2002) similarly found that mothers’ ‘insightfulness’ about their children's behaviour correlated with maternal sensitivity and infant attachment, and mediated the link between the two.
Mentalisation, reflective functioning, mind-mindedness, and insight-fulness all refer to a specific capacity on the part of parents to respond to their children not just with passive understanding, but appropriately. Ainsworth's initial formulations had from the start differentiated between mothers’ capacity to respond to infant signals, and the ‘appropriateness’ of that response, but this distinction was not incorporated into the global sensitivity scales used by the early researchers. ‘Appropriateness’ assumes that a caregiver needs two-stage skills: first the capacity to ‘read’ her infant – that is, to mentalise; second, to gauge her responses in the light of those ‘readings’. By developing accurate ways of measuring this ability, Fonagy, Slade, Meins, and Oppenheim and Koren-Karie have significantly narrowed the elusive transmission gap.
Since their early finding that reflexive function protects adults who have themselves had traumatic childhoods from inflicting insecurity on their children, Fonagy and his team have further developed the notion of mentalising. They suggest that children who have been ‘mentalised’ by their caregivers will themselves be able in time to read the minds of others: indeed secure children outperform their insecure counterparts in ‘theory of mind’ tasks aged three. This ability, they suggest, arises out of a) joint pretend play with adults, b) intense mutually enjoyable conversations also with care-givers, and c) peer-group interaction, especially with siblings (Allen et al., 2010).
Mentalising capacities are intimately linked with self-agency (Knox, 2011), so crucial to psychological well-being. The consequences of impaired mentalising can be seen in its starkest form in child maltreatment. Here the child is exposed to the ‘double whammy’ of a maltreating caregiver, by definition unable or unwilling to mentalise the impact of the neglect and/or abuse they inflict, and unable to foster the very mentalis-ing capacity that would help the child make sense of their maltreatment.
Thus far our discussion has confined itself mainly to childhood attachment relationships, and their long-term consequences, with the bulk of the research carried out by developmental psychologists. A new chapter in the attachment paradigm opened up in 1987 when two social and personality psychologists, Hazan and Shaver (1987), published their landmark paper on adult ‘romantic attachments’. Following Bowlby's suggestion that love relationships in adults are shaped by the attachment dynamic, they decided to see whether the Ainsworth categories of secure and insecure attachment applied to adult romantic attachments. Using a self-report methodology, they found that individuals’ experiences in romantic relationships followed the secure/avoidant/anxious–ambivalent typology described by Ainsworth. The distribution of the three types of romantic attachment in a non-clinical sample of adults corresponded closely with those found in children (56 per cent secure; 24 per cent avoidant; 20 per cent anxious–ambivalent).
That they asked this particular question was itself revolutionary; equally significant, however, was their use of a descriptive quick-and-easy self-report method for measuring attachment styles. In their initial study, as well as in a later study on love and work (Hazan and Shaver, 1990), they asked subjects to choose among three prototypical descriptions of the way they felt in relationships: comfortable with closeness (secure), preferring distance (avoidant) and wishing for greater closeness (ambivalent). Subjects’ choices were highly correlated with their experience of romantic love, their attitude to work and ambition, as well as mental models of self and others and of childhood relationships.
Hazan and Shaver's original three-category attachment styles questionnaire was modified in a crucial way by Kim Bartholomew (1990), who proposed a fourth attachment style, and developed the Relationship Questionnaire. She suggested that – instead of being described along a single dimension (secure/insecure) – attachment dispositions be described in terms of two orthogonal dimensions, anxiety and avoidance. Secure individuals, she proposed, are low on both; the ambivalently attached/preoccupied are high on anxiety but still cling to their attachment figure in a way that is, as it were, not distant enough. The innovative aspect of her approach entailed dividing avoidant people into two types depending on the anxiety dimension. Those with high avoidance and high anxiety she described as fearful, that is, people who long for, but fear closeness. In the dismissive avoidant pattern, where avoidance is high but anxiety low, defensiveness operates more successfully, with individuals in this group less psychologically troubled than their fearfully avoidant counterparts.
The original attachment style prototype measures were later replaced by a thirty-six-item questionnaire that yields continuous scores along both attachment and avoidance dimensions (Experience in Close Relationships Questionnaire (ECR) Brennan et al., 1998). These methods, widely used today (Mikulincer and Shaver, 2007), can be used to assess attachment across a range of adult relationships.
These refecting, self-report methods are distinct from the more cumbersome, but possibly more meaningful, activating techniques of the Strange Siutation and the AAI (see Maunder and Hunter, 2012). Researchers have found correlations between the AAI and ECR to be relatively weak, suggesting that there may be important differences in what each is measuring. As mentioned, the question of whether attachment security/insecurity varies continuously or discontinuously in the population, and therefore whether categorical or dimensional classifications are most appropriate remains an area of active debate (Fraley and Spieker, 2003).
While romantic attachment researchers have focused less on disorganised/unresolved attachment than their developmentalist counterparts, it is possible that Bartholomew's ‘fearful avoidants’ overlap with this group. Another angle is the view that severity is a more important dimension than the specific attachment subtype assigned, and that Bartholomew's fearful group, who combine fear and avoidance, are more severely insecure and therefore more prone to psychopathology (Maunder and Hunter, 2012).
Staying within the romantic attachment framework and methodology, Mikulincer and Shaver (2007) have aligned attachment with psychodynamic perspectives, while pioneering the use of standard psychological science methods such as measuring reaction times. For example, they have explored a fundamental aspect of attachment – the care-giving behavioural system – showing how compassion and empathy can be contextually enhanced even in the insecurely attached, by subliminal priming with thoughts, words, or names associated with attachment (Mikulincer et al., 2005).
Another important feature of Mikulincer and Shaver's contribution is their development, (see Marris, 1996; Kraemer and Roberts, 1996; see also Chapter 10) of the sociological implications of the attachment paradigm. In a moving epilogue to their monograph (Mikulincer and Shaver, 2007) they show how social cohesion is promoted by secure attachment, and how more caring and coherent societies foster families’ capacity for altruism.
Crowell, like Hazan and Shaver, examined attachment in adult romantic relationships (Crowell et al., 2002). Rather than rely on self-report measures, she and her colleagues developed the Current Relationship Interview (CRI) that assesses representations of attachment, classifying the subject as secure, dismissing, or preoccupied in relation to their current partner. Those classified as secure on the CRI reported greater levels of satisfaction and commitment overall; when both partners were insecure, a number of markers of aggression were elevated and partners reported lower levels of marital satisfaction and closeness. Crowell found – particularly for secure individuals – that attachment classification remained stable over the course of the transition to marriage. Insecure adults can shift to a secure status in the early months of marriage, but many remain dissatisfied and struggle with feelings of abandonment and anger. Those who had managed to separate themselves to some extent from their families of origin were more likely to move from insecure to secure status; the dismissive group was more resistant to change than the ambivalently attached. The Unresolved group was the least stable.
Crittenden's (2006) Dynamic Maturational Model reframes attachment classifications as self-protective (that is, defensive) strategies learned in interaction with caregivers, emerging against the backdrop of maturation and individual biological differences. She also proposed a ‘circumplex’ model, with a range of subtypes that have implications for understanding and treating developmental psychopathology.
Other important contributions in the adult romantic attachment paradigm include Waters and Waters’ (2006) notion of ‘secure base scripts’ as narrative manifestations of attachment, and Feeney's account of how a partner's capacity to provide a secure base increases exploration. Feeney and Van Vleet (2010) studied the ways in which attachment patterns affected spouses’ ability to support one another when given laboratory-based tasks. They delineate the ‘dependency paradox’ – that is, the more supported a person feels, the less dependent and the more able s/he is to explore. Their research identifies the three components of security-promoting relationships – availability, non-intrusive support, and encouragement, a simple yet profound formula for attachment's role in psychological health.
Some of the findings of post-Bowlbian research are summarised in Table 5.1 which attempts to depict a picture of benign and vicious cycles of security and insecurity. Secure mothers are responsive and attuned to their babies and provide them with a secure base for exploration. They are able to hold them, delight in them, and cope with their discontent and aggression in a satisfactory way. These mothers have a balanced view of their own childhoods which, even if unhappy, are appraised accurately. Their children, secure as infants, grow up to be well-adjusted socially and to have a realistic self-appraisal and a sense that separation, although often sad and painful, can be coped with. Secure mothers and secure children have a well-developed capacity for self-reflection and narrative ability, and convey a sense of coherence in their lives.
Insecure children, by contrast, especially if avoidant, tend to have had mothers who found holding and physical contact difficult, who were unresponsive to their infant's needs and not well attuned to their rhythms. These mothers tended to be dismissing about their relationships with their parents and to be unable to tell a vivid or elaborated story of their own childhoods. As they grow up, avoidant children tend to be socially isolated, to show unprovoked outbursts of anger, to lack self-awareness and to be unable to give a coherent or textured accounts of themselves and their lives.
Are we seeing in these insecure children the roots of adult personality difficulty and disturbance? If so, what can we learn about the psychological mechanisms that may underlie these disorders, and do they provide clues as to how psychotherapy might help to reverse them? We shall discuss these issues in the two subsequent chapters. Our concern here is to try to conceptualise how maternal handling becomes internalised as infant psychology. In Fraiberg's (Fraiberg et al., 1975) famous metaphor: ‘In every nursery there are ghosts. These are the visitors from the unremembered pasts of the parents; the uninvited guests at the christening.’ How do parental ghosts get incorporated into the internal working models of the infant? Three inter-related ideas, all of which we have already touched on, can be used to clarify this: avoidance of painful affect (Grossman and Grossman, 1991), consistency and coherence of internal working models (Bretherton, 1991a), and mentalising (Fonagy, 1991; Fonagy et al., 1991a, 1991b).
Bowlby viewed the capacity to experience and ‘process’ negative affect – to feel and resolve the pain of separation and loss – as central to psychological health. This intuition has been confirmed in fMRI studies showing that parents of securely attached infants accept and respond to their infants' distress and happiness equally, while parents of insecure children tend to activate brain areas associated with disgust, and to have reduced oxytocin (the ‘bonding’ hormone) levels when faced with their children's misery (Strathearn et al., 2009). The latter group either ignore their infants' distress (avoidant pattern) or become over-involved, panicky and bogged down in it (ambivalent pattern).
We can speculate that because these parents have not been able to deal with or ‘metabolise’ (Bion, 1978) their own distress, they find it difficult to cope with pain and anger in their infants and so the cycle is perpetuated. The infant is faced with parents who therefore cannot ‘hold’ (Winnicott 1971) their children's negative feelings of distress or fears of disintegration. The child consequently resorts to primitive defence mechanisms in order to keep affects within manageable limits. Aggressive feelings may be repressed or split off, as in the avoidant child who does not react to his mother's absence but then shows overt aggression towards toys or siblings. The insecure–ambivalent child by contrast may show over-compliance or a ‘false self’ (Winnicott, 1971), or ‘identification with the aggressor’ (A. Freud, 1936).
Major parent–child mis-attunements characterise Disorganised Attachment and include such phenomena as gaze aversion, self-injury (such as head banging), freezing, dissociation, or unprovoked rage. Disorganised Attachment, as we have seen, is associated with child maltreatment, severe socio-economic stress, child psychopathology, and may presage Borderline Personality Disorder in adolescence and adulthood, manifesting with self-injurious behaviour such as wrist cutting, overdosing and substance abuse.
One can imagine continuities between infant physiological experience and psychological structures which evolve through childhood into adult life (see Figure 5.1). From maternal consistency comes a sense of history: the reliability of the mother's response to the infant becomes the nucleus of autobiographical competence. From maternal holding comes the ability to hold one's self in one's own mind: the capacity for self-reflection/mentalising, to conceive of oneself and others as having minds. The insensitivity and unresponsiveness of the mother of the insecure infant is not necessarily mean or abusive, but rather a failure to see the world from the baby's point of view, to ‘take the baby's perspective’ (Bretherton, 1991a). The mother who cannot act as an ‘auxiliary ego’ for her child exposes him or her to inchoate and potentially overwhelming feelings when that child is faced, as will be inevitable (and in the case of an insecure mother to an excessive degree), with loss and separation anxiety. We are moving here beyond attachment as protection from predation (Bowlby's original idea), to a view in which the key role of attachment relationships is affect regulation. In this model, the parent's role is to help the child ‘co-regulate’ emotions, and the process by which that ability gradually becomes internalised (Schore and Schore, 2008) is self-regulation.
Fonagy (Fonagy et al., 1991b) sees coherence as a central feature of parents classified as Secure-autonomous on the AAI:
The coherence of the parent's perception of his past derives from his unhindered capacity to observe his own mental functioning…. This coherence is the precondition for the caregiver to … provide an ‘expectable’ or ‘good enough’ environment for the infant…. A child may be said to be secure in relation to a caregiver to the extent that his or her mental state will be appropriately reflected on and responded to accurately.
Without some sense of coherence and benignity towards oneself, self-reflection becomes distorted or even impossible, as the following case of Borderline Personality Disorder illustrates.
Anna was a single parent in her mid-twenties. She became depressed and suicidal when her child was 6 months old. Her mother had been in hospital a lot when Anna was a small child because of TB, and at one point (when Anna was ten) had left the home for a while to live with another man. Despite this neglect Anna saw her mother as ‘perfect’, someone whose standards she could never match, and herself as hateful. She had two distinct ‘selves’: one competent, intelligent, well-organised, cheerful, compliant, pretty; the other dark, despairing, longing to die. In hospital at times she avoided eye-contact, secreted razor blades and frequently cut herself, was morose and monosyllabic, and would occasionally have outbursts of rage. At other times she was a model patient and collaborated enthusiastically with her therapeutic programme. She was discharged from hospital and started weekly analytic therapy, but once more became suicidal and was readmitted. She complained that she found psychotherapy very difficult because it meant that she had to think about herself. That entailed getting in touch with her self-hatred:
‘Whenever I look into myself I come across the feeling that I want more than anything to die. I am forced to stay alive because of my daughter. Coming to talk to you reminds me of all the things I don't want to think about.’
Her wish to harm herself arose whenever she was faced with painful feelings of separation; for example, when she was on her own in the evenings. The origins of an almost unbridgeable split between her compliant and defiant selves, and the difficulty in reaching her real pain and hope, could be seen in terms of the feeling that she had never felt securely held by her mother as a child, and was reminded of the anxiety and pain of this whenever she was on her own. She needed to be held by the hospital – sometimes literally – before she could begin to think about ‘holding herself in mind’ in therapy, and to feel a measure of security about her capacity to be a mother to her daughter. As suggested by Figure 5.2, her mental state could be represented by a series of parallel concentric circles of ‘holder’ and ‘held’, container and contained.
Borderline patients like this provide adult examples of insecure infants, whose care-givers have been unable to reflect on and so metabolise their infants' feelings of pain on separation. They are then forced to survive as best they can, using splitting, isolation and self-harm as ways of coping and self-soothing. When they become parents they often hope to break the cycle of insecurity which has jeopardised their own development, through vicariously providing their own child with security. Tragically however this often merely perpetuates the cycle of insecurity. Mental health services and therapists typically counter-transferentially re-enact a repetition of the unresponsiveness and breaks in care that the patient experienced as a child. Anna became passionately involved with one male nurse – paralleling her long-repressed desire to have an exclusive relationship with her mother – and did very well, until he was transferred to another ward.
Bowlby depicted healthy internal working models as subject to constant revision and change in the light of experience. Anna exemplified how in pathological mental states there is often a sense of repetitiousness and ‘stuckness’ in therapy. Bretherton (1987) has speculated about why it might be that internal working models in insecure attachment are particularly resistant to change. She sees mental structures as organised hierarchically from low-level ‘event-scripts’ (Waters and Waters, 2006), such as ‘when I hurt myself my mother comes to comfort me’, through intermediate generalisations like ‘my mother is usually there when I need her’, to basic assumptions: ‘my mother is a loving person. I am lovable and loved.’ Insecure individuals not only have negative core assumptions, but, because communication between different levels of the hierarchy is distorted and restricted, are unable to revise these models in the light of experience. Anna's basic assumption – ‘I am hateful’ – remained impervious to contrary evidence provided by the love of her boyfriend and the care of her therapists:
What seems to differentiate the Internal Working Models of secure and insecure individuals is in part their content, but also their internal organisation and relative consistency within and across hierarchical levels…. Reconstruction of working models cannot be achieved [simply] by ‘lifting repression’ or removing barriers which allow well-encoded, but hitherto inaccessible information to come into conscious awareness. Something much more akin to complete reorganisation and reinterpretation may be necessary.
(Bretherton, 1991a)
The therapeutic challenge is that for such re-organisation to occur, patients, and their therapists with them, need the support and time to tolerate and survive periods of extreme vulnerability. Only then can new, trusting, and more flexible and adaptive internal working models emerge.
Bowlby's grafting of the experimental methods of ethology to the insight of psychoanalysis bore rich fruit. The research we have surveyed in this chapter has revealed some of the relational elements underlying psychological health: a sense of security, of efficacy; of being loved and having the capacity to love; of being a person in the world like others and yet with one's own unique biographical trajectory; of being able to withstand the failures, losses and disappointments that are the inevitable consequence of the ‘thousand natural shocks that flesh is heir to’. Attachment research reveals some of the ingredients that make up good parenting: appropriate responsiveness, attunement, holding, mentalising. We turn now to the implications of these findings for psychotherapy, and to further exploration of the relationship – at times avoidant, at times ambivalent – between Attachment Theory and psychoanalysis, in the hope of finding a more secure and coherent synthesis.
