Mary Main was one of Ainsworth’s first graduate students at the Johns Hopkins University. However, her distinctive training prior to graduate work meant that she came to the study of child development from an unusual angle. As an undergraduate she studied classics and natural sciences at St John’s College, Maryland. This included four years of courses in literature and philosophy. Since Main had married one of her philosophy professors, Alvin Main, if she wanted to continue to graduate school she would need to find a programme nearby. Following her lifelong interest in language, Main applied to Hopkins to study psycholinguistics.1 However, her application was accepted by Ainsworth for graduate work on child–caregiver attachment. Main completed her doctorate under Ainsworth in 1973. Her project was the third study to use the Strange Situation, after Ainsworth and Sylvia Bell, and the results were part of the 106 cases analysed in Patterns of Attachment. In 1973 she was appointed to a faculty position at the University of California, Berkeley. Settling in at Berkeley was disrupted by the death of Alvin Main. However, during the 1970s she managed to begin a longitudinal study in the Bay Area, conducting the Strange Situation with both mothers and fathers. Already in her dissertation, Main’s interest in the work of Robert Hinde had led her to study conflict behaviours observed in the Strange Situation. She pursued this inquiry in her Bay Area sample, which led to the introduction of the disorganised attachment classification. A subset of families from the sample were invited back to the laboratory when the child was aged six. Main and her group inductively identified associations between the infant Strange Situation classifications and assessments of the six-year-old’s behaviour and family drawings. The Strange Situations also had associations with the form taken by a parent’s autobiographical narrative. This latter finding led to the development of the Adult Attachment Interview (AAI).2 The Berkeley study has been described as ‘the most influential, in-depth, and complex study of intergenerational factors in attachment’ of its era.3 Together with her second husband and collaborator Erik Hesse, Main introduced the hypothesis that disorganised attachment may be caused by caregivers’ behaviour that alarms their infant, and that this alarming behaviour can be predisposed by unresolved experiences of loss or trauma. Building on the work of Bowlby and Ainsworth, contributions by the Berkeley group revolutionised and redefined the methodology and theory of attachment research.
Main was the first of Ainsworth’s graduate students to receive a faculty position and establish an independent research group. As a result, the most pressing item of business was replication. Whereas Ainsworth’s original study in the mid-1960s had 26 infant–mother dyads, and Patterns of Attachment could report 106 infant–mother dyads, Main submitted a proposal in the mid-1970s to the William T. Grant Foundation for a much larger study, including both mothers and fathers. The proposal was accepted and Main was able to hire research assistance to support the recruitment of 189 families in 1976–77. For this replication and extension of Ainsworth’s study, Main’s sample was selected precisely to be low risk. All the families were middle- or upper-middle class. No teenage mothers were included, and none of the parental couples were divorced. Families were screened out if the infant had been born prematurely or had a low birth weight, or if there had been problems with the delivery. Families were also screened out if infants had experienced a separation of two weeks or more from their attachment figures. And only Caucasian or Asian families were included, to exclude as a confound the adversities black families faced with stigma and oppression. Families were also screened out if the father was unemployed, or if infants were in more than 25 hours of daycare per week, though parents were not excluded on the grounds of mental illness.
Deliberately unrepresentative of US families, Main’s sample was intended to be low risk, as a point of comparison against which later studies addressing specific risks or particular populations could be compared.4 However, it was also anticipated to be large enough for further detailed study of the difference between secure and avoidant attachment, which was Main’s central interest in the mid-1970s.5 Main’s inclusion of fathers in the Berkeley longitudinal study can be situated in the context of shifts in American public and academic discourse in the 1970s that urged recognition of the significance of the father as a parent and as an attachment figure.6 However, Ainsworth’s decision to focus on mothers was a pragmatic one based on her limited resources and the childcare practice of the time in Baltimore. A decade later, distal changes in discourses around the family combined with the better resources of the Berkeley study, leading to an expanded lens on parenting as well as a larger sample.
However, in fact Main did not ultimately have the funds to arrange rigorous coding for all 378 Strange Situation recordings. A report was made by Main and Weston in 1981 from a subset of 46 families who were invited to take part in both the Strange Situation at 12 and 18 months, and in an additional assessment of prosocial behaviour.7 Of the infants seen with mother at 12 months, 68% of the dyads were classified as secure, 28% avoidant, and 4% ambivalent/resistant. For these same infants seen with father at 18 months, 59% of dyads were classified as secure, 35% as avoidant, and 6% as ambivalent/resistant. The distribution was therefore ‘highly compatible’ with the distributions from Ainsworth’s sample, and indeed are well aligned with subsequent distributions.8 Main and Weston reported that the attachment classification of an infant with one parent had a very weak, statistically insignificant association with the attachment classification of an infant with the other parent.9 Furthermore, there was no significant association on the proximity-seeking, contact-maintaining, avoidance, or resistance scales.10
Main and Weston also reported that the attachment classification with each parent made an independent contribution to prosocial behaviour. This was assessed in a rather unusual laboratory-based procedure at 12 and 18 months. With a parent in the room, infants were approached by a clown who sought to play with the child. The clown approached the child with a mask, then removed the mask, called to the infant, and somersaulted about. The clown was asked by an adult to leave, and began to protest wretchedly: ‘Extremely realistic sobbing. The cry lasts 50 seconds. Its realism is so strong that many adults hearing this cry for the first time are somewhat shaken.’11 Prosociality was rated if infants engaged positively with the clown when invited to play and seemed sympathetic to the clown’s crying. Children with two secure attachment relationships scored highest on prosociality; those with two insecure attachment relationships scored lowest; and those with one secure and one insecure attachment were rated in the middle.12 Children with avoidant attachment relationships were found to often remain impassive in response to the crying clown, and their mothers were more likely to show a derisive facial expression when the clown began crying.13 The Clown procedure appeared to especially stress children in avoidantly attached dyads. Unable to seek their caregiver directly for support, around half showed a variety of ‘conflict behaviours’ such as ‘rocking back and forth whilst staring into space; assuming odd postures; engaging in odd tension movements; sudden inappropriate or empty laughter; vocalising to the wall in a “social” manner; an odd “frozen” facial expression; lying on the floor in foetal position with eyes closed’.14 These behaviours were especially common in those who were avoidantly attached with both parents. Main took this to signify that conflict behaviours were elicited especially when the child experienced the relative unavailability of both caregivers as a safe haven. In a later reflection, however, Main concluded that the association between avoidance in the Strange Situation and conflict behaviours in the clown study may have been caused by caregivers ‘who were harsh or frightened in addition to simply displaying aversion to physical contact’.15 The implication is that the close relationship Main initially would perceive between avoidance and conflict behaviour was, in part, an artefact of a sample in which the antecedents of avoidant attachment and the antecedents of conflict behaviours happened to co-occur.
When participants were aged around six years old, in 1982, 40 families were invited back to the laboratory, stratified by infant attachment classification.16 No families with parents who had divorced in the intervening years were included.17 A small grant from the Society for Research in Child Development helped cover the costs of the study, and Main’s lively and intriguing classes attracted a talented group of Berkeley graduate students to work on the project: Donna Weston, Carol George, Judith Solomon, Nancy Kaplan, and Ellen Richardson. Each family was first visited at home by a member of the team, to take consent and build rapport. On arrival in the laboratory, the family viewed excerpts together from one of the films made by James Robertson (‘Thomas: ten days in fostercare’). Study of family responses to the distressing, evocative film of a 14-day separation and reunions with father and mother was the basis of Carol George’s doctoral dissertation with Main.18 George found that the parents from insecure dyads and children from insecure dyads were more likely to show signs of anger, disgust, and sarcasm in response to viewing the child’s behaviour in the film. Yet as well as a stimulus for George’s study, the film served as a prime for the attachment behavioural system and caregiving behavioural systems ahead of subsequent assessments.19 The parents then left the room for individually administered life-history interviews (the AAI). The children remained in the playroom with an examiner. For 15 minutes they were asked to make a drawing of their choice, then one of their family. Next, the examiner asked the child to respond to six pictures of child–caregiver separations, giving their thoughts on what the separated child might feel or do (the Separation Anxiety Test).20 The child and parents were reunited after about an hour’s separation; in half the cases the mother returned first, and in half the cases the father. The reunion was filmed.
Main’s approach to the analysis of her six-year data was deeply influenced by the five years she had spent as a graduate student within Ainsworth’s research group. Main had been drenched to the bone in Ainsworth’s ideas at a point that attachment research as an empirical paradigm was little more than this single laboratory at Baltimore. Ainsworth’s identification of three patterns of infant attachment was felt by Main with incredible force. Why three? Researchers outside of the little laboratory often simply drew the conclusion that the result was arbitrary, primarily an artificial construction of diversity into pragmatic categories.21 But Main had conducted 50 Strange Situations as part of her dissertation research, and was therefore in a position to see the Ainsworth system from the inside and to identify something remarkable: it was astonishingly un-arbitrary. Naturally there were marginal cases between the three groups. Yet the categories themselves seemed to Main to work unnervingly, miraculously well in offering the three basic images to which all the infants approximated in her dissertation sample. There really did seem to be, beyond any reasonable expectation, three patterns of attachment displayed not just in the 26 middle-class Baltimore infant–mother dyads recruited by Ainsworth for her initial study, but by the 106 middle-class infant–mother dyads from the studies by Ainsworth, Bell, and from Main’s doctoral work.
The effectiveness of the Ainsworth classifications seemed to Main to imply something about the fundamental reality of human emotional life. It is hardly unusual for a doctoral student to look out at the whole world through the lens of her supervisor’s work. Chapter 2 discussed how Blatz’s work on security played a foundational role for Ainsworth in her thinking about adult personality, the behaviour of infants in Uganda, and ultimately individual differences in the Strange Situation. However, when Main started to look at the world through the lens of the Ainsworth infant attachment classifications, two surprising things happened. First, Main came to the exhilarating conclusion that the Ainsworth patterns represented the three basic strategies used by all humans, whether infants or adults, for handling distress in interpersonal contexts. Second, exceptions to these patterns therefore took on special significance and interest. Main’s own powers as an observer of all scruffy particularities became combined with a theoretical focus on exceptions, developing a fourth category for behaviour suggesting a disruption in strategy.22 On this basis, Main developed an account of individual differences in attachment as reflecting strategies for the direction of attention with respect to attachment-related perceptions and memories.
These ideas are not widely understood, especially as several key texts by Main were ultimately not published. However, they shaped the development of the AAI and other measures developed by Main, Hesse, and collaborators in the Berkeley group in the 1980s and 1990s. The theory and research of Main and colleagues has repeatedly been situated as a ‘revolutionary shift’ in attachment research.23 Allen and Miga described how the work ‘permits and indeed forces a new conceptualization of attachment, and opens up important avenues for assessing the attachment system beyond childhood’.24 Holmes described it as the start of attachment research ‘Phase 2’, and Toth observed that it brought about the interest in attachment theory among the clinical community that Bowlby had sought but not managed on his own.25
The work of Main and colleagues has offered innovative, persuasive methods and theory with great heuristic power and value for researchers, especially in allowing them to extend the applications of attachment theory beyond infancy. However, at times, the qualities of these methods and theory have led them to enter into circulation without fine-grained distinctions being drawn, especially in the use of concepts and categories. This concern has been widely recognised by developmental attachment researchers in recent years, including by Main and her collaborators themselves.26 There are a variety of reasons for this, which are discussed in this chapter. Of course, matters such as terminological precision and the exact articulation of categories are not really the priorities of pragmatic researchers, for whom the essential concern is whether measurement and prediction can be pursued more effectively or persuasively.27 Yet, looking over the decades in a historical perspective, the pragmatic concerns of individual researchers have at times led to collective problems. Misunderstandings and miscommunication that are only minor irritants or not particular priorities for any individual researcher may, for a field, cause wide-ranging issues, played out incrementally over decades, with costs mounting. In particular, researchers have tended to neglect or skip over the underpinning logic of the contributions of the Berkeley group, to talk past one another as different meanings are given to concepts, and to neglect questions with important consequences for theory, research, and clinical practice.
During the years that Main was completing her doctorate in the Ainsworth laboratory, a fundamental question was the meaning of the behaviour shown by infants in Type A dyads. In 1969 Bowlby published Attachment, situating proximity with the caregiver as the set-goal of the attachment system. Avoidance of the caregiver on reunion therefore seemed at first sight to falsify Bowlby’s theory. In fact, Ainsworth’s home observation data were beginning to reveal a context for this behaviour, in the less-sensitive caregiving received by infants in avoidantly attached dyads. A year into her doctorate, at Ainsworth’s urging, in 1970 Main posted a manuscript to Bowlby for feedback. In a wide-ranging theoretical text entitled ‘Infant play and maternal sensitivity in primate evolution’, Main made a variety of gutsy points. One was to question Bowlby’s assumption that the environment within which humans evolved would reward a single ‘setting’ for the attachment system. Main proposed that the attachment system would need to be capable of calibration to a variety of environments, favourable and adverse. Sensitive caregiving is optimal, and the provision of a secure base would help a child to explore and learn. However, less-sensitive caregiving could be expected to elicit responses that would support survival even in adverse conditions. The vigilance of infants in ambivalent/resistant dyads was readily explicable in this account, as it helped retain proximity to a caregiver in potentially dangerous environments.28 Avoidant behaviour, however, remained a puzzle for which Main, as yet, had no explanation.
Main’s 1973 doctoral research headlined results showing the higher scores on cognitive development of the children from secure dyads.29 However, perhaps the most critical finding for Main’s later thinking was the observation that infants in secure dyads in her sample had the longest attention spans during play as toddlers. This suggested to her that children in insecure dyads had other behavioural systems or responses active and sapping attention from exploration. During her doctorate, Main also served as a coder for Berry Brazelton in a micro-analysis of filmed interactions between five mother and infant dyads, which occurred weekly over the first months of life. The focus of the analysis was on ‘cycles of looking and non-looking, or attention and nonattention’.30 Though Brazelton had come into the study with the assumption that the infant would display continuous attention within interactions, the stop-frame analysis conducted by Main revealed a flexible movement between looking at and away from the caregiver. The researchers concluded that ‘looking away behaviour reflects the need of each infant to maintain some control over the amount of stimulation he can take in via the visual mode in such an intense period of interaction’.31 This looking away strategy was adopted more frequently when the caregiver was insensitive in offering stimulation that was not well aligned with the infant’s pacing. Drawing on contemporary ideas from ethology, the researchers called this an ‘approach-withdrawal model’, a term that had been used to describe the way that animals might flexibly use approach and withdrawal behaviours to modulate the intensity of stimulation, and in this way remain well regulated.32 This conclusion was supported by the finding that infant looking away behaviours were less common when caregivers were alert to indications that the infant was becoming ‘upset or disintegrated’ or otherwise having difficulty regulating:33
The authors criticised Bowlby for focusing too strictly on behaviour, following his ethological models, and his neglect of the role of attentional processes as a flexible resource for maintaining regulation and coherently structured interaction. They speculated that the interdependency of rhythms of attention and inattention within the relationship between infant and caregiver ‘seemed to be at the root of their attachment’.35
After Main finished her doctorate in 1973, she continued reflecting on avoidant attachment. Her assumption was that, in contravening the predicted set-goal of the attachment system so directly, it would be ‘predictive of interactive and affective disturbance’.36 She developed scales for analysing caregiver–infant touch in the Ainsworth home observation data. Main found that the mothers in dyads classified as avoidant in the Strange Situation rebuffed physical contact with their babies much more frequently at home. They often rejected their infants’ attachment behaviour, and more frequently displayed anger and flat affect in response to displays of distress. Over time, this increased the frequency with which their infants made tentative or circuitous approaches. In February 1974, Ainsworth wrote to Bowlby that ‘we have found plenty of evidence that the mothers of A babies dislike physical contact, and that it is through behaviour relevant to physical contact that they (at least in large part) express rejection. Mary’s theory is that this puts babies in a double bind, for they are programmed to want contact and yet are rebuffed (or at least have unpleasant experiences) when they seek it.’37
Main’s undergraduate background in liberal arts at St John’s College made her aware that novelists such as Hardy and Dostoevsky had ‘long been aware of the attraction irrationally implicit in rejection’.38 Whilst at the College, she married her philosophy professor, Alvin Nye Main. In the early 1970s, Alvin Main drew Mary Main’s notice to a potential evolutionary basis for the attraction that may be prompted even in the context of rejection.39 In The Voyage of the Beagle, Darwin noted that the amphibious sea lizard, when frightened, would never flee towards the water. To test this further, Darwin repeatedly frightened sea lizards, and then threw them into the water. They repeatedly swam right back to him on the land, directly towards their ‘attacker’. Darwin interpreted this otherwise strange and counterproductive behaviour in terms of the sea lizard’s evolutionary history:
Alvin and Mary Main discussed the analogy with the infant of a caregiver disposed to rejecting or angry behaviour. For the sea lizard, ‘if the source of the attack is “the shore” itself, the shore is nonetheless returned to as a haven of safety’.41 Similarly, according to Bowlby’s theory of the attachment behavioural system, the infant would be disposed by an evolutionarily channelled mechanism to seek proximity with the caregiver when alarmed as the haven of safety, even if the caregiver is also experienced as rejecting or threatening.42—hence a ‘double bind’.
Alvin Main died of cancer in the summer of 1974, just following the end of Mary Main’s first year as an assistant professor at Berkeley. In the grief of this loss, Mary Main experienced a period of writer’s block that lasted around two years.43 One text written during the period was a paper presented by Main and her friend Everett Waters, delivered in July 1975 to the International Society for the Study of Behavioural Development at the University of Surrey. Waters had earlier studied gaze aversion shown by infants in relation to the stranger. He proposed that gaze aversion can best be understood as an attempt to ‘cut off’ the stimulus from perception, which serves to modulate tendencies that it might evoke.44 This agreed with the conclusions of Main’s work with Brazelton. Main and Waters applied this idea to the behaviour of infants in Type A dyads seen in the Strange Situation.45 They speculated that, as well as modulating arousal, visual cut-off allows the infant to remain close to a rejecting or angry parent—without ‘falling into “all or nothing” response patterns which cannot be terminated voluntarily such as approach for comfort, or crying’.46 In Ainsworth’s observations, the children from dyads classified as Group A were precisely, and apparently paradoxically, those who showed more distress and clinginess at home than other children (Chapter 2). In Main’s only published work from this period, a book chapter in a volume dedicated to the memory of her late husband, she proposed that the child who shows avoidant behaviour turns to objects of attention in the environment that will not add to heartache, or that might even provide the relief of distraction. In this way the child may retain control and flexibility of response.47 Main offered her qualitative impression, however, that infants in avoidantly attached dyads were still more disposed to show conflict behaviours in the home than infants in secure dyads.48
Ainsworth had repeated the Strange Situation with the first wave of her sample two weeks after the first administration (Chapter 2). All seven of the infants in avoidantly attached dyads displayed conflict behaviour as they broke from their avoidance in panicked approach to the caregiver. The intensified activation of the attachment system from yet more separations, Main supposed, made cut-off and therefore avoidance impossible. As a consequence, these infants were no longer able to stay well regulated and near to the caregiver. Instead they were thrown into the ‘all or nothing’ response of direct approach, despite their concerns about the rebuff their experiences of their caregiver had led them to expect. The infants’ approach was mottled by conflict behaviours as markers of this concern.49 These reflections were further developed when Main spent September 1977 to May 1978 as a Fellow at the Centre for Interdisciplinary Research in Bielefeld, West Germany. The focus that year was on the application of biological principles to observable behaviour, and the other Fellows represented future leaders across a wide variety of disciplines. Alongside Main, other attendees included the biologists Richard Dawkins and John Maynard Smith, Robert Hinde’s student and collaborator Patrick Bateson, Harry Harlow’s student and collaborator Stephen Suomi, and the child psychiatrist and psychoanalyst Robert Emde.50 The Fellowships were coordinated by the ethologist Klaus Immelmann and by Klaus and Karin Grossmann, who were in the process of attempting the first replication of Ainsworth’s study outside of America. Fellows engaged in structured and unstructured group conversations at least two or three times a week with biologists and ethologists.
During her Fellowship in Bielefeld, Main applied ethological principles to the interpretation of avoidant attachment behaviour as seen in the Strange Situation. Ethology asked four questions of behavioural sequences (Chapter 1): first, the contribution of the behaviour for species survival or reproduction; second, how the behaviour came about through natural selection; third, the behaviour’s underpinning mechanisms; and fourth, how it develops over the lifespan. Putting these questions to avoidant behaviour in the Strange Situation set Main on a collision course with Bowlby. Bowlby, and following him Ainsworth, had presumed that the environment within which humans evolved would have predisposed infants to seek their caregiver when alarmed, as their ultimate source of protection. Other behaviours were therefore abnormal and pathological. However, Maynard Smith and Dawkins were in the process of introducing a new account of the evolutionary biology of mating and reproductive behaviour, based on game theory. In game theory, the actions of individuals are treated as moves in a game, responsive to the environment and its potential costs and benefits. Applied to evolutionary processes, game theory suggested that species may develop a repertoire of strategies that can be initiated in response to various expectable environments and their costs and benefits.
In papers from 1979, Maynard Smith and Dawkins argued that the diversity of successful mating strategies could be explained if it was assumed that natural selection had predisposed a repertoire of behavioural patterns for achieving reproductive success in diverse circumstances.51 The direct achievement of mating would be sought when conditions were favourable. However, a range of more circuitous routes to mating success were also available when a direct approach was not readily feasible. The latter were termed ‘conditional strategies’ by Maynard Smith and Dawkins. The term is a technical one. It does not imply that the direct achievement of a function or goal is ‘unconditional’; direct expression of a goal-oriented response is usually dependent on a facilitative environment. Rather, the term ‘conditional’ was intended to refer to expectable behavioural sequences oriented towards at least partial achievement of a function or goal that would be preferred if the direct strategy could be anticipated to be unsuccessful. Main took this model of reproductive strategies and applied it to the Strange Situation. Just as the environments for reproductive success varied for the insects, reptiles, and birds of interest to Dawkins and Maynard Smith, so the caregiving environments of infants was also one with expectable variation in human evolutionary history.
From Bielefeld, Main sent Bowlby a manuscript entitled ‘Avoidance of the attachment figure under stress: ontogeny, function and immediate causation’ towards the end of 1978. She started from the ethological premise that ‘maternal investment varies. Some desert and all must wean their infants. There must be a strategy for survival under these conditions.’52She acknowledged that, at first sight, avoidant behaviour seen in the Strange Situation appears to disprove Bowlby’s position that proximity will be sought when the attachment system is activated. However, the observation can be reconciled with the theory if it is assumed that ‘avoidance of the attachment figure may function as a kind of conditional strategy for proximity maintenance’. Conditional proximity supplies a stand-in for a relationship that would offer genuine welcome. The infant can remain close by to a caregiver who might rebuff them if they attempted a direct approach, a rebuff that would be emotionally painful and might further reduce the caregiver’s availability. To set out the apparent paradox: the relinquishment of full proximity implied by avoidance is part of its effectiveness, part of its basis for a qualified hope of a qualified caregiver availability.
Main later developed these ideas in two groundbreaking publications: a short note from 1979 entitled ‘The “ultimate” causation of some infant attachment phenomena’, and the 1981 chapter ‘Avoidance in the service of proximity: a working paper’ published in a book of papers by the Fellows at Bielefeld and co-edited by Main. In both texts, Main criticised Michael Lamb and colleagues, who had stated that Bowlby’s position implies that infant avoidant behaviour is maladaptive since proximity with the caregiver is the basis for protection.53 Main countered with an alternative reading of Bowlby. Bowlby himself had referred in passing to ‘insecure attachment’ as a ‘strategy’ in Separation, though his discussion had been of the clinginess of the ambivalent/resistant pattern.54 And Main highlighted Bowlby’s emphasis on the evolutionary basis of behaviour. From an evolutionary perspective, avoidance could be interpreted as a proactive response by the infant that ‘paradoxically permits whatever proximity is possible under conditions of maternal rejection’.55 Evolutionary processes would have selected for avoidance as one part of the infant repertoire for responding to caregivers, since infants who are able to avoid antagonising their caregivers or making demands that their caregiver will rebuff are more likely to have survived. Infants successfully utilising an avoidant strategy maintain an indirect but real proximity to their caregiver, as well as the regulatory control to continue to be responsive to the environment.56
Main also considered the ethological question of the behaviour’s underpinning mechanisms. She regarded the phenomenon as, essentially, ‘active visual, physical and communicative avoidance’ of the caregiver in the context of an activation of the attachment system.57Yet, drawing on her work with Brazelton and with Waters, the paradigmatic form of avoidance for Main was gaze aversion. In Bielefeld initially, and then to a greater extent in 1980 during a second visit to the Grossmanns after their move to Regensburg, Main gradually came to the conclusion that gaze aversion was the most potent external marker of a fundamental internal process underpinning the visual, physical, and communicative qualities alike. This internal process was the redirection of attention away from the caregiver. This would inhibit the activating conditions of the attachment behavioural system. Gaze aversion and other forms of avoidant behaviour were therefore not just keeping the wish to approach from the caregiver’s notice. They were also keeping this wish from the baby itself.58 A version of ‘Avoidance in the service of proximity’ was later published in a celebratory volume in honour of Bowlby called The Place of Attachment in Human Behavior, edited by Colin Parkes and Joan Stevenson-Hinde. Other than some sections cut for concision, the only substantive change came in the conclusion. There it was argued that in infant avoidance of an attachment figure, ‘we might better conceive of thought and behaviour as becoming actively reorganised away from the parent and the memory of the parent’.59
However, in the reception of Main’s ideas from this period, the idea of attachment behaviour as reflecting attentional processes was little noticed. Instead, her application of evolutionary game theory in conceptualising the Ainsworth classifications as ‘strategies’ took centre stage, without bringing the attentional model with it from the wings. Given that both ideas are emphasised across Main’s writings in the 1980s and 1990s, the period in which the received image of her ideas was established, this reception is curious at first sight. Three suggestions can be offered for the greater popularity of the idea of attachment strategies to the exclusion of attentional processes in the reception of Main’s ideas. First, attention seemed to already be encompassed by the concept of ‘internal working models’. Main’s emphasis on attention therefore did not seem salient or novel, even if, in practice, when internal working models were defined or operationalised the focus was rarely on attention. Second, the concept of attachment strategies addressed a problem with Bowlby’s theory. Bowlby’s presumption that the secure response was the one primed by evolution offered no basis for understanding why avoidant and ambivalent/resistant behaviour could be seen in sample after sample studied using the Strange Situation. By contrast, Main’s account of individual differences in attachment as underpinned by attentional processes solved no existing problem. Additionally, methodologies for the study of attentional processes in young children, such as eye-tracking technology, were still emerging.60 A third reason for the predominant emphasis on ‘strategies’ in interpretations of Main’s theory may have been the timeliness of the concept. A sudden burst of appeals to the idea of ‘strategy’ occurred across the human sciences in the late 1970s, especially in conceptualisations of the formation of family units.61 It was a concept that allowed social scientists to model expectable, patterned responses by individuals to their social environment, whilst keeping open the possibility for change. The idea of ‘strategies’ permitted the explanation of patterns of social interaction within and by families, as well as offering a lens on the techniques, including self-regulation, used in order to maintain this stability.62
With her account of attentional processes underpinning attachment strategies, Main was developing a model of individual differences seen in the Strange Situation. The infant in a securely attached dyad could flexibly turn attention to the environment or the caregiver in a manner responsive to the changing situation. Main regarded this strategy as reflecting a lack of concern regarding access to proximity with the caregiver, permitting the child’s focus to be responsive to the changing environment of the Strange Situation. The infant in an avoidantly attached dyad, by contrast, was conceptualised by Main as attempting a partial reorganisation of attention away from the caregiver and towards the environment, in order to raise the threshold for activating and lower the threshold for terminating the attachment system. Gaze aversion, turning attention to the toys, and facing away from the caregiver all served to keep the caregiver out of the infant’s focal awareness, thinning the influx of perceptions that would otherwise activate the attachment system (and that might also activate anger). This then raised the question of the evolutionary basis and underpinning mechanisms of ambivalent/resistant behaviour in the Strange Situation.
As early as 1970 Main had come to conceptualise the proximity-seeking and clingy behaviour of infants in ambivalent/resistant dyads as explicable in an evolutionary sense: vigilance regarding the availability of the caregiver might be expectable and helpful in some environments.63 Main’s further reflections on ambivalent/resistant attachment in the late 1970s were informed by a groundbreaking paper by Robert Trivers.64 Whereas Bowlby had typically treated the infant’s caregiver in terms of the caregiving behavioural system, Trivers emphasised that caregiving was only one priority for a parent. Other demands include the parent’s own survival, maintenance of sexual and social relationships, and the care of other offspring. There was every potential for conflict between parents and offspring. In fact, Bowlby had already noted this in passing, in his observation of angry and distressed protest among human toddlers during weaning or when their mother turns attention to a new baby.65 Bowlby’s remarks were influenced by the observations of his friend and colleague Robert Hinde, whose research group had observed angry and distressed protest behaviour by young rhesus monkeys in response both to weaning and to short-term separations from their mother.66 Trivers was fascinated by Hinde’s findings, and especially the observation that the more frequently infants were pushed away by their mother prior to separation, the more distress, tantrums, and clinging the infants showed on reunion. Trivers interpreted these responses as indicating ‘that the infant interprets its mother’s disappearance in relation to her predeparture behavior in a logical way: the offspring should assume that a rejecting mother who temporarily disappears needs more offspring surveillance and intervention than does a nonrejecting mother who temporarily disappears’.67 The clinging, protesting behaviour observed by Hinde and others was regarded by Trivers as serving to signal to the caregiver that the child needs more investment before they would be able to survive alone. He regarded such signalling as part of the primate behaviour repertoire, selected by evolutionary processes, since it may have contributed to infant survival under conditions where caregiver investment in the child might otherwise not be forthcoming.
Influenced by Trivers’ ideas, in a paper from 1979 Main identified that Trivers’ argument has strong relevance for thinking about ambivalent/resistant attachment.68 A lowered threshold for activation of the attachment system, and the alternation of angry and distressed behaviours, could be regarded as a ‘conditional strategy’. It may well serve to retain the attention of the caregiver and prompt the activation of the caregiving system. If the child feels invisible to the caregiver, displays of distress, anger, and proximity-seeking/contact-maintaining can be anticipated to increase the child’s visibility. However, it is a conditional rather than a primary strategy, since it is unnecessary when the attachment system can be satisfied directly through proximity-seeking. Furthermore, as Bowlby noted in Separation, infant anger may coerce caregiver attention to attachment signals, but it also has the potential for backfiring by antagonising the caregiver.
In 1982, Jude Cassidy—then a doctoral student with Ainsworth—spent a year in Main’s laboratory at Berkeley. Main and Cassidy reflected on what ambivalent/resistant attachment would look like after infancy, and concluded that it would appear as behaviour that attempted ‘to exaggerate intimacy with the parent, dependency on the parent, and his or her relatively immature status’ with at least some ‘hostility or physical ambivalence’ evident.69Whereas avoidant attachment represented a conditional strategy premised on the minimisation of attachment signals, ambivalent/resistant attachment included an intensification of attachment signals as well as their punctuation by anger. This was an idea that Cassidy brought back to Ainsworth’s group at Virginia, where it generated a great deal of discussion, and was considered in print by Cassidy and Roger Kobak in 1988.70
In a paper published in 1994, Cassidy offered an interpretation of avoidance and ambivalence/resistance in terms of emotion regulation, as ‘minimising’ or ‘maximising’ the attachment behavioural system itself (Chapter 5).71 Cassidy’s account aligned with Bowlby’s discussion of defensive exclusion as inhibiting the activation of behavioural systems, without specification of what exactly was being excluded (Chapter 1). It also aligned with other voices, such as Schore, calling at the time for the reinterpretation of attachment as, in general, the process through which self-regulation is achieved within close relationships.72 It should be noted though that in contrast to Cassidy’s more diffuse position, Main emphasised that these observable differences in strategy are underpinned by differences in attentional processes.73 It is unclear the extent to which Cassidy realised that her description of minimising and maximising departed from Main in this regard. However, Kobak astutely observed already at the time that two different models of ‘minimising and maximising’ seemed to be in play. For his part, he strongly aligned himself with the centrality of attentional processes, criticising Main for not doing more to highlight to readers the fundamental role of attention in her conceptualisation of conditional strategies.74
For Main, avoidance was based on the direction of attention away from the caregiver, as well as other perceptions and memories that may activate the attachment system.75 The infant’s behaviour seemed to Main to be oriented to maintain the caregiver’s peripheral rather than focal attention, presumably since the caregiver tended actually to be most available when not directly concerned with the baby: ‘if the mother picked them up, they turned away and, as though attempting to distract her attention from themselves, pointed to toys and other aspects of the environment’.76 In pointing to a toy on reunion, a behaviour that had perplexed Ainsworth, Main saw a tactic through which the infant in an avoidant dyad could manage the caregiver’s attention, so that the child remained in the caregiver’s awareness but not directly the object of concern.77 By contrast, according to Main, distress and anger were intensified for infants in ambivalent/resistant dyads by attentional vigilance towards the availability of the caregiver, as well as other possible perceptions and memories that might hold information relevant to the caregiver’s availability. In turn, this strategy centred the caregiver’s own attention on the child:
The term ‘conditional strategy’ was drawn from ethological reflection on alternative forms of adult sexual behaviour in non-human animals. Main similarly did not limit conditional strategies to childhood in her theory. She proposed that the same strategies could be observable in caregiving behaviour: either minimisation of attention to the child and reduced activation of caregiving, or intensification of concern with the child.79 For instance, following Trivers, she acknowledged that caregivers’ past experiences or present adversities may lead them to minimise activation of the caregiving system, for instance to avoid overwhelming feelings evoked by the child or to ensure resources remain available for other challenges.80 The minimising conditional caregiving strategy, Main suspected, may also serve to prime an infant for an environment in which early independence might be beneficial. Conversely, the maximising conditional strategy may serve to prime an infant for an environment in which prolonged dependence on relationships would be especially salient and valuable. However, Main was at pains to emphasise that the adoption of one caregiving strategy or another need not be conscious, and that it would be a category error to regard parents as motivated by the evolutionary function of a strategy.81 Instead, the proximal cause will be the demands of the present.
In a paper from 1990, ‘Cross-cultural studies of attachment organization’, Main argued that ‘the maintenance of differing conditional strategies entails the utilisation of similar cognitive mechanisms across individuals’.82 This made clear that she was playing for grand stakes; her proposals amounted to no less than a global model of human emotional life. Elaborating on the 1990 claim in a later work, Main asserted that ‘there exist species-wide abilities that are not part of the attachment system itself, but can, within limits, manipulate (either inhibit or increase) attachment behavior in response to differing environments’.83 As described in Chapter 1, already in 1956 Bowlby, Ainsworth, and colleagues had written that ‘the personality patterns of children who have experienced long separation tend to fall into one or other of these two opposite classes’: either ‘over-dependent’ and ‘ambivalent’ or ‘mother-rejecting … having repressed their need for attachment’.84 This was already a momentous claim. However, whereas Bowlby and Ainsworth discussed the two classes as responses to separation, in her 1990 paper Main resituated these classes as the fundamental alterations of the attachment system and caregiving system, or potentially any behavioural system.
Main’s model of conditional strategies was not, therefore, merely a localised psychological theory to account for the Strange Situation, but a philosophy of human experience in general, with resonances of Plato and Kant. These resonances signal the important, if complex, role of philosophy on the development of Main’s ideas. The only mention of Main’s parents in her published works is the remark that ‘philosophy was much admired by my parents, who had introduced me to Plato, Kant, and several Eastern philosophies by age 10’.85 As an undergraduate, she re-encountered Plato and Kant at St John’s College in Maryland in the classes of her future husband, Alvin Main.86 Now, in her 1990 paper, she was proposing that behind the diversity of apparent infant behaviours there lies three essential forms for responding to distressing and challenging situation (cf. Plato, Aristotle); the basis for these differences stems from the structure of human experience, which shape human perception, language, and behaviour (cf. Kant).
Though the three Ainsworth categories were identified using 23 middle-class infants in the mid-1960s in Baltimore, for Main the categories exceeded this particularity to express the three basic ways that humans can respond to distressing and challenging situations. In the context of worries or other troubling feelings, there are three basic approaches: we can communicate about our feelings to someone we anticipate or hope might help us; we can keep our feelings to ourselves; or we can make our distress and frustration someone else’s problem. As we saw in Chapter 2, Ainsworth took from Blatz the idea that the people we may need to depend upon are independent of us in a deep sense, and this can be regarded as a source of worry or as a source of reassurance, depending on what this freedom has implied for us in the past. Main’s conditional strategies were responses to this predicament, based on the selective exclusion of information that would risk reducing the availability of the caregiver as a secure base and safe haven:
An infant in an ambivalent/resistant dyad directs attention away from potential information that might suggest that the environment is unthreatening and that the caregiver is available. This infant therefore sees clearly, indeed overclearly, the human predicament that attachment figures are independent of us in worrying ways. Infants in an avoidant dyad direct attention away from potential information that might elicit alarm, distress, or a tendency to approach the caregiver for comfort. They see clearly, indeed with a reductive clarity, that as humans we are partly independent of our attachment figures. Each of the conditional strategies is acutely and effectively attentive to some aspect of reality, and in this way enacts a method of responding to alarm that may be of survival value under conditions where that aspect is of special importance. This may provide important information for children about how to calibrate the demands they make on the world in order to achieve what nurturance and resources may be available. The conditional strategy may also be experienced as a kind of ‘secondary felt security’, despite being held in place by anxiety; the reason for this is that the conditional strategies nonetheless offer predictable and therefore reassuring access to some sense of closeness and regulation.88 However, for Main, each conditional strategy must also depend upon selective exclusion of another aspect of reality.89 There will be, she predicted, a price to be paid for the conditional strategy in the long run, to the degree that this information about attachment relationships and about their own affective life remains lost or relatively opaque to the individual.90
With Main’s model of Group A and Group C as conditional strategies, attachment theory became a global account of human emotion and relationships.91 For Main, two conditional strategies existed for the attachment and the caregiving systems, and potentially for other behavioural systems. Its output could be minimised or intensified, and the basic architecture for these strategies would lie in the direction of attentional processes.92 The magnitude of the impact for subsequent attachment theory of the encompassing conditional strategy model of individual differences in attachment cannot be overestimated. It has had more sway for later attachment theory than even the theory of attachment as a behavioural system that it ostensibly modified. Few developmental attachment researchers in the twenty-first century have made more than token mention of behavioural systems, or developed specific hypotheses from this concept (though see Chapter 5). Since the 1990s, the idea of attachment as a ‘behavioural system’, grounded in human evolutionary history, has rather the status of a memento that the developmental tradition of attachment research is pleased to have, and without which they might feel bereft, but which never gets actually brought out for further examination, except perhaps when teaching.93
By contrast, the image of human difference as strung out along axes of minimising and maximising has been the beating heart of attachment research for three decades, and central to hypothesis generation and the interpretation of empirical findings regarding individual differences. It has formed the ‘grid of specification’ according to which kinds of attachment behaviour in infancy and beyond have been divided, contrasted, related, grouped, classified, and derived in relation to one another.94 The idea of minimising and maximising strategies is frequently cited as originating with Main, sometimes with Cassidy and Kobak, and occasionally with Hinde.95 However, most often the image has been taken as simply a timeless part of attachment research as a paradigm, implied by Ainsworth’s introduction of three patterns of attachment. This interpretation has been supported by Main’s tendency to attribute her own ideas to Ainsworth, where these grew out of the soil of Ainsworth’s own thinking.96 It has likely also been supported by the elegant simplicity, amounting to apparent obviousness, of the theory of individual differences divided into minimising and maximising.
Ainsworth was, understandably, rather astonished that Main transformed her categories for the Strange Situation into an encompassing philosophy of existence, applicable to all behavioural systems. In a letter to John Bowlby in March 1984 (embarrassingly posted by accident, in fact, to Main),97 Ainsworth expressed ‘unease’ with Main’s ambitious and universalising proposals:
However, as she learned about Main’s theories over the course of 1985 and became increasingly impressed with the promise of Main’s AAI, Ainsworth remarked to Bowlby: ‘You were right that I am in a sense a student of Mary Main’s.’99 Ainsworth’s support played a critical role in establishing Main’s account of conditional strategies as the central image of individual differences within attachment theory, as a paradigm primarily concerned with individual differences. In particular, Ainsworth’s late paper with Eichberg reporting an exceptionally strong relationship between the infant classifications and the new AAI was of great symbolic as well as scientific importance.100 Ainsworth supported and encouraged her different graduate students; Inge Bretherton, Everett Waters, Patricia Crittenden, and Bob Marvin were all making both methodological and theoretical innovations at the turn of the 1990s that built on Ainsworth’s contribution.101 However, by then the Ainsworth and Eichberg paper was already in print, and quickly came to be widely interpreted as confirmation of Main’s extension of the Ainsworth categories across the lifespan, passing the baton of Ainsworth’s role as the field’s method-giver.102 In addition to Ainsworth’s support, Main’s characterisation of avoidance and ambivalence/resistance as conditional strategies was seen as a powerful explanatory tool. And the AAI, the first trainings in which by Main were co-taught with Ainsworth, was regarded as an exciting methodological development. In her final writings, Ainsworth urged her successors to retain an open-ended system, and not to close either theory or method around her categories.103 However, a central goal of the second generation of attachment researchers was the construction of a cumulative empirical research paradigm of replicated results, and for this a settled coding system, rather than further exploration, was treated as desirable.
The conventional reference for the theory of conditional strategies over the decades has been Main’s 1990 paper ‘Cross-cultural studies of attachment organization’.104 It was more elaborate than ‘The “ultimate” causation of some infant attachment phenomena’ from 1979, which is more of a note than a full paper. Her 1981 book chapter ‘Avoidance in the service of proximity’ was, by its own admission, meandering. It was also published in an edited volume in a context in which journal articles, across the social sciences, were coming to receive comparatively greater prominence. However, as a consequence, readers have generally missed the role of attention as the architecture underpinning Main’s conceptualisation of conditional strategies. This was mentioned to a greater or lesser extent in all her work of the period—except the 1990 paper.
Main observed that the conflict between anger and attachment could flood out and interrupt the smooth expression of the attachment system. However, most ambivalent/resistant infants used the oscillation between distress and anger in quite a smooth way to retain the attention of their caregiver: when they were put down they cried to be picked up; when they were picked up, they squirmed away. Just as avoidance was paradoxically in the service of proximity for Group A infants, the distress and conflict of Group C infants appeared to in the service of organisation via proximity. In Main’s perspective, the ambivalent/resistant classification seemed ‘much the least well organised’ of the Ainsworth classifications.105 And in some work under Main’s supervision, the ambivalent/resistant and unclassifiable infants were grouped together for analysis on the basis that both displayed conflict.106 Nonetheless, ultimately, Main concluded that ‘like an avoidant baby, a resistant baby may be said to be “organized” in the sense of having a singular attentional focus’, i.e. retaining the attention of their caregiver.107
Main went to visit Bowlby in March 1978, and they discussed these ideas. In the course of these discussions Bowlby was persuaded that the attachment system may have evolved with a repertoire of strategies.108 They discussed early draft material towards his book Loss, and in the final version he cited Main’s perspective with approval. Bowlby had already from 1977 taken an interest in caregiving and compliant behaviour shown by children towards parents when they might not receive adequate care (Chapter 1). Main’s proposal seemed to fit with this, since avoidant, caregiving, and compliant behaviour had in common the substitution of direct attempts to seek the caregiver’s support for an alternative strategy that would be more effective given parental behaviour.109 The exclusion of information that might activate the attachment system for children in this situation would have the predictable outcome of increased caregiver responsiveness and so survival value within human evolutionary history.110 In Loss, Bowlby therefore situated the avoidant conditional strategy within the broader observation that ‘as children know in their bones, when mother is prone to be rejecting it may be better to placate her than to risk alienating her altogether’. However, he also acknowledged Main’s emphasis on attention as the underpinning infrastructure of avoidance as a conditional strategy:
Against his earlier position and also the common image of his work still in circulation today (Chapter 1), Bowlby can be seen here in Loss following Main in turning away from an image of the display of attachment behaviour as natural and best, towards acknowledgement of the evolutionary function and the at least short- or medium-term benefits of diverse forms of infant behaviour. And Bowlby would continue to refer to avoidance as a ‘strategy’ in his subsequent writing, acknowledging both short-term advantages and potential survival value and the longer-term contribution the strategy might make to mental illness.112 Such treatment of avoidance as a conditional strategy fitted with Bowlby’s wider transition in his late work from regarding proximity as the set-goal of the attachment system to seeing this as the physical and attentional availability of the caregiver.
However, though recognised by Bowlby, fundamental aspects of Main’s account of conditional strategies have been missed by the subsequent attachment literature. Klaus and Karin Grossmann suggested that a key reason for this is that attachment researchers have generally regarded Main’s introduction of the ‘disorganised attachment classification’ as superseding any early reflections.113 As a result, very few attachment researchers have read any of Main’s papers from prior to 1990. Attachment researchers repeat—and repeat—the idea of ‘minimising’ and ‘maximising’ strategies in summaries of the paradigm and in interpreting empirical results. Yet the fact that the original idea was of minimising or maximising attentional processes, specifically, is essentially unknown. Main made several efforts to clarify the centrality of attention to her account of individual differences in attachment. The most strenuous was her decision to title a paper from 2000 ‘The organised categories of infant child and adult attachment: flexible vs inflexible attention under attachment-related stress’.114 Yet besides her most immediate collaborators such as Erik Hesse and Marinus van IJzendoorn,115 and a few others with personal links to Main at one period or another such as Roger Kobak, Pehr Granqvist, and Carlo Schuengel,116 researchers generally interpret Main as describing the minimisation or maximisation of either attachment behaviour or of distress (or both), rather than recognising her proposal of a causal account with attentional processes at the centre. This may have been facilitated by a lack of clarity regarding the relative overlap or differences between Main’s specific concern with attentional process and Bowlby’s descriptive concept of defensive exclusion, since in Loss Bowlby treated the two accounts as aligned (Chapter 1). Difficulties in identifying and understanding Main’s position have also been facilitated by Cassidy’s influential reinterpretation of minimising and maximising strategies in terms of ‘emotion regulation’ in the 1990s.117 In any case, eclipse of the role of attentional processes in individual differences for Main has made for much more general and less parsimonious hypothesis-generation. It has additionally obscured the nature of the links Main perceived between infant behaviour and individual differences in later development.
In her doctoral dissertation Main had documented various forms of conflict behaviour in the Strange Situation, including ‘stereotypies; hand-flapping; echolalia; inappropriate affect (inexplicable fears, inappropriate laughter) and other behaviours appearing out of context’.118 She had also been impressed by an incident during her doctoral research. She had been meeting with an infant, Sara, and her mother:
Sara received the highest score for avoidance in the sample in the Strange Situation, and her mother had the highest score for aversion to physical contact with her infant. Main concluded that Sara’s approach-avoidance conflict caused by the lightning was too intense to permit an organised avoidant response. The result was a redirection of attachment behaviour towards Main herself. For a time, interested to see if this effect would be replicable, Main planned an empirical study in which ten infants avoidantly attached with both parents and a comparison group of infants would be placed equidistant between the experimenter and the mother and exposed to a sudden, loud crashing noise and flickering lights.120 Fortunately for the infants, this study was not carried out.121
One reason may have been that it was not necessary: Main found that she could reliably observe conflict behaviour in other contexts. One was among young children who had experienced abuse. Main’s graduate student Carol George proposed that conflict behaviour was expectable when young children experienced physical maltreatment by their caregiver. George observed ten physically abused toddlers and ten matched controls in a San Francisco daycare. She conducted careful qualitative observations, making detailed narrative notes. Based on analysis of these notes, George and Main reported that ‘all of the abused children but none of the controls were observed to respond to peer affiliations with approach-avoidance behaviour,’ such as approaching a professional carer but with their head averted.122
Additionally, Main found that she could reliably elicit conflict behaviour in the laboratory. Conflict behaviour was frequent, especially among infants in avoidantly attached dyads, in response to the unnerving crying clown both in Main’s Berkeley sample and in the Grossmanns’ Regensburg sample where the Clown procedure was also applied. In 1981, Main, working together with her graduate student Donna Weston, formulated an unpublished ‘Scale for Disordered/Disoriented Infant Behaviour’ for the Clown study. The original version appears to be lost, but a version of the scale was typed up by Main for Karin Grossmann in June 1982.123 The manuscript shared with Grossmann is a nine-point scale, indexing behaviours including ‘stereotypies, episodes of immobilisation, disoriented behaviour, misdirected behaviour, sudden disordered outbursts of activity, and sudden uninterpretable noises or movements’. In the 1982 manuscript, behaviour is identified by Main as ‘disordered’ based on the ‘extent to which such behaviour may be indicative of difficulties in functioning’, for instance by virtue of lacking either ‘orientation’ or ‘purpose’.
Main was startled and intrigued by the fact that so many of these odd behaviours in human infants resembled the conflict behaviours described by Hinde in animals (Chapter 1): simultaneous or sequential contradictions in approaching the caregiver, misdirected approaches, poorly coordinated combinations of movements, freezing in place, or signs of confusion, out-of-context anger, or tension on reunion with the caregiver. In some cases, a child’s behaviour in the Strange Situation seemed interrupted or misshapen by conflict behaviours. However, in other cases, the infant’s response on reunion with their caregiver was dominated by other behaviours, making an Ainsworth classification impossible. Hinde was in Berkeley for the year 1979–80 as the Hitchcock Professor at the University of California, so Main was able to discuss her observations and ideas with him; she later described the influence of Hinde on her thinking about conflict behaviours shown in the Strange Situation as ‘strong and direct’.124
In a paper from 1981, Main and Weston published an interim report on 46 families where the Strange Situation had been coded for both infant–mother and infant–father interaction. They reported that 12.5%125 could not be classified using the Ainsworth system: ‘infants were not judged unclassifiable if they merely showed conflicted behavior or behaved oddly during the strange situation: one infant stared, talked to the wall, and indeed seemed almost to hallucinate, but was nonetheless regarded as secure in relation to the parent. Infants were judged unclassifiable only if their social and emotional behavior toward the parent could not be encompassed by the present Ainsworth system.’126 However, in fact the large majority of the unclassifiable infants did show conflicted, confused, or apprehensive behaviours.
Yet Main and Weston’s report on ‘unclassifiable’ cases was only on the Strange Situation procedures they had coded by that point. This was a fraction of the hundreds of Strange Situations conducted by the research group. The others still needed to be coded. In 1977, Judith Solomon joined Main’s lab following a graduate focus on ethology and comparative psychology, which included attention to conflict behaviours in animals. Yet Main left soon after Solomon’s arrival for Bielefeld, returning to Berkeley only in 1978. In her absence, Main instructed Solomon to learn how to classify the Strange Situation guided by Donna Weston and by feedback that Main sent by post.127 Solomon began to compile detailed notes on cases she found difficult to classify, for discussion with Main on her return. At the same time, Solomon began to study a sample of maltreated infants in the Strange Situation with Carol George—a research project that was ultimately not completed or published. In coding these two samples, Solomon observed a variety of behaviours discrepant with the Ainsworth coding protocols, though they were particularly common in the maltreated sample: apparent signs of depression in infants; indications that an infant was attempting to muster a coherent strategy of approach or avoidance but failing to achieve this; infants initially approaching the caregiver but then veering off; and disoriented behaviours (e.g. the child leaves its arm hanging in the air).
The first use of a ‘D’ classification in the Berkeley study occurred in August 1979.128 This was a case coded by Solomon. The infant’s behaviour on reunion displayed a whole variety of conflict behaviours. After attempting at length to work out a best-fit Ainsworth classification, Solomon eventually gave up: ‘Well, it is not A and it is not B and it is not C. I’m going to call it D.’ Solomon’s first use of the ‘D’ label was in pique. However, from 1979 Solomon and Main began thinking about and discussing a ‘D’ category for the Strange Situation. It was in line with Hinde’s emphasis on conflict behaviours and on Main and Weston’s existing work on cases unclassifiable by the Ainsworth coding protocols. Additionally, the fact that such behaviours were more common in the maltreatment sample Solomon was also coding spurred interest in their meaning.
Though they encased their observations within a category, Main and Solomon did not assume that the various forms of discrepant behaviour all meant the same thing. They were thinking in terms of the work by ethologists on conflict behaviour (Chapter 1):
Based on observations across various species, Hinde articulated very clearly the different conditions that would lead to different forms of conflict behaviour. It would undoubtedly have been possible for Main and Solomon to develop scales for conflicted, confused, or apprehensive infant behaviours. The use of differentiated scales within an overarching category was precisely the solution that Main’s research group later adopted in the 1990s when studying parental behaviour.130 That some infants scored on more than one scale would hardly have been a problem. Yet in the 1980s, in a field of empirical inquiry grounded on Ainsworth’s Strange Situation and her patterns of attachment, categories were currency; to a large extent, categories formed a horizon of how data could readily be conceptualised and discussed at the time by attachment researchers. The publication of the influential DSM-III in 1983 (Chapter 1), as a clinical system based on categories, may also have played a distal role in supporting category-based thinking.
In characterising conflict behaviours seen in the Strange Situation as a category, Main and Solomon sought to support their attempts to win attention to a phenomenon—or interrelated phenomena—that they saw as important, despite obstacles to such recognition. The obstacles were substantial. There was intense resistance to making any change to the Ainsworth coding system for the Strange Situation. Additionally, many of the behaviours lasted only a few seconds, and could easily be discounted as simply babies doing odd things because they are figuring out motor coordination. To gather more data and to support a claim to wider relevance, Main’s laboratory began to collect unclassifiable tapes from other researchers working with high-risk samples such as Mary J. O’Connor, Elizabeth Carlson, Leila Beckwith, and Susan Spieker. Drawing on an analysis of both the Berkeley Strange Situations and these additional recordings from other samples, in the winter of 1982 Main and Solomon began work on what would be their 1986 chapter announcing ‘discovery of a new, insecure-disorganised/disoriented attachment pattern’. This represented a shift in terminology. In initial discussions with Solomon, Main used the terms ‘disordered’ and ‘disoriented’ as overarching labels for the forms of conflict behaviour, following on from the use of these terms in the coding system for the Clown study. However, the term ‘disordered’ was judged to sound pejorative. It was not clear what the different behaviours meant, and so it was regarded as premature to describe them as a mental disorder.
In looking for a more descriptive word, the term ‘disorganised’ was available to Main and Solomon. The term had entered common use in neurology in the 1940s and 1950s to refer to the potential for strong feelings to be experienced as overwhelming. Affects such as anxiety, anger, awe, and ecstasy could be so intense and absorbing they make a person lost in the affect and unable to respond to the cues of the situation they are in.131 The neurological literature used ‘disorganised’ to refer to the state following such intense and absorbing affects. It essentially meant ‘overcome’ by the feeling. Bowlby introduced the term into attachment theory in his 1960 paper on ‘Separation anxiety’, in order to propose that, whether in childhood or adulthood, grief could also be a disorganising affect. Having emphasised the value of the concept of ‘disorganization’, Bowlby then promised that ‘this is a concept to which we shall be returning in a paper to follow’.132 The promise was left unfulfilled, eliciting letters from readers requesting more detail about this idea of ‘disorganisation’ and why Bowlby thought it so important.133 In Attachment, Volume 1, Bowlby returned to the issue of disorganisation. He restated his earlier claim that behaviour can become uncoordinated in the context of certain intense emotions: ‘Above a certain level, however, efficiency may be diminished; and, when in an experimental situation total stimulation is very greatly increased, behaviour becomes completely disorganised.’134 Here he added, however, that such overwhelming intensity is specifically expected in the context of conflicts between strong motivational systems, and ‘in some cases, indeed, the behaviour that results when two incompatible behavioural systems are active simultaneously is of a kind that suggests pathology’.135 He then reviewed Hinde’s discussion of the various forms of ‘conflict behaviour’ seen in animals when two competing behavioural tendencies were activated. These included rapid transition between one tendency and the other, poorly coordinated combinations, freezing in place, misdirected movements, or signs of confusion or tension.136
Main’s colleagues at Berkeley, Block and Block, also drew on this term ‘disorganisation’ from neurology. They used the word to mean ‘immobilised, rigidly repetitive or behaviourally diffuse’ flooding behaviours, which could be expected when a child was experiencing ‘a difficulty in recouping’ in the face of behavioural conflict and distress.137 The concept of disorganisation may have also appealed to Main and Solomon in the context of a technical use of the term ‘organisation’ that had sprung up following Ainsworth, Sroufe, and Waters (Chapters 2 and 4), who had cut the term ‘organisation’ loose from its ordinary language meaning and given it a technical sense: behaviour coordinated to achieving the set-goal of the attachment system. Behaviour that did not seem oriented to the achievement of this set-goal was therefore described by Ainsworth as ‘disorganised’ as early as 1968 in correspondence138 and 1972 in print.139 These were, importantly, the exact years in which Main was Ainsworth’s doctoral student. What the different forms of conflict behaviour observed by Main and Solomon had in common was that they did not seem like components of a coherent and coordinated attempt to achieve proximity. Avoidance could be interpreted as ‘in the service of proximity’, and therefore oriented to the achievement of the set-goal of the attachment system. By contrast, conflicted, confused, or apprehensive behaviours did not seem ‘organised’ to Main, in the technical sense of the word.140 It may well have seemed logical, therefore, to call them disorganised.
At a four-day workshop in 1985 at the University of Washington, Ainsworth sat on the floor to be as close as possible to the screen as Main showed her tapes coded with the new disorganised classification. At the end of the event, Ainsworth wrote to Bowlby that she and ‘everyone there was most impressed with the need for adding a new “D” or disorganised category to the classification system’.141 Bowlby also became convinced that the behaviours identified by Main and Solomon were likely ‘of great clinical concern’, and he expressed pride in their work.142 However, in the margins of his personal copy of Main and Solomon’s 1986 chapter ‘Discovery of an insecure-disorganized/disoriented attachment pattern’, he wrote that the authors would have done better to call it a ‘status’ because the unitary term ‘pattern’ may result in confusion if readers interpret it in the Ainsworth sense. In this marginalia, he observed that Main would likely agree with this reasoning, since she had indicated to him in a discussion in March 1986 that, in her view, ‘trauma to the attachment system causes disorganisation of behaviour but does not create a new category’.143
However, one key advantage of making a new classification was that it cleaned up the existing categories: many of the children who could now be classified as disorganised in at-risk samples had previously been classified as ‘secure’ because, despite manifest displays of conflict or confusion, they had protested the departure of their caregiver and been comforted on reunion. This was a particular problem, for instance, in the influential Minnesota Longitudinal Study of Risk and Adaptation (Chapter 4) and in clinical samples.144 Ainsworth had also observed conflict behaviours in the samples reported in Patterns of Attachment, and had even entered them into the coding protocol as characteristic of the B4 subtype (Chapter 2). Main was able to demonstrate in Washington that conflicted, confused, or apprehensive behaviours were not limited to this subtype, but could be found across the Ainsworth patterns of attachment. In fact, Main could show examples of Strange Situation procedures where forms of conflict, confusion, or apprehension so disrupted reunions that it was the dominant response. As she wrote in a paper composed in 1986, the extent of conflict behaviour was in theory on a spectrum. In a small number of cases, though, the conflict was so pervasive that it drowned out other responses: ‘disorganization operates as a category only in extreme cases, being otherwise … a dimension’.145
By the time of their 1990 chapter offering protocols for coding the new category, Main and Solomon had closely analysed 100 recordings of infants from low-risk samples and 100 recordings from high-risk samples.146 They proposed certain infant behaviours to be indicative of a ‘disorganised’ attachment response. They clustered the identified behaviours into seven indices based on how they appeared:
Sequential displays of contradictory behaviour; Simultaneous display of contradictory behaviour; Undirected, misdirected or incomplete movements; Stereotypies, mistimed movements and anomalous postures; Display of apprehension of the caregiver; Overt signs of disorientation.
To facilitate coding, Main and Solomon presented general guidelines and a nine-point scale.147 This scale is partly a measure of the extent of inferred disorganisation of the attachment system, and partly a ranking of how certain a coder is that they are seeing behaviour suggesting disruption of the attachment behavioural system. A score above 5 is sufficient for placement of the dyad into a D classification.148
As Main and Solomon acknowledged, behaviours pertaining to the first five indices were already discussed by Hinde and Bowlby as classic ‘conflict behaviours’ (Chapter 1). Main and Solomon introduced two further kinds of behaviour, based on their analysis of the recordings: apprehension directed towards the caregiver, and disorientation or confusion on reunion or in proximity with the caregiver. They knew that apprehension did not really fit under the term ‘disorganisation’, since the behaviour could well be behaviourally coherent and coordinated. However, at a more abstract level of analysis, apprehension of the caregiver nonetheless seemed in conflict with proximity, considered as the set-goal of the attachment system.149 Unlike the Ainsworth patterns which tended to be more discrete, Main and Solomon found that infants who showed one kind of conflict, apprehension, or disorientation also regularly showed some other kind. This led them to regard the phenomena as highly related, even if not necessarily all of a piece. As a result, the different behaviours were grouped together as a ‘disorganised’ category.
Yet their use of the term ‘disorganised’ differed from the dictionary definition. And they did not clarify this for the reader at the time, something both authors now regret. The dictionary, everyday meaning of the term ‘disorganization’ suggests randomness and a lack of predictable responsiveness to contingencies: ‘to destroy the system or order of; to throw into confusion or disorder’.150 According to the connotations of the word in ordinary language, what Main and Solomon appeared to be proposing was a category of undifferentiated chaos. The scientific concept was taken hostage by its ordinary language connotations.151 In Spangler and Schieche, for instance, the authors wrote that ‘as disorganized infants, by definition, do not have any coherent strategies, behavioral regulation is restricted or even not possible at all’.152 Likewise, Wanaza and colleagues wrote that ‘disorganization is defined as the collapse of attachment strategy under conditions of stress; under such conditions, disorganized individuals select a set of behaviors that are irrelevant to their need for downregulation of discomfort’.153 Commentators at the research–practice interface have criticised the idea of attachment disorganisation as clinically evocative but unhelpful, since the concept means incomprehensibility and therefore offers no clues to clinicians about how to proceed.154
In fact, the term ‘disorganisation’ was used in three ways in Main and Solomon’s 1986 and 1990 chapters—not one of which aligns well with the dictionary definition. The term was used by Main and Solomon to describe (i) observable behaviour that seemed to lack order or relevance to achieving proximity with the caregiver; (ii) a disruption of the attachment behavioural system, caused by past experiences of child–caregiver interaction and inferable from observed behaviour; (iii) and the label given to the category used for coding the Strange Situation.155 In using the same term ‘disorganised’ to refer to both behaviour and/or psychological process and/or the overall category, Main and Solomon had a specific aim, though it was not well articulated at the time. ‘Disorganisation’ was used as a conceptual tool for picking out ‘an observed contradiction in movement pattern, corresponding to an inferred contradiction in intention or plan’.156 The theoretical stakes of using the term ‘disorganized’ to mean both behaviour and psychological process was the claim that the diverse behaviours picked up by the Main and Solomon indices could well have different antecedents and sequelae, but what they had in common was that they suggested disruption or breakdown at the level of the attachment system. This was the rationale for the ‘radical notion that the many, highly diverse indices of disorganization and disorientation can be placed under one heading’.157
In the early 1990s, Main created a revised version of the Main and Solomon coding protocols, though she did not publish it. This version is distributed to trainees at the Minnesota Strange Situation training institute, run by Alan Sroufe and Betty Carlson.158 Main’s revisions predominantly entailed adding further forms of behaviour under the seven indices. For instance ‘overbright greeting’ was added to Direct Indices of Apprehension (Index VI), presumably as this was interpreted as an early form of caregiving or compliance by the child towards the caregiver. Main also added guidance to coders on ‘Major Considerations’ for coding disorganised attachment. The primary consideration was ‘Is the behaviour inexplicable (no evidence of immediate goal or rationale) OR is the behaviour explicable only if we presume: a) The baby is afraid of the parent; b) The baby is inhibited from approach without being able to shift attention to the environment?’159 Here we can see that disorganisation is operationalised as behaviours that seem confused, apprehensive, or conflicted. Also visible is the technical distinction for Main between avoidance and disorganised attachment, hinging on the fact that the former can direct attention to the environment in order to maintain regulation, whereas in the latter this conditional strategy of the use of attention to circumvent approach/avoidance conflict is not feasible. However, this manuscript remained unpublished, and generally available only to junior researchers who are tasked with coding. With exceptions, many senior researchers in the field of attachment research have therefore not known how the category is actually coded, for instance that only some infants will look apprehensive—others will look confused or conflicted (or a combination of the three). Senior researchers have tended to be unaware of the significant role coders must give to inferences about how the infant is directing attention in the Strange Situation.
In putting confused, apprehensive, and conflicted behaviours together, Main and Solomon did not intend to imply that they all meant the same thing. In fact, they stated in print that ‘our discovery of the D category of infant Strange Situation behaviour rested on an unwillingness to adopt the “essentialist” or “realist” position regarding the classification of human relationships’.160 Their epistemological position, like Bowlby’s, treated categories as ‘provisional, albeit nonarbitrary approximations’,161 and Main and Solomon end their 1990 chapter with an extended criticism of essentialist approaches to categorisation in the history of biology. Yet these remarks by Main and Solomon on the dangers of essentialism have never been cited or discussed. In general, the ensuing literature respectfully cites but gives little evidence of having directly read the book chapters introducing the classification. Main and Solomon were widely interpreted as introducing an exhaustive addition to the Ainsworth system. They have therefore been accused of being bent upon ‘reducing complex human experience to typologies’.162 Likewise, it has been assumed that their category aimed simply to soak up possible variation in human behaviour beyond the Ainsworth patterns and treat it all as evidence of dysfunction.
For instance, Gaskin argued that ‘the category is really just a residual one’: rather than itself designating any meaningful phenomena, the existence of the classification ‘might be seen more productively as evidence of the inadequacy of the three attachment classifications’ introduced by Ainsworth on the basis of her Baltimore middle-class sample.163 This misunderstanding has a history that goes back to the very introduction of the classification. Mark Cummings was one of the editors of the volume within which Main and Solomon’s 1990 chapter was published. In his contribution to the volume, he argued that ‘deviations from expected sequences do not constitute a sufficient criterion for classification’.164 Against what he took to be Main and Solomon’s perspective, he proposed that D behaviours could not all be expected to reflect the same process of breakdown of ‘general functioning’, and that therefore the category lacked coherence and meaning.165 A better criterion for disorganisation, Cummings argued, would be behaviours that do not appear to function to achieve ‘felt security’, which Sroufe and Waters had proposed, as the set-goal of the infant’s attachment system (Chapter 4).166
Psychological constructs are usually subjected to statistical analysis to see which elements cluster together and which among these clusters are especially involved in associations with other variables of interest. By contrast, though it is not uncommon to hear praise of the ‘strong psychometric properties’ of the Strange Situation procedure and its classification as a whole,167 in fact no psychometric study has ever been reported on the infant disorganised attachment classification. Though perhaps less in social psychology than in developmental psychology (Chapter 5), psychological scientists are generally satisfied to work with instruments that have expectable correlates and fit with a conceptual framework, no matter that psychometric work is incomplete. Furthermore, psychometric work receives little professional reward for psychological researchers. Another potential obstacle is that a sample of adequate size for psychometric research is needed—though in the case of disorganised attachment this cannot have been the only factor in play, since a number of such samples have long been available now. Rather, it was simply assumed that disorganisation was best treated as a single category.
No assessment was made, for instance, of whether the one-to-nine scale would be more serviceable than the category. And no exploration was made of the component parts of disorganisation, for instance to economise the coding system. The underdeveloped and rather complicated coding protocol made application of the classification time-consuming and, beyond this, very hard to learn. The result has been a significant bottleneck for research using the Strange Situation due to the small number of reliable coders and the disincentives of investing time and resource in learning the system. It has also limited the circulation of important practical knowledge, leading to frequent speculative discussions about disorganisation among attachment researchers that have been wholly cut loose from how it is actually coded.168 Yet prediction of negative outcomes was treated by researchers—and subsequently by clinicians and policy-makers—as validity. Indeed, the disorganised classification also quickly bore fruit in predictive significance.
The short-term test-retest stability of disorganised attachment as assessed twice using the Strange Situation was r = .35. This was regarded as at least adequate to validate the measure, and potentially rather high considering that the classification was, at times, made on the basis of quite fleeting behaviours169—though stability between infancy and toddlerhood or preschool has been found to be much lower.170 Incidence of the disorganised attachment classification was also discovered to be more common in clinical samples, samples known to social services, and in samples facing multiple adversities, suggesting that it reflected some adverse experience or process. In community samples with relatively few adversities, around 15% of infants show a sufficiently high degree of confused, disoriented, or apprehensive behaviours towards their caregiver in the Strange Situation for a disorganised classification to be assigned. However, this increases to a majority of infants from families drawn from samples facing multiple compounding adversities, or where maltreatment of the child has been documented.171 Of particular importance for acceptance of the new classification, however, was the prediction of conduct problems. An early and influential study by Lyons-Ruth and colleagues, reported in 1991, found disorganised infant attachment associated with aggressive behaviour in preschool in a sample of dyads known to child protection services.172
The link between the infant disorganised attachment classification and later aggressive behaviour has been confirmed by numerous studies. By 2010, Fearon and colleagues could report from a meta-analysis of 69 studies a modest but material association (d = .34) between disorganisation and various externalising behaviours (including attentional problems).173 The association was stronger in samples where families were known to clinical or professional services (d = .43) or facing low socioeconomic status (d = .44). Indeed, for many samples, disorganised infant attachment only predicted later externalising problems in conjunction with other forms of adversity.174 For disorganised attachment, as for the Ainsworth avoidant classification (Chapter 2), the Strange Situation conducted with two-year-olds seemed to predict later externalising problems more effectively than when it was conducted with infants, perhaps because one of the particular developmental challenges of this age is learning to regulate frustration in relationships.175 Researchers have also documented other sequelae, including small-to-moderate associations with social competence and friendships (d = .25).176
A meta-analysis by Groh and colleagues found no association with later depression or anxiety.177 But other mental health outcomes have been documented. For instance, a classification of disorganised/disoriented attachment in infancy predicts severity of later post-traumatic stress disorder (PTSD) symptoms following a trauma—an association that was found not to be attributable to the many co-occurring risk factors.178 In her landmark 1998 paper, Carlson also documented associations in the Minnesota Longitudinal Study of Risk and Adaptation between infant disorganised attachment and general mental health in early adulthood, with a particularly strong association for dissociative symptoms.179 The Carlson paper helped secure perceptions of disorganisation as, in Sroufe’s words, an ‘incredibly powerful construct, being among other things the strongest single predictor of later psychopathology’.180 Yet Groh, Fearon, van IJzendoorn, Bakermans-Kranenburg, and Roisman have presented meta-analytic findings inconsistent with Carlson’s apparent emphasis on the implications of infant disorganised attachment for general psychopathology, finding instead stronger links with externalising problems specifically.181 The Minnesota findings are discussed further in Chapter 4.
The Main and Solomon 1990 chapter laid out protocols for coding the classification. Each of the seven Index headings was followed by a variety of concrete examples of kinds of behaviour seen by Main and Solomon in their review. Some of the examples are italicised, which means that behaviour of this sort may, on its own, be sufficient for a D classification. Other examples are not italicised, which means that only when other kinds of conflicted, confused, or apprehensive behaviour is present should a D classification be considered. The rationale was not made explicit by Main and Solomon for the inclusion of particular examples, or for which examples are italicised, or how to weight observations that differ from the examples. This has contributed to the bottleneck in reliable coders. A careful retrospective examination of the protocol, conducted together with Solomon, reveals that behaviours are included when they are conflicted, confused, or apprehensive. And, again in retrospective examination, italicisation was based on six factors: (i) frequency of a behaviour, (ii) its pervasiveness or duration, (iii) its abruptness in behavioural sequence, (iv) the extent to which it occurs either close to reunion or in physical proximity with the caregiver, (v) whether it can be better explained as a reaction to the immediate environment, and (vi) the extent to which the infants’ responses to their caregiver suggest the experience of apprehension.182
This latter item in the protocol reflects a theory that Main had been developing with her husband and collaborator Erik Hesse. Hesse entered undergraduate study in psychology at Berkeley as an adult, after studying to become a professional musician. Following completion of his studies in 1981, he worked as a research collaborator with Main, for instance as one of the coders for the Kaplan drawing system (discussed in the section ‘The Kaplan and Main family drawing system’). At the start of her relationship with Hesse, Main had been thinking intensively about the role of avoidant behaviour in the Strange Situation. She concluded that a rejecting or angry parent would distress an infant, activating the attachment system and a desire for proximity, but simultaneously would evoke a tendency to withdraw from the parent. Avoidant behaviour would circumvent the ensuing approach–avoidance conflict.183
Hesse’s ingenious proposal was that one cause of the conflicted, confused, and apprehensive behaviours that Main and Solomon were seeing in the Strange Situation would be experiences of alarming behaviour by the caregiver, the haven of safety sought by the attachment system. This would not permit the redirection of attention that, for Main, provided the infrastructure for a conditional strategy: ‘the infant may at times be experiencing a fear or distress too intense to be deactivated through a shift in attention (the Ainsworth A pattern), yet at least momentarily cannot be ameliorated through approach to the attachment figure (the Ainsworth B and C patterns)’.184 This predicament could be produced if ‘the fear the infant experiences stems from the parent as its source’, due to the injunction of the attachment system to approach the caregiver when alarmed.185 Alarming behaviour by an attachment figure would cause an injunction to both approach and take flight, blocking the redirection of attention upon or away from the caregiver that underpins the two forms of infant conditional attachment strategy. Main and Hesse put forward this idea in a 1990 chapter that immediately followed the Main and Solomon protocols, establishing it essentially and fundamentally within reception of the D classification. In their writings of the period, and in personal communication with colleagues, they proposed that ‘it is the interjection of fear into the caregiving experience that is essential to developing a disorganised/disoriented attachment’.186
Supporting evidence for association of the caregiver with fear as one pathway to disorganised infant attachment was already available by the time of Main and Hesse’s 1990 chapter, from Dante Cicchetti and his research group (Chapter 4). The Harvard Child Maltreatment Project was initiated in 1979 by Cicchetti and Aber, as a large longitudinal study, modelled on the Minnesota Longitudinal Study of Risk and Adaptation.187 A small subsample were seen in the Strange Situation and classified using the Ainsworth categories.188 Twenty-two dyads were known to child protection services for abuse and/or neglect of the child or an older sibling; 21 were a matched comparison group. However, a third of the sample were classified as securely attached according to the Ainsworth coding protocols. There were a few different proposals available at the time for additions to Ainsworth’s system for use with clinical samples. One was that of Ainsworth’s student Patricia Crittenden, who observed that some abused infants show both avoidant and resistant attachment (A–C).189 However, the Main and Solomon disorganised attachment classification appealed to Cicchetti and colleagues as more encompassing; it was assumed that finer-grained distinctions would be worked out later. The disorganised classification also had emerging evidence of predictive validity.190 The data were recoded for disorganisation by Cicchetti and Barnett; Barnett was blind to the children’s maltreatment status; 82% of the infant–caregiver dyads known to child protection services were classified as showing disorganised attachment and 14% classified as secure; in the comparison group, 19% were classified as disorganised and 52% classified as secure. At 18 months, 61% of dyads known to child protection services were classified as showing disorganised attachment, compared to 29% in the comparison group.191
In the empirical developmental attachment tradition initiated by Ainsworth, it was widely held that home observations were needed to validate new measures or the adaptation of existing measures. By the early 1990s, Ainsworth was expressing her dismay that ‘so many attachment researchers have gone on to do research with the Strange Situation rather than looking at what happens in the home … in part it has to do with the “publish or perish” realities of academic life’.192 And Sroufe and Waters agreed that ‘the strange situation is used only because it can stand in place of attachment as it would be observed in the home’.193 As such, ‘any procedure claiming to assess attachment (even variations or new applications of the Ainsworth procedure) must be anchored to observations of the attachment–exploration balance in the natural environment. That is the crucial criterion.’194 Though the Strange Situation is often described as the ‘gold standard’, Sroufe and Waters characterised detailed naturalistic observations as the true ‘standard’ for assessing attachment. Cicchetti and colleagues revealed one kind of experience in the home environment that could substantially increase the likelihood of disorganised attachment. Yet Main and colleagues regarded it as implausible that in the low-risk Bay Area sample, all infants who received a disorganised attachment classification had received maltreatment from their caregiver. And, as mentioned in the section ‘A new category’, Main and Solomon found that apprehension of the caregiver and disorientation were more common in the tapes they reviewed from maltreatment or very high-risk samples, whereas these behaviours were uncommon in the infants classified as disorganised in the Bay Area sample.
Conversations between Main and Hesse led to the conclusion that there could be other forms of alarming caregiver behaviour besides abuse. Main’s written notes from home observations during her doctorate offered too little detail to be the basis for a published study, and her group had not collected new observations of the child’s behaviour at home for the Berkeley study. So she was not in a position to examine how disorganised attachment in the Strange Situation related to a child’s behaviour at home in samples not specifically known to child protection services. Ainsworth regarded a home observation study as ‘highly desirable’ for the validation of the classification.195 Likewise, van IJzendoorn alleged that ‘since the D category has not been widely validated against home behavior, its meaning is not yet fully clear’.196 To answer such concerns, Main, though she had no home observation recordings, did have observations of infant free-play sessions in the laboratory with both the child’s mother and father. If alarming behaviour in playful and other more ordinary interactions was shown predominantly by parents in dyads classified as disorganised in the Strange Situation, then this would offer one form of ecological validation for the classification, and at the same time help shed light on its cause or causes.
From the early 1990s, Hesse undertook doctoral research at Leiden University, supervised by van IJzendoorn. As part of this research, Hesse and Main began to formulate a coding system for alarming caregiver behaviours, based initially on theory and then on review with their graduate students of 13 observations from the Berkeley sample of infant–mother or infant–father interaction during unstructured play from before the Clown session.197 Many of the potentially alarming behaviours they observed were fleeting; the coding system developed by Hesse and Main to identify and capture these moments is penetrating, subtle, and extremely difficult to learn (even on the rare occasions on which training has been offered). It was dubbed the ‘FR coding system’, named after two central components: frightening and frightened caregiver behaviours. Examples of frightening behaviour provided by the coding manual were directly threatening or predatory behaviours, such as bared teeth, or lunging or looming at the infant without markers of play or prior warning. Merely angry behaviours by the caregiver, or ordinary forms of ‘culturally sanctioned’ discipline, were not included.198 Frightened behaviours shown by the caregiver towards the infant were also identified, such as sudden withdrawal from the baby or a strong startle response in response to the infant’s approach. Such displays were anticipated to alarm the infant when they sought the caregiver as a safe haven. A third kind of behaviour identified by Hesse and Main was apparently dissociative behaviours by the caregiver, which were assumed to potentially frighten an infant through the caregiver’s inexplicable attentional unavailability.199 The role of dissociative caregiver behaviours was proposed to Main and Hesse by their friend Giovanni Liotti, whom they had met at Bowlby’s 80th birthday party in 1987 and then visited in Rome in 1990.200
In the course of the 1990s, Hesse and Main continued to scope potentially alarming caregiving behaviours through review of 62 further Berkeley recordings of caregiver–infant interaction and also recordings drawn from four diverse samples where Strange Situation data were also available. One was Karlen Lyons-Ruth’s sample of families known to social services in Boston.201 Another was a doctoral project by their student Mary True, who had recorded mother–infant interactions among the Dogon of Uganda.202 A third sample was collected by Debby Jacobvitz in Texas; though a community sample, rates of alarming caregiver behaviours and of disorganised attachment were unusually high.203 And a fourth sample of bereaved mothers was collected by Carlo Schuengel at Leiden for his doctoral research.204 Both study of the Berkeley sample205 and Schuengel’s Leiden sample206 revealed that in fact dissociative caregiver behaviours were the most predictive of disorganised attachment, more so even than directly frightening caregiver behaviours, though the two were intercorrelated.207
By 1995, Hesse and Main’s work on the coding system had led to the addition of three further kinds of behaviour.208 These were not in themselves assumed to be directly alarming, but were held to index the caregiver’s potential for one of the three primary forms of alarming behaviour. These were: (i) behaving in a timid or deferential way toward the child;209 (ii) sexualised behaviours toward the infant, suggesting confusion between the caregiving and sexual behavioural systems as well as a lack of capacity to monitor action; and (iii) behaviours by the caregiver that are coded in the Main and Solomon indices for infant disorganisation, such as misdirected behaviours or briefly moving in a stiff, asymmetrical manner. These latter behaviours were assumed to represent some significant disruption at the level of the caregiving system and therefore serve as distal markers of alarming caregiving. In the Berkeley sample, Hesse and colleagues found that these additional forms of FR caregiving behaviour were not associated with infant disorganised attachment; however, they suspected that this may have been a function of their low-risk cohort.210
Though only a tiny number of researchers have been trained by Main and Hesse to use it, the FR coding system itself has been widely praised. For instance, Rothbaum and Morelli, major critics of Ainsworth’s work, praised the FR system as less prone to unacknowledged ethnocentric bias than Ainsworth’s sensitivity scale.211 Furthermore, empirical evidence has accumulated in support of the predicted association between infant disorganised attachment in the Strange Situation and caregiver behaviours coded with either the initial or the expanded version of the FR coding system. Madigan, Bakermans-Kranenburg, van IJzendoorn, and Moran reported from a meta-analysis that FR caregiver behaviour during observations accounts for 13% of the variance in attachment disorganisation.212 Cassidy has criticised Main and Hesse for overclaiming the importance of the frightening/frightened pathway to disorganised attachment, since in fact ‘the relation between frightening/frightened behaviour and infant disorganisation has been found to be relatively weak’.213 There is certainly much unexplained variance, suggesting the role of factors besides frightening, frightened, or dissociative caregiving. Nonetheless, these are clearly causally important: the meta-analysis by Madigan and colleagues reveals that parents who displayed these behaviours were 3.7 times more likely to be part of dyads coded for disorganised attachment in the Strange Situation (r = .34, N = 851).214
An expanded version of the FR coding system was developed by Lyons-Ruth, Bronfman, and Parsons (the Atypical Maternal Behavior Instrument for Assessment and Classification (AMBIANCE)). This assesses five dimensions of disrupted parental affective communication: negative-intrusive behaviour; role confusion; disorientation; affective communication errors; and withdrawal from the child. AMBIANCE is a more encompassing measure, with equivalent prediction of disorganised infant attachment to the FR system.215 Therefore, on the grounds of parsimony, some attachment researchers have expressed a preference for FR.216 However, training is available annually in AMBIANCE, unlike the FR system. Efforts to economise AMBIANCE measure for use in routine clinical practice may contribute to the further popularity of the measure in the future.217
An important meta-analytic finding by van IJzendoorn and colleagues was that disorganised infant attachment was only weakly predicted by caregiver insensitive behaviours in general. Van IJzendoorn and colleagues concluded that the FR pathway was therefore distinct from insensitivity.218 Subsequently, however, Bailey and colleagues as well as other researchers provided evidence suggesting that the FR pathway and insensitivity overlap in high-risk families.219 Proposals have also been put forward regarding the role of specific forms of insensitive care in disorganised attachment, including by Lyons-Ruth. Using AMBIANCE, Lyons-Ruth and collaborators identified a specific association between caregiver withdrawal from the child and disorganised attachment.220 Another pathway to disorganised attachment has been suggested by researchers such as Spangler, who have argued that infant temperament or genetic factors may predispose at least some forms of disorganised attachment.221 Main and Hesse have by and large been sceptical of such temperamental or genetic explanations and have pointed to the fact that classifications of disorganised attachment with different caregivers have little association, and that studies of temperamental or genetic antecedents have had a poor record of replication.222 However, they have been intrigued by findings by van IJzendoorn and colleagues that suggest gene × environment interactions in the prediction of disorganised attachment.223
Yet, just as the ordinary language connotations of ‘disorganisation’ misled readers of Main and Solomon, so did the evocative connotations of the term ‘fear’ magnetise readers of Main and Hesse’s chapter. Close examination of the Main and Hesse 1990 text from the vantage of the present suggests that the term ‘fear’ was used in a way that was insufficiently specified. The 1990 chapter fell subject to a danger already identified by Bowlby: ‘unfortunately in colloquial English the word “fear” is used in many senses, often being synonymous with expectant anxiety and sometimes with fright’.224 Main and Hesse meant to convey that one pathway to disorganised attachment may be when children come to associate their caregiver, for whatever reason, with feelings of alarm.225 However, as a result of the way they used the term ‘fear’, it has been widely assumed by readers that Main and Hesse were proposing that disorganised attachment represents a child afraid of their caregiver. It has been common to encounter statements even by authorities such as van IJzendoorn and Bakermans-Kranenburg referencing Main and Hesse as saying that, for instance, ‘the essence of disorganized attachment is that the child is at times scared of the attachment figure’.226
To take further examples of the misapprehension of Main and Hesse’s position, Paetzold and colleagues—the researchers in the social psychological tradition who have given most notice to disorganised attachment—stated that, according to Main and Hesse, ‘infants in the disorganized category develop a fear of their attachment figures because these figures display frightening behaviors in their daily interactions with their children’.227 Sinason wrote that ‘disorganized attachment refers to grossly disorganized behavior on the part of the infant or child: apprehension in the presence of the mother or primary caretaker’.228 And Rees wrote to fellow paediatricians that ‘disorganized patterns arise if pervasive abuse leaves children ineffective both in self-sufficiency and in using relationships, lacking understanding of their own and others’ feelings. Safe independence is unlikely and criminality in adulthood common without recovery.’229 Reification of the fear–disorganisation relationship also appeared in proposals for National Institute for Health and Care Excellence guidelines on child attachment for clinicians in the UK.230 However, following feedback from Main, Hesse, Lyons-Ruth, and others,231 these guidelines were amended in the final version to avoid implying that frightening caregiver behaviour is the primary route to disorganised attachment.232
One contributing factor to misapprehensions regarding the role of fear in disorganized attachment may have been the illustrations in Main and Solomon’s 1990 chapter.233 This comes into relief if the illustrations in the original manuscript and the published chapter are compared. In both versions, pen drawings were presented of infants showing behaviour listed in the coding protocols. The drawings were tracings of the film negatives of video recordings of infants in the Strange Situation. In the original manuscript, the illustrations are of a variety of the behaviours. In the published version, the only set of illustrations is of explicit apprehension seen on reunion with the caregiver. There is a drawing of a child covering her mouth on reunion with her caregiver, which might also suggest confusion, but the drawing and Main and Solomon’s commentary suggest that the behaviour should be interpreted as indicating apprehension. In general, the facial expressions in the drawings convey terror and a crumpled misery. Though the drawings selected for inclusion in the published version are somewhat higher-quality drawings (less blotchy, more human-looking), their selection was also partly shaped by theory. They align much better with the Main and Hesse hypothesis that fear is implicated in at least some forms of disorganised attachment. Illustrations of visible fear of the caregiver served as a powerful encapsulation and expression of the process theorised to be underlying the diversity of disorganised behaviours. However, a disadvantage of privileging of illustrations of fear was that it was unrepresentative. There are a variety of behaviours used to classify disorganised attachment, including conflicted and confused behaviours without apparent apprehension of the caregiver.234 And, in fact, visible apprehension of the caregiver in the manner of the drawings is very uncommon in community samples, and relatively rare even in samples facing significant adversities.235 Incidence of clear apprehension of the caregiver in maltreatment samples has not been reported but is an outstanding question of the utmost significance, not least in terms of understanding the potential for measurement variance across samples among dyads who receive a disorganised attachment classification. As Padrón, Carlson, and Sroufe indicated, if one group of children show predominantly outright apprehension of the caregiver and another group predominantly non-frightened behaviours in the Main and Solomon indices, both might receive a disorganised attachment classification but have very different correlates.236
The choice of illustrations by Main and Solomon may well have had particular importance for the imagination of the disorganised infant by the readers of their chapter.237 The image of fear on reunion is central, for example, to explanations of the category by researchers in papers,238 discussion of disorganisation in textbooks for psychology students,239 and guidance provided for social workers and other child-safeguarding practitioners in using disorganised attachment as an indicator of child maltreatment.240 Waters criticised Main and Hesse for conveying an impression of disorganised attachment behaviours as always caused by fear, and all in the same way, blocking attention to the potential diversity of aetiological factors.241 Lieberman criticised Main and Hesse on related grounds: for lack of clarity regarding whether disorganised attachment is essentially an expression of PTSD in children or an independent (or overlapping) construct.242 And Bakermans-Kranenburg and van IJzendoorn drew attention to the potential ‘heterogeneity of the mechanisms leading to disorganized attachment’.243 In a 2006 article, Hesse and Main stated their wish that they had made it clearer that they intended their emphasis on frightened or frightening caregiver behaviour as ‘one highly specific and sufficient, but not necessary, pathway to D attachment status’.244 In this paper, Hesse and Main acknowledged that how their earlier work was framed and argued appears to have misled readers.245
When participants in the Berkeley study were six years old, 45 families were invited back to the laboratory, stratified by infant attachment classification. Whilst fathers and mothers went to be interviewed, the children were asked to do a drawing of their family. They were then asked to consider six pictured parent–child separations and offer their thoughts on what the pictured child might feel and do. They played in a sandbox for a quarter of an hour, before their parents returned—counterbalanced so that some fathers returned first and some mothers. In the early 1980s, the Berkeley group developed coding systems for these observational assessments. A coding system for the family drawings was developed by Main and her doctoral student Nancy Kaplan.246 The response to the pictures of separation was analysed with an adaptation of a coding system developed by Klagsbrun and Bowlby.247 And a coding system was developed for the reunion episodes by Cassidy and Main.248 A coding system specifically for the verbal interaction between child and caregiver on reunion was also developed by Amy Strage, a research associate in the group between 1983 and 1985.249
Findings from these analyses and those with the AAI were reported at a dedicated symposium in 1985 at the Society for Research in Child Development held in Toronto, and a summary published as the paper ‘Security in infancy, childhood, and adulthood: a move to the level of representation’ in a monograph for the Society, edited by Everett Waters and Inge Bretherton. New classifications for coding a mother’s autobiographical discourse in the AAI had a strong correlation (r = .62) with the classification of that parent–child dyad in the Strange Situation five years earlier, and a marked correlation also for fathers (r = .37).250 The prediction from infant attachment to responses at age six was an astonishing 68% to 85% in the case of the mother–child dyads. In the case of the father–child dyads, prediction from the Strange Situation to reunion behaviour was equivalent. But prediction was weak or non-significant for the other tasks. Main and colleagues supposed that, in contrast to a direct reunion with the parent, these latter measures were less effective at priming mental processes specific to the relationship with a particular parent. Their interpretation was that since almost all of the children in the subsample had their mother as their most familiar and primary caregiver, undifferentiated attachment primes such as a family drawing made children think about their mother rather than, or more than, for father.
In the early 1980s, even close collaborators and friends of Bowlby acknowledged that an inherent limitation of attachment research was that there was no way of measuring attachment after infancy.251 Its relevance to other questions and concerns, such as therapeutic work with adults, seemed limited. Main and colleagues appeared to conclusively demonstrate that this was not so. Certainly the attachment behavioural system would change very dramatically with maturation. However, having conceptualised attentional processes as the underpinning architecture of individual differences in the Strange Situation, Main was in a position to extrapolate upwards and examine how attention was directed in relation to attachment figures in other contexts. For her, the Strange Situation procedure, children’s drawings and discourse, and adult autobiographical discourse all had in common not any specific behaviour but a question about the direction of attention in relation to attachment figures. In ‘A move to the level of representation’, Main stated that across the different measures:
To situate and legitimate her ideas within the tradition of Bowlby’s theory, on the first page of the ‘A move to the level of representation’ paper, Main described the individual differences characterised by her measures as based on differences in ‘internal working models’.253 She defined these, however, in a technical way to mean biases and filters in the processing of attachment-relevant information: ‘We define the internal working model of attachment as a set of conscious and/or unconscious rules for the organization of information relevant to attachment and for obtaining or limiting access to that information, that is, to information regarding attachment-related experiences, feelings, and ideation’.’254 In terms of positioning within the academic field, claiming that the AAI assessed Bowlby’s classic concept of ‘internal working models’ had clear symbolic advantages for gaining recognition and acceptance of Main’s work.255 However, the ‘internal working model’ was a problematic concept. Main did not mark for the reader that her technical definition departed from the way that the term had been used in Bowlby’s work. There it could be used to mean almost anything cognitive. In particular, it could sometimes refer to mental representations of specific caregivers and sometimes to any cognitive process relevant to attachment (Chapter 1). One unfortunate result was that readers were led away from the fundamental logic through which Main extrapolated from the Strange Situation to coding systems for later in the lifespan, making this leap seem mysterious, and making the development of precise hypotheses for testing more difficult.256
In fact, for Main, the common thread linking the Strange Situation with her group’s new measures was the ‘rules for the direction and organisation of attention’.257 Whereas in traditional psychoanalytic theory, defensive processes operate on mental content, in Main’s account, attentional processes are used to exclude information that might result in patterns of attention and behaviour anticipated to be unhelpful based on representations of past experiences. Whilst representations might guide the direction of attention, for Main it was attentional processes that were ultimately what animated individual differences. So, for instance:
The introduction of multiple new assessment methods was reported in the same chapter, alongside new reflections on Bowlby’s concept of the ‘internal working model’. As a consequence, all the explanations were highly compacted and their theoretical underpinnings are only sketched. Details of the six-year assessments remained obscure—with some measures, such as the family drawing, not reported at all. In the years after the ‘A move to the level of representation’ paper, these further details were not forthcoming. Kaplan did not publish the results from her dissertation. Strage did not write up her work. Cassidy published an analysis of the results of her dissertation work259 but did not publish the coding system. The adaptation of the Separation Anxiety Test by the Berkeley group was also left unpublished, though it influenced work by Bretherton, Cassidy, George, and Solomon, and others on story-stem and doll-play approaches to the assessment of attachment, which have since flourished.260 In 1986, Main completed a book manuscript with details of each of the coding systems, which was due to be published by Cambridge University Press. However, Main’s sense that the systems were not yet adequately finished contributed to delays in the book’s publication. One factor may have been that very few ambivalent/resistant dyads were available for the six-year follow-up, so the ‘maximising’ aspects of the coding systems was relatively underdeveloped. Main may also have had concerns that the coding system was ‘overfitted’ to her Berkeley sample, and wanted to see replications and take the chance to amend the systems before making them publicly available. Ultimately the book remained unpublished, though it circulated to several colleagues in manuscript form.
Over the decades, only a handful of individuals were ever trained in any of the six-year measures by Main.261 Main also did not delegate the matter of training to any colleagues—until July 2019, when Naomi Gribneau Bahm, Kazuko Behrens, Anne Rifkin-Graboi, and Deborah Jacobvitz received certification to deliver trainings in the Main and Cassidy reunion system. (The same researchers will also be trained and certified to deliver training in the FR coding system from summer 2020.)262 However, in the meantime, the lack of trained coders produced a gap in the availability of consecrated measures of attachment. For school-age children, alternatives such as the Manchester Child Attachment Story Task and the Child Attachment Interview have been developed.263 Nonetheless, one consequence of the de facto unavailability of the Berkeley six-year measures has been that, with few studies being conducted, over time publications that have used these measures have become regarded as rather niche.264 A second consequence has been that what research has been conducted with the Berkeley six-year measures has almost always occurred ‘off the books’, without training or a standardised test to confirm reliability against other research groups.265 Researchers have been forced to proceed on the basis of the scanty information available in the public domain. A rather sad case is the work of Dallaire and colleagues, who describe their attempt to reconstruct a coding manual for the family drawing coding system, having carefully scoured the little information made publicly available by Main’s research group.266 Finally, a third consequence has been that new measures to address this gap have emerged only slowly, since they had to contend with the position of Main’s six-year system as the orthodox and authoritative approach to assessment; as in the case of the Manchester Child Attachment Story Task, often researchers sought validation of new assessments against the Berkeley measures, treated as the gold standard.267 They also often emulated the Berkeley four-category model. Lyons-Ruth, Bureau, and colleagues have been especially active in the development of observational measures that offer equivalent categories to those identified by Main.268 However, it is telling that they have tended to develop measures for alternative ages, such as middle childhood and seven- to eight-year-olds. This would appear partly motivated by a desire to make assessments available for these age groups, but also, implicitly, to avoid the appearance of competition with Main’s six-year measures, which remain widely perceived as the ‘gold standard’ even if no training has been available and few studies conducted over the past decades.
Access to two unpublished books by Main offers a chance to flesh out the brief description of the reunion and family drawing measures in Main’s 1985 ‘A move to the level of representation’ paper. The first book is Behaviour and the Development of Representational Models of Attachment from 1986. In one of Bowlby’s last pieces of professional correspondence, composed less than two months before his death in October 1990, he wrote to his Portuguese translator Sonia Monteiro de Barros: ‘I wonder if you are following Mary Main’s recent work? … She has a book in press with Cambridge University Press which will be very remarkable.’269 This book essentially comprises the coding manuals for the Strange Situation; the AAI; the Strage and Main discourse system; the Kaplan family drawing system; and the Main and Cassidy reunion system.270 The Kaplan adaptation to the Separation Anxiety Test was not described. The second book is Four Patterns of Attachment Seen in Behaviour, Discourse and Narrative: An Abstract for Psychoanalysis from 1995.271 This book was written in honour of the psychoanalyst Joe Sandler; when he died just as the book was ready for submission, Main chose not to submit it to the publisher. The 1995 book contains a description and commentary on the measures from the 1986 book, together with a discussion of attachment and psychoanalytic theory and of epistemological aspects of Main’s work. The six-year systems, little discussed in print, help reveal the logic of Main’s approach to the more famous AAI, as well as her thinking about classification and method in developmental science. Together they offer deeper entry than the published record into Main’s approach to the measurement of attachment and the theoretical commitments that organise this approach.
Central to Main’s work in the 1980s was a new, inductive approach to the development of coding systems. Guess and uncover is a memory game for children. In the game, each player secretly writes a number pattern that follows a rule. For instance, 3, 6, 10, 15, 21, 28. They then cover all but two numbers from the right-hand-side. One player attempts to predict the third number in their friend’s pattern. If they are correct (a ‘hit’), they get to continue to guess the fourth number, and so on. If they do not guess correctly (a ‘miss’), the number is revealed, but their opponent gets to take a turn. The game uses inductive processes in the identification of a pattern, forming the basis for a deductive system that can support further extrapolation. Main used this game as a methodology for developing new coding systems. This is only mentioned in print in any detail in Main and Cassidy’s 1988 paper, and even there it appears with sufficient brevity that it was not clear how someone else might use guess and uncover in a scientific context.272 The approach has almost never been mentioned again by subsequent attachment researchers,273 who instead tend to chalk up Main’s new coding systems to her genius, solely and unreplicably.274 Yet there is no opposition between recognition of Main’s creativity and consideration of the ‘guess and uncover’ methodology used in the development of her coding systems. Furthermore, inattention to or lack of awareness of Main’s approach has stopped later researchers from either using or evaluating it.
Main first applied this methodology in guiding the work of a new, young research assistant in her group, Ruth Goldwyn. Edward Goldwyn had made a documentary film for the BBC in the early 1970s, depicting the work of Ainsworth at Baltimore.275 During his time working on the project, Goldwyn became friends with Mary Main. Some years later, when he went to Berkeley to make another film, he and his 16-year-old daughter Ruth stayed with Main. By the time she finished secondary school, Ruth had no clear sense of her future plans. Therefore, as a gap-year, in 1981 she went to Berkeley and joined Main’s group. Goldwyn was invited to read books by Bowlby and Hinde, and Darwin’s book on the origin of emotions, and she audited Main’s classes.276 She also worked with Kaplan in recruiting and collecting data in the six-year follow-up, including conducting half of the interviews with parents. Once the data were collected, however, Goldwyn was not immediately needed for any other projects. Though study of events reported in the interview with the parents was part of George’s doctoral project, George went on maternity leave after completing the interview transcriptions.277
Main was intrigued by a particular transcript: ‘Although the speaker was clearly essentially coherent and collaborative, Main noted that he went to slightly unusual lengths in describing tender, emotionally affecting aspects of his life, lingering in somewhat lengthy descriptions of his loss of a beloved family member … like a B4 infant, this speaker to a slightly unusual degree seemed to be attempting to draw and maintain the interviewer’s attention, and (not untowardly) to evoke a sympathetic response.’278 And, remarkably, this father was part of a dyad classified B4 five years earlier. Main set Goldwyn, her teenage research assistant, a ‘guess and uncover’ challenge. Could she match the transcripts of the parents’ autobiographical interviews to their child’s Strange Situation classification from five years earlier?
Ultimately the result was the basis for the AAI coding system, which is discussed later in the chapter in the section ‘Adult Attachment Interview’. The system was based on ‘guess and uncover’ on a sample of 36 transcripts, and then applied to 66 further cases: ‘Developing the system involved moving (blind) through each transcript in the development sample, and in each instance using feedback (“correct” or “incorrect”) with respect to the infant’s attachment classification to that adult) to refine and to further develop the rule system. This is a slow-moving but highly profitable method of rule development, and it was used in the creation of every succeeding system.’279 Part of what made the method so powerful and creative compared to the standard hypothesis-testing tradition in psychology was that both ‘hits’ and ‘misses’ fed fine-grained inductive information into a theory-driven deductive system.280 It was therefore exceptionally well adapted for discerning unexpected links in new data on the basis of a pre-existing frame of reference, in this case the infant attachment classifications which Main had come to regard as reflecting the basic forms of human emotional life. As Main put it, ‘hypothesis-testing procedures were combined with inductive processes’.281 Assuming the accuracy of the infant classifications, the benefit would then be that ‘this method maximises the likelihood of the discovery of new, attachment-associated patterns’.282
In her unpublished 1995 book Four Patterns of Attachment Seen in Behaviour, Discourse and Narrative, Main offered further reflection on the guess and uncover methodology.283 She linked it to the idea of the ‘hermeneutic circle’ in the work of hermeneutic philosophers such as Heidegger, Jaspers, and Gadamer: ‘In this circle, the relations between the whole and the part move forward with each repeatedly transforming and extending understanding of the other. Thus, each part contributes to the understanding of the whole which, transformed, then directs attention to new parts. These again alter the meaning of the whole.’284 For Heidegger and later philosophers in this tradition, the hermeneutic circle was a description of how all human understanding operates. Mortal and mutable, each of us interprets new perceptions always in the context of our horizon of pre-existing assumptions, which in turn can be modified by the new perception.
Main regarded the guess and uncover method as a sharpening and intensification of this process of ‘dialectic’ between parts and wholes. The horizon of assumptions gets brought to an explicit point every time the researcher dares to extrapolate from their existing rules in guessing the next case, and the encounter with novelty is likewise sharpened as a source of inductive learning by the repeated feedback of both ‘hits’ and ‘misses’ and attempt to then revise existing rules. The process is then extended at a meta-level as the rules generated, ideally, on a sub-sample are applied to the sample as a whole. The system is then fine-tuned and subjected to attempts at replication on further samples. Through this process, the coding system is likely to be changed by learning from the first few additional samples.285 Over time, the rate of learning slows, as the system is stabilised by a system of rules that are effective enough to deal with most novelty. Further refinements may occur, in principle. Without fail, every one of Main’s coding systems identifies itself as unfinished and with future changes expected, though none has seen revision now for decades.
Main offered a characterisation of this hermeneutic circle within Ruth Goldwyn’s ‘guess and uncover’ challenge on the AAI transcripts:
Reflecting on the ‘guess and uncover’ approach used by Main and colleagues in the development of their measures, this approach has some quite clear limitations. Main was well aware that the approach risked over-fitting coding systems to the data of the original sample. It also took continuity as a methodological assumption: this is no problem when recognised as an assumption. Yet with few details about ‘guess and uncover’ publicly available, attachment researchers and the wider community of readers naturally interpreted Main’s identification of continuities as a research finding. This contributed to an expectation of continuity in subsequent research, and to publication bias against reports of discontinuity.287
With ‘guess and uncover’ interpreted as proof of continuity rather than as a technique for generating hypotheses, it was wholly to be expected that researchers in the 1990s would be concerned when findings showed a ‘transmission gap’ between AAI classifications and Strange Situation classifications.288 However, it was also wholly to be expected that later researchers would face a growing weight of evidence suggesting much weaker continuity than in Main’s findings, as they have done in the past few years (Chapter 6).289 In the meantime, however, over decades the assumption of continuity, together with the dominance of the idea of minimising and maximising strategies, has powerfully shaped the imagination of attachment research. For instance, new measures developed for different age-groups have tended to retain the same categories used by Main. Few attachment measures have had their categories developed inductively, based on theory about age-appropriate tasks and naturalistic observation. They did not seem much need, since the relevant categories already seemed to have been set out in advance by Main. The Minnesota group are among the exceptions (Chapter 4), but their measures have had less theoretical influence, and uptake, than those of the Berkeley group.
At the same time, for the purposes of making novel observations and links, the particular strengths of ‘guess and uncover’ must be acknowledged. The integration of inductive and deductive strategies within the approach was central to Main’s elaboration of her new methods, and to the novel hypotheses that they permitted. The hypothesis of a link between a dyad’s attachment classification in the Strange Situation and qualities of the parent’s discourse was perhaps the most remarkable, and is discussed further in the section ‘Adult Attachment Interview’. However, ‘guess and uncover’ was also integral to Main and Cassidy’s identification of the ‘controlling’ categories in the six-year reunion assessment.
Building on the success of Goldwyn’s use of ‘guess and uncover’ on the AAI transcripts, the method was next applied by Cassidy during her visit to Berkeley in 1982. Several video-tapes from the six-year reunion were stolen in a theft, and some tapes were unusable due to technical difficulties. There were 33 children available with reunions with both parents for Cassidy to analyse:
The Main and Cassidy rule system was then given to an independent researcher to recode the whole sample including, it would appear, the development set. The independent researcher found 84% of child–mother reunions predictable from the infant attachment classification, and 62% of child–father reunions.291
Main and Cassidy classified dyads as secure if, on reunion at age six, the child remained calm and relaxed, but also expressed open pleasure on the parent’s return, and initiated conversation or interaction with the parent (or seemed ready to communicate if he or she did not initiate interaction). Responsiveness to the parent was often displayed as a ready expansion of the parent’s own remarks, continuing the conversation. In the Main and Cassidy system, dyads were classified as avoidant if the child responded only minimally. The six-year-old child did not turn away from the caregiver, or partially stonewall them, as might an avoidant infant (especially the A1 subgroup). Instead, the child turned down the intensity of interaction to simmer whenever opportunities arose. For instance several seemed very interested in examining the toys just as they saw their parent enter. Neither anger nor affection was much in evidence. Dyads were classified as resistant if on reunion there was ‘exaggerative or maximizing of relatedness to the parent’ by the child, for instance ‘putting the arm around the parent and head-cocking while looking towards the camera (as though posing for a mother–child portrait), speaking in a baby-like voice, or sitting on the parent’s lap.’ There could be some signs of frustration with the parent, though this was much less in evidence than in the infant category.
Main and Cassidy created two categories of disorganised attachment for the six-year reunion system. In both cases the child’s behaviour towards the caregiver was controlling. The first to be identified was behaviour that was punitive and directive: one previously disorganized child ordered the mother to ‘Sit down and shut up, and keep your eyes closed! I said, keep them closed!’292 Another form of punitive behaviour was rejection of the parent through a hostile and implacable silence, timed in such a way as to leave the parent humiliated or embarrassed. A further child from this group ‘pushed the door shut as his mother tried to open it … another said immediately and angrily on his mother’s entrance, ‘Don’t bother me’.’293 In the other category the control was solicitous and caregiving. In one such dyad, the child asked the parent what she had been doing and how she was feeling, then carefully invited her to play, reassuring her that it would be fun for her. Another asked: ‘Are you tired, Mommy? Would you like to sit down and I’ll bring you some [pretend] tea?’294 One form of solicitous behaviour that particularly interested Main was ‘overbright’ behaviour, in which the child showed a nervous, skipping cheerfulness on reunion. Main and Cassidy observed that ‘brightness and caregiving of this kind do not merely alleviate the parent’s present but temporary depression or anxiety, but ultimately are used to relieve or forestall the experience of anxiety by the child’:
In the case of controlling-punitive behaviour, the anger system appeared to have been recruited in the service of attachment. Main and colleagues theorised that angry behaviour permitted the child to regulate the relationship, and ensured a conditional kind of proximity, even if it did not afford comfort.296 In the case of controlling-caregiving behaviour, the caregiving system appeared to have been recruited in the service of attachment. As Bowlby had described in Loss just a few years earlier, caregiving behaviour displayed by a child to a caregiver might help keep the caregiver near and prop up the adult’s own capacity to offer care (Chapter 1). The child was therefore also offered conditional access to proximity, even if this proximity was achieved at the expense of acknowledgement of the child’s own attachment-related feelings.
Though there were too few instances for Main and Cassidy to make a third category, the researchers also noted that some children displayed sexual or romantic behaviours towards their parent. They described this as an ‘emerging’ category, which would probably achieve full category status after more cases had been seen.297 Again, this could be regarded as the recruitment of components of what would later become the sexual system in adolescence, in the service of maintaining the potential for closeness with the attachment figure.298 Some further six-year-olds displayed conflict behaviours or disorientation on reunion with their caregiver, with similar kinds of behaviours as those in the Main and Solomon infant system. Initially, Main and Cassidy also had a category for ‘confused’ behaviour on reunion; however, this was not elaborated in later versions of the system.299
Ainsworth criticised Main and Cassidy for their use of the label ‘disorganised attachment’ to characterise controlling-punitive and controlling-caregiving behaviours in the six-year system. She did not think it was an appropriate description for behaviours that clearly seem environmentally responsive and coherently sequenced.300 Controlling-caregiving behaviour (and perhaps controlling-punitive behaviour) could predictably offer access to conditional proximity with the caregiver, making it ‘organised’ by Main’s own earlier definition. In the controlling-punitive and controlling-caregiving categories, attention may also be coherent, in being focused on the adult. Furthermore, the use of anger or caregiving (or sexuality) in the service of attachment could be a behavioural repertoire channelled by human evolutionary history, like the avoidant and ambivalent/resistant strategies. However, the term ‘disorganised’ was retained by Main and Cassidy to signal the longitudinal continuities with the infant category. It also served to mark two ideas held by Main and Cassidy regarding these children. A first was that the controlling behaviour suggested a chronic obstacle to the termination of the attachment system sometime earlier in the child’s experience. Anger or caregiving would not have been pressed into service, they felt, if the child had received predictable access to proximity with their attachment figure in another way, whether through directly communicating distress and seeking care or through an avoidant or ambivalent/resistant conditional strategy. Secondly, a central aspect of the attachment system in childhood is that children seek care from someone stronger and/or wiser when they are distressed. However, controlling-punitive and controlling-caregiving behaviour involve an inversion of power between child and parent. There is therefore assumed to be a symbolic ‘disorganisation’ of the attachment system, even if behaviour is coherent, in the chronic confusion of roles.301
Controlling-punitive and controlling-caregiving behaviour may be smoothly sequenced and ultimately achieve a conditional proximity with the caregiver. Both punitive and caregiving behaviour allow children to approach, and their behaviour may help them retain the attention of the caregiver—understood to be a requisite to any successful functioning of the caregiving system. The fact that two main forms were found at age six, and that the evolutionary function of the controlling-punitive and controlling-caregiving behaviour could be inferred, suggests that they could be regarded as ‘conditional strategies’. However, Main has never referred to the controlling-punitive and controlling-caregiving categories as ‘conditional strategies’—even though this has become common usage among close collaborators such as Cassidy and Liotti.302 She does not offer an explanation for this. The Main and Cassidy six-year system was an extrapolation from the infant system, so it is quite possible that simply by definition for Main any behaviour labelled ‘disorganised’ could not be a conditional strategy, with the connotations of the term itself shaping theory. Perhaps, however, it could also be that for Main a conditional strategy had to offer an alternative way of achieving the goal of the behavioural system. Other attachment researchers defined the goal of the attachment system more widely as the availability of the caregiver or the maintenance of ‘organisation’, and so the controlling behaviours seem to be strategic in this sense.303By contrast, Main’s model of the attachment system was formed by Bowlby’s early control system model in which proximity terminated the system. The controlling-punitive and controlling-caregiving six-year-olds, even if they do achieve proximity to hit or settle their parent, do so primarily to the rhythms required for regulating the parent. Such behaviour can be inferred to ultimately benefit the attachment system in offering some proximity, but it is not clear that the system can be terminated if this rhythm is generally unrelated to the child’s attachment needs.
Stability between infant attachment and the Main and Cassidy six-year system was high. All 12 infants classified as secure with mother in infancy were classified as secure five years later. Six out of the eight avoidant dyads received the same classification, as did eight out of twelve disorganised dyads.304 This led Main and Cassidy to conclude that controlling-punitive and controlling-caregiving behaviour can be regarded as an attempt to organise and regulate the behaviour of a caregiver who might otherwise be alarming. For instance, both behaviours might retain the attention of a caregiver struggling with trauma, who might otherwise enter into episodes of abrupt unavailability. Cassidy recruited 50 dyads in Charlottesville for a replication of the six-year reunion study. Intercoder reliability was 90% (κ = .84) for the organised categories. However, eight of the twelve category disagreements involved the D category. In a reapplication of the procedure after one month in Charlottesville, stability for the disorganised classification was poor: only seven of fourteen (50%) children classified as controlling in Session 1 were controlling in Session 2.305 This may be one reason why Main regarded the system as unfinished and has only offered training on it to a few researchers over the decades.306
An important later study using the Main and Cassidy system was conducted by Solomon and George, former doctoral students in Main’s group. During the separation they had the children in their sample complete a doll-play procedure featuring story stems regarding getting hurt, a threatening monster, separation, and reunion; meanwhile, their mothers were asked to complete the AAI and an interview about their experiences as a parent.307 The six-year reunions were coded by Cassidy. Solomon and George found that the mothers of secure six-year-olds had a distinctive quality in offering a balanced and detailed provision of helping the child cope with worries. The mothers of avoidant six-year-olds had a distinctively rejecting quality both towards the child and towards the caregiver’s own difficulties, with a pervasively negative attitude towards themselves and their child.308 And the mothers of controlling six-year-olds described feeling helpless and out of control in their relationships with the child: ‘Some children in the controlling group were described as wild or helpless and vulnerable (e.g., locking the mother out of the house, persistent bed-wetting, wild tantrums). Other children were described as precocious or powerful (e.g., comedian-like behavior, amazing acting abilities, caregiving skills, supernatural powers, special connections with the deceased).’309 All of the caregivers classified as unresolved on the AAI were part of dyads classified as controlling on reunion; indeed, there was a match between six-year reunion and AAI classifications in 81% of cases (κ = .74).310
In their doll-play, the children classified as controlling demonstrated themes of fear and danger, in contrast to the rest of the sample: ‘the narrative structure of these stories can best be described as chaotic and flooded. Catastrophe, sometimes multiple catastrophes, often arise without warning; dangerous people or events are vanquished, only to surface again and again. Objects float and have magical, malignant powers; punishments are abusive and unrelenting’; ‘fantastic disasters frequently arose without warning. Further, children hastily gave post hoc explanations for these as though they were surprised and disturbed by the direction the story had taken.’311 Five of the seven children classified as controlling-caregiving seemed ‘extremely uncomfortable with the task and did not want to enact a story’. When given repeated prompts, their narratives were brimming with themes of ‘chaos and disintegration’.312 Solomon and George observed that for the controlling group as a whole, ‘fears disrupted their doll-play by flooding the content in a chaotic and primitive way or were inflexibly contained through a brittle strategy of inhibition of play’.313 Children from dyads classified as controlling were more likely to be judged as having behavioural problems by their teachers and by their parents, whereas there were no differences between the secure and organised-insecure dyads on these measures.
After her work on the six-year reunion system, Cassidy undertook a large project with another of Ainsworth’s former students, Robert Marvin, in the development of a preschool system for coding the Strange Situation. These efforts were supported by a working group, including Ainsworth and Main, funded by the MacArthur Foundation.314 Unlike the Ainsworth coding system, which developed inductively from naturalistic observation, the Cassidy and Marvin coding protocol elaborated ‘up’ from the infant coding system and ‘down’ from the six-year reunion system.315 Main’s ‘guess and uncover’ method was used with a set of 300 Strange Situations drawn from various American (and one British) research groups. The MacArthur preschool system is based in the first instance on common themes between the two systems. First, ‘the pattern shown by infants classified insecure/avoidant (pattern A) is strikingly similar to that shown by six-year-olds classified insecure/avoidant: children at both ages avoid intimate interaction or contact, maintain an affective neutrality, and convey the impression that the parent’s return is of no particular importance to them’. Second, ‘the pattern shown by infants classified securely attached (pattern B) is strikingly similar to that shown by six-year-olds classified securely attached: children at both ages show interest in proximity or at least interaction with the attachment figure’. Third, the ambivalent preschool classification incorporates the resistance of the Ainsworth infant system and the intensified signalling of relatedness to the parent of the Main and Cassidy system. Immature behaviour contributes to a subclassification of ambivalence, on the analogy of passivity in the Ainsworth infant system. Fourth, disorganised attachment is coded on the basis of either Main and Solomon behaviours from the infant system or the controlling-caregiving or controlling-punitive behaviours from the Main and Cassidy system. In 2002 Cassidy and Marvin added some additional subclassifications, including a depressed subclassification and one for children who seemed overtly fearful of their caregiver.316
The Cassidy and Marvin system has been used very widely and, with annual training available, new studies using the approach continue to regularly appear. As such, in all likelihood it represents the most important legacy of the Main and Cassidy system, where training has not been available. There has been a severe bottleneck for relevant research in middle childhood over the decades. For instance, the only large randomised study of adoption after institutionalisation, the English and Romanian Adoptee project, conducted a separation-reunion procedure with their participants at age six. Due to the de facto lack of availability of the Main and Cassidy method, the researchers chose to code their data using the Cassidy and Marvin system. They found uninterpretable results.317 The researchers concluded that the Cassidy and Marvin system was not appropriate for their study. Given the unique nature and global importance for policy of the English and Romanian Adoptee project, this was a wasted opportunity on a historic scale.
Given the lack of availability of training, the most—and potentially the only—especially influential study since the 1990s to have used the Main and Cassidy system was a longitudinal study by Ellen Moss and colleagues. With a community sample of 120 French Canadian children, they coded the Strange Situation at 2.5 years using the Cassidy and Marvin system and a separation-reunion procedure coded with the Main and Cassidy system at 5.5 years. Though trained in the Cassidy and Marvin system, it is not clear that Moss and colleagues were trained in the Main and Cassidy system, or sought (or could seek) interlaboratory reliability. The researchers found moderate stability of attachment classifications over the preschool years (secure: 72%, z = 6.0; avoidant: 44%, z = 3.6; ambivalent: 62%, z = 5.0; disorganized: 77%, z = 5.6). None of the dyads classified as disorganised at 2.5 years was classified as secure at 5.5 years. Changes from secure to disorganised attachment were associated with changes in caregiver behaviour towards the child,318 lower marital satisfaction, and events such as bereavement and parental hospitalisation. In addition to the separations associated with hospitalisation, Moss and colleagues offered the speculation that ‘parental hospitalization is also likely to compromise the child’s confidence in parents as a source of protection and security’.319
In an expansion of the sample by Moss and colleagues to include a number of five-year-olds, 68% of the children classified in the D group showed one of the controlling responses on reunion; 32% showed Main and Solomon-style disorganisation. Maternal report of marital conflict and unhappiness was associated with the display of Main and Solomon indices on reunion, but not with the Main and Cassidy controlling behaviours: ‘The unpredictability and overwhelming nature of the family environment, which is disrupted by severe marital conflict, may compromise the possibility of the child’s forming an organized integrated model of attachment, even one of a role-reversed controlling nature.’320 The researchers did not include measures of exposure to violence, abuse, neglect, or other forms of adversity that might have shed further light on the experiences of these children. However, contrary to the researchers’ expectation, Moss and colleagues found no association between maternal depression and any of the forms of disorganisation, including the controlling strategies. Though given that this was a community sample, they by no means ruled out that such an association would be seen in a population facing more adversities. Moss and colleagues did find, though, that families of children who developed a controlling-caregiving pattern were more likely to have experienced a major bereavement. Children who showed controlling-punitive behaviours on reunion with their parent were rated by teachers as displaying more conduct problems at school; children who showed controlling-caregiving behaviours were rated by teachers as displaying more anxiety and depression at school.
Another of the six-year systems developed by the Berkeley group was a system for coding family drawings. The use of drawing tasks as a projective measure had roots in the work of both Bowlby and Ainsworth.321 The children in the Berkeley six-year follow-up had been asked to complete a family drawing. Like Ainsworth’s use of projective measures (Chapter 2), Kaplan and Main intended the family drawing task as a standardised, ambiguous, and evocative stimulus, to elicit individual differences based on past experiences in the family relationships. It was also well aligned with Kaplan’s background in arts and humanities from her undergraduate studies at Sarah Lawrence College.322 The ‘guess and uncover’ methodology was, once again, utilised in the development of the Kaplan and Main drawing system. However, rather than use a development sample and then apply the scheme to the rest of the data, the number of family drawings was judged too small for this approach. Kaplan and Main therefore performed no external test, but used the whole sample in the development on the scheme: ‘Moving blind through the set of drawings while developing the rules for identifying the classifications, the two original judges independently obtained hit-rate [with the infant classifications] of close to 80%. The drawings and the rule-system were then given to a third blind judge who had not participated in the development of the system: this judge obtained an overall hit-rate of 76%.’323
Family drawings classified as secure by Kaplan and Main had the general quality of being ‘rational and realistic’. Kaplan and Main found that figures tended to be complete, with eyes, nose, mouth, hair, hands, and feet. This was despite the fact that there was no relationship between an independent assessment of drawing skill and attachment classification at age six. Drawings classified as secure had figures that seemed well grounded, with legs planted in a way that suggested that the floor beneath them, even if invisible, was adequately stable. Figures appeared in the middle of the picture, and did not have their arms closed off from one another or seem to be in over-rigid postures. When height differences were characterised, no figure loomed over anyone else. Objects and houses might appear in the picture, but did not dominate the scene in a way suggesting that attention was being directed away from the attachment figures. A curious aspect of the secure classification in the Kaplan and Main system is that they specifically stated that the absence of the self from the picture is not relevant to the classification. This was partly the result of their inductive methodology, since the absence of the subject was seen in a number of pictures where the child had previously been classified secure in infancy. However, the conclusions also seems theoretically motivated. Whereas ‘internal working models’ are generally described by other attachment researchers as representations of the self and others (Chapters 1 and 5), the secure classification on the six-year drawing system shows that this was not Main’s primary concern. Depictions of the self are not necessarily relevant: what mattered was the kind of attention directed towards or away from attachment figures and attachment-relevant experiences (Chapter 2).
Family drawings classified as avoidant by Kaplan and Main had many merry figures with little individuation. They also often lacked arms, or had their arms in postures not suitable for physical contact with others. Kaplan and Main found this especially interesting, since when asked to do a different drawing, of a bear, arms were then present and the apparently unwelcoming posture was absent. The figures also seemed rather isolated from one another on the page. The overriding characteristic was a picture of a family with attention directed away from the potential for the members’ potential vulnerability to or intimacy with one another. Family drawings classified as resistant tended to draw out-of-scale family figures, placed very close together, sometimes even with overlapping arms or shoulders. Several features of the drawings emphasised vulnerability. For instance, some of the children drew themselves with round bellies; half of children in this category, and no other child in any other classification, drew an unusual slant to the neck and head relative to the shoulders and the rest of the body, a posture that suggested coy submission. However, there were relatively few ambivalent/resistant children in the sample, and the classification was somewhat unfinished.324
Drawings classified as disorganised had some disrupted or chaotic elements, such as the child having to redo and redo the task or being unable to attempt it effectively. Kaplan and Main also identified two elements that corresponded to the punitive and caregiving behaviours that had been observed by Cassidy in the reunion episodes. A first was the presence of ominous elements in the picture, such as black clouds looming over the figures, dismembered or floating body-parts, or rows of skeletons in the background.325 A second was over-bright elements ‘which at first sight appear simply very sweet or very cheerful (e.g. a family standing on a row of hearts with a small sun drawn directly over the mother’s head; a huge sun with a smiling face dominates the figures in the picture)’. This was different to the addition of hearts and flowers added by many other children, since here the cheerful elements seemed to dominate the image and to have a role in emotionally propping up the characters rather than expressing their inner state.
Though an unpublished system, the research community was intrigued by the Kaplan and Main coding system, since it seemed to offer a special access to the child’s ‘internal world’.326 It was also the least resource-intensive measure to date produced by the developmental tradition of attachment research. An attempt to replicate the Kaplan and Main study was conducted by Fury, Carlson, and Sroufe using data from the Minnesota Longitudinal Study of Risk and Adaptation. No training by Kaplan or Main or attempt to achieve cross-laboratory reliability was reported.327 The children in the study were eight- to nine-year-olds. Fury and colleagues had concerns about some aspects of the system. First, ‘many of the Kaplan and Main signs that required a large measure of subjectivity (e.g., “faint ominousness”; “pained smile”) were excluded’.328 Second, some of the signs seemed overparticular, and a product of overfitting of the coding system to the Berkeley sample. Indeed, they found that only one of the indicators of avoidance from the Berkeley drawing study—the positioning of figures’ arms—was associated with avoidant attachment in the infant Strange Situation in the Minnesota sample. Fury also alleged that the semi-inductive ‘guess and uncover’ methodology used by Kaplan and Main was overly scattershot in the identification of specific markers, and that Kaplan and Main offered too little in the way of principles regarding ‘precisely how various descriptors were organised and distributed’.329 In the Fury study, none of the individual markers of disorganised attachment in the drawing system was associated with disorganised attachment in the infant Strange Situation. However, when signs were combined into aggregate indices, the expectable associations appeared, though associations were weak or modest (avoidance, r(168) = .15, p < .05; resistance, r(168) = .27, p < .001).330 The best association of the drawing system was not with the infant Strange Situation, but with a global measure of mental health: a .32 association with resistance in girls (though no association for boys) and a .36 association with avoidance in boys (though no association for girls).331 Main was dismayed by these results, and in a 1995 chapter concluded that ‘children’s drawings should not be used in the assessment of attachment’: at most, they should be ‘regarded only as correlates of early attachment’.332
The Minnesota findings were replicated by Madigan and colleagues with 127 dyads drawn from a moderate-risk sample in Ontario, with a significant proportion of children diagnosed with forms of heart disease. Again, no training by Kaplan or Main or attempt to achieve cross-laboratory reliability was reported. These researchers found few links between infant attachment and specific markers in the drawings made by the seven-year-olds in their sample. However, they found that the match between the infant and the seven-year attachment classifications was 80%, with most of the mismatches occurring for the avoidant infant category: over half were classified as secure in the drawing system. The low stability in the high-risk Minnesota study and the high stability in the lower-risk Berkeley and Ontario samples is in line with findings regarding the stability of attachment classifications in general (Chapter 2). Madigan and colleagues speculated that a higher proportion of disorganised cases in the Minnesota study may also have played a role in the different results.333
Running contrary to Main’s 1995 claim that the drawing system should no longer be used as a measure of attachment, in a 2016 paper Gernhardt and colleagues observed that the system had, in fact, become increasingly popular over the subsequent decades. Suspicious of the measure, they wondered about its cross-cultural applicability. In their study, Gernhardt and colleagues used the drawing system—though without the disorganised classification—with 32 six-year-old children from Berlin and 21 six-year-old children from rural farming villages around Kumbo in Cameroon. Once again, no training by Kaplan or Main or attempt to achieve cross-laboratory reliability was reported. Their results showed that most of the pictoral markers of insecurity were dramatically more common in the Kumbo sample. For instance, neutral or negative facial affect characterised every drawing by the children from Kumbo, compared to a tiny fraction of the children from Berlin. Likewise, lack of individuation, arms downwards, and unusually small figures were common features in the Kumbo pictures, whereas they were rare in the Berlin pictures. Nearly half of the Kumbo drawings omitted the mother, whereas none of the Berlin drawings did so. Gernhardt concluded that the Kaplan and Main system is essentially a cultural artefact, without any cross-cultural validity. This conclusion is questionable since they did not examine whether the results of the drawing measure correlate with expectable individual differences in the children’s lives or behaviour. Their view was that to do so would have been culturally inappropriate and potentially pathologising. Nonetheless, Gernhardt and colleagues do effectively make the point that many markers of insecurity identified by Kaplan and Main on the Berkeley sample need to be treated with caution when applied to contexts with different ways of thinking about, representing, and experiencing family life.334 This caution is especially important since the underpinning principles for the Kaplan and Main system, which could have been used by researchers to responsively elaborate the specifics of the coding system to achieve greater cross-cultural applicability, remain unpublished and, additionally, underdescribed even in the unpublished texts.
During the six-year follow-up at Berkeley, while their children were busy with tasks such as completing the family drawing, mothers and fathers were interviewed separately about their experiences relevant to attachment. Bowlby’s 1980 book Loss had been shared with Main in draft a few years earlier. In this work Bowlby was interested in the variety of forms that memory can take, and especially the work of Tulving. According to Tulving, ‘episodic’ information comprises temporally dated episodes or events and of relations between such episodes or events, as experienced by an embodied subject; by contrast, ‘semantic’ information contains generalised propositions about the world.335 Bowlby argued that whereas episodic information derives primarily from an individual’s own embodied experience, semantic information is more explicit, linguistically encoded knowledge, in particular what others have told the individual, especially as a child.336 Kaplan, George, and Main therefore developed an interview protocol to include prompts for both descriptions of key attachment relationships and specific supportive memories.
The AAI began with a question, developed by Kaplan on the basis of her clinical interests,337 asking speakers to describe their relationship with their parents. Main added a request that speakers choose five adjectives to describe each relationship, a prompt for semantic memory.338 Speakers were then asked to explain what made them choose each adjective with reference to illustrative memories, a prompt for episodic memory. They were asked what they did when they were upset in childhood, a verbal analogue in the AAI for the separations and reunions experienced by a child in the Strange Situation procedure. Speakers were also asked whether they could remember being physically held by their parents for comfort as a child. This question was likewise a verbal analogue, this time for Main’s ‘Aversion to Physical Contact’ scale for infants, which had proven singularly effective in discriminating caregivers in avoidant dyads in the Baltimore and Berkeley samples (Chapter 2). Another question asked by the interview protocol concerned whether speakers had any experiences of feeling rejected as a child, and if yes, why they now thought their parents had behaved as they did. Additionally, participants were asked whether they had undergone any major separations from their parents in childhood, building on Bowlby’s classic concern (Chapter 1). They were also asked about occasions when parents might have been threatening to them, following on from Main’s interest in the role of threatening behaviour in the aetiology of avoidant attachment. In line with a particular concern of Kaplan and Main’s, and the topic of Bowlby’s most recent book on Loss, participants were questioned about experiences of loss and bereavement. Participants were also asked about why their parents behaved as they did, and whether their early experiences of care have affected them as an adult.339
Any one of these questions might be asked of a person in certain contexts in ordinary life. However, the requirement to respond to this cumulative incision into speakers’ experience of their attachment relationships was a demanding and potentially distressing challenge. It cut away the familiar, conventionalised rhythms of interaction that usually offer us mooring against memory. At the same time, a chance to talk to an interested person about personal experiences without interruption, exasperation, or contempt is something quite special. Amidst all its bustle, it is rare in ordinary life that we are freely given sustained and patient attention by another. Furthermore, the overall tone of exemplary incidents from our childhood is something that we have little reason to share as adults, and may have few prompts to consider in any structured way. There is a quality of matter-of-fact strangeness to the intimacy of delivering or receiving an AAI.
Applying the ‘guess and uncover’ approach, the first pattern identified by Goldwyn was that some parents described their childhoods in flatly glowing terms. There were no problems; everything was fine. Yet, intriguingly, rather than belonging to the accounts of parents from securely attached dyads, these accounts of perfect childhoods belonged to the parents from avoidantly attached dyads. Main and Goldwyn therefore termed this ‘idealisation’. In a conference paper giving the first public report from this research, Main described a ‘mother who stated initially that her mother “was a good one” and that they had “a fine relationship” later in the interview told us—as though spontaneously—that she had painfully broken her hand as a child. Although she had been in pain for weeks she had not told her mother because her mother would have been angry. This incident was recounted in almost the third person—i.e. “but one couldn’t tell her”. When this mother was seen in the Strange Situation with her infant, the behaviour of the child led to the dyad receiving top scores for avoidance.340 Another facet of the transcripts of parents from avoidantly attached dyads was that though the parent asserted an ideal childhood, in fact they could supply few or no concrete memories of experiences of intimate and caring interactions. They often seemed to show little regard for “the need to depend on others” or “recognition of missing and needing others or being missed and needed by others”.’341
Furthermore, there were indications in the transcripts that at points speakers had experienced rejection in their attachment relationships. Yet this did not seem to inform the speaker’s overall characterisation of the relationship, which was resolutely positive. Main ran an initial test on the first 26 transcripts and found a substantial association between idealisation of the parent and indications of early experiences of rejection (r = .51).342 She also found a robust association between reports of early experiences of rejection in the AAI and displays of aversion to physical contact with the infant in the Clown session: r = .56 for mothers and r = .63 for fathers.343 However, Main had the qualitative impression that this association was stronger for speakers where rejection was inferred by the coder, rather than consciously and clearly reported by the speaker.344 Discussions between Goldwyn and Main identified what appeared to be the common thread: ‘we may conceive of the AAI (like the Strange Situation) as creating conditions that arouse and direct attention toward attachment’.345 Just as the infant in an avoidant attachment relationship directed attention and behaviour away from the caregiver on reunion in the Strange Situation, the parent in interview was directing attention and discourse away from the actual events of their childhood and associated feelings.346 In both cases there is a swerve of the heart.347 Main theorised that ‘whether or not the move is conscious and deliberate, the experience of rejection by the mother in childhood has led to a shift of attention from attachment figures, experiences and feelings in adult life’.348 Such experiences of rejection might include a parent’s aversion to physical contact, but could also be produced by other forms of rebuff, incomprehension, or disparagement. In turn, infants of these caregivers show avoidant behaviour in the Strange Situation, keeping potential distress away from their own and their caregiver’s attention. Main wrote to Bowlby, concluding that ‘it appears that the infant does represent in its behaviour the parents’ life, unconscious models and intentions’.349
During a visit to Berkeley in December 1982, Ainsworth looked at transcripts together with Main. They came to the conclusion that the idealised narratives of parents from avoidant dyads represented a reliance on semantic memory, at the expense of episodic memories that might have qualified such undifferentiated accounts.350 This helped account for why speakers often reported a lack of memory for specific occasions when asked to substantiate the positive general account offered of their attachment relationships. Conversations with Ainsworth in 1982 revealed her qualitative impression that in the Baltimore sample, too, conversations with the mothers of avoidant dyads suggested idealising of their own childhoods.351 Main and colleagues initially termed the speech of these caregivers ‘detached’, since its defining feature was a minimisation or disavowal of negative experiences, the feelings associated with these experiences, and of the value of relationships. However, this term from Bowlby was ambiguous, and Ainsworth thought that it risked connoting a lack of attachment (Chapter 1). Detached speech was therefore renamed ‘dismissing’ (labelled ‘Ds’).352
Another pattern identified by Goldwyn using ‘guess and uncover’ was that several speakers seemed highly concerned with grievances; they remained angry and preoccupied with their past and present relationships with attachment figures to such an extent that they often took long conversational turns and lost track of the question. They seemed focused on the vivid recollection of their childhood attachment relationships. Their goal at times seemed to become that of making a case against their parents, rather than responding precisely to the interviewer’s questions. In the terms that Main had developed for thinking about the Strange Situation, these speakers were directing attention towards attachment-relevant cues, even if this was at the expense of cooperation with the interview. The speech of these caregivers was initially termed ‘enmeshed-conflicted’ (E); this was changed to ‘preoccupied’ as the first term was ultimately considered stigmatising, though the category label ‘E’ was retained. Preoccupied speech seemed to Main and Goldwyn to be analogous to the infant of an ambivalent/resistant dyad in the Strange Situation, who is anxious at even the prospect of separation, ‘unable to direct attention to the environment, expresses strong and sometimes continual fear and distress, and seems constantly directed toward the parent’.353 And indeed, some of the preoccupied speakers in the Berkeley sample had been part of such dyads five years earlier.
Goldwyn also discerned qualities that seemed to distinguish the speech of parents who had been part of securely attached dyads five years earlier. These speakers described both the positives and negatives of their early relationships with good balance, even if these relationships had been difficult. Conversations between Goldwyn, Main, and Hesse suggested that a central property of the transcripts of secure infants was that the speaker appeared to have ready access to both semantic and episodic memory and so could coordinate a response to both kinds of prompts.354 Just like the secure infant in the Strange Situation could turn attention to the toys or the caregiver in response to changing environmental cues, the speakers seemed to be able to attend to both semantic and episodic memory, and positive and negative experiences, in a flexible way. In fact, the caregivers of dyads classified B3 in the Strange Situation in the Berkeley sample displayed these features most prototypically.355 Main, Goldwyn, and Hesse therefore termed the speech of these caregivers ‘secure-autonomous’, not because the speaker seemed independently minded, but because the defining feature of the transcript was the absence of restrictions in the discussion of experiences. The category label given to these speakers was F, since they seemed ‘freed through experience or thought to recognise the importance of attachment relationships, yet independent enough to evaluate them’.356 This seemed to be in contrast to parents who had been part of avoidant and resistant dyads, where there seemed to be ‘restrictions of varying types … placed on attention and the flow of information with respect to attachment’.357 In a handwritten annotation on a very early draft of the coding system, Main remarked that in the most prototypical of secure-autonomous speakers, ‘there is a striking ability to integrate existing information’.358
This developing coding system was clearly unusual. Whereas most psychological measures for coding interviews focused on coding answers to particular questions, the developing Main, Goldwyn, and Hesse method for coding the AAI examined the transcript as a whole, across the dance and drift of spoken discourse. The questions induced speech about personal experiences without respite: ‘in clinical terms, we would say that the objective of this interview is to “surprise the unconscious” with respect to attachment, through repeated, insistent probing’ combined with a lack of reciprocity from the interviewer.359 However, the most distinctive feature of the AAI as a scientific protocol, as well as perhaps its most audacious and uncanny feature, was its focus on apparent restrictions on information, which put the coder in the position of comparing ‘the subject’s own semantic categorisations of her experiences’ with what ‘we could discern [of] actual experience’.360
As such, the coding focused neither on the subject’s stated opinions, nor on the individual’s inferred actual history, since this is unknowable from an interview. The focus was instead on the manner in which the speaker attended to and communicated about attachment-related experiences. The coder was asked to consider both the internal consistency of the account and the extent to which the speaker’s interpretation seemed plausible based on the episodic information provided.361 In this, the coder had to consider the world from the speaker’s point of view, and take a step back to evaluate the speaker’s account of the world and appraise potential distortions of information. In this sense, the coding system had clear analogies to the clinical interviewing of the period in the USA, placing the coder in the position of clinician. Through the 1970s, under the influence of psychoanalysis, a dominant trend in clinical interviewing in the context of mental health was to seek to identify psychological defences or confusions expressed through speech and decipherable by the clinician.362 However, from the 1980s, under the influence of the American Psychiatric Association’s DSM-III, clinical interviewing increasingly sought to identify specifiable markers for categories of mental pathology out of the particularities of what and how individuals report their experiences.363 The AAI had elements of both forms of clinical interview, which made it well suited and timely for clinicians with a psychoanalytic background in providing a scientific basis and classifications for blockages or disruptions in how individuals are able to know or regulate themselves.
Main and colleagues felt that ultimately ‘we were attempting to trace what could happen to information regarding negative or rejecting attachment experiences, other than their easy and coherent recognition and evaluation’.364 On the one hand, it could be defensively excluded through lack of memory and/or idealisation. On the other hand, it could become a preoccupying focus of attention in a way that hindered evaluation.365 Main described the object of the interview as the adult’s state of mind with respect to attachment.366 However, the phrase is quite misleading. Not least what is in question is not what psychological research generally means by a ‘state’: a transitory experience responding to an external prompt without the expectation of stability over time.367 Unfortunately, the pivotal article in which Main set out and described the concept of ‘state of mind’ was accepted by the journal Developmental Psychology but then withdrawn by Main as ultimately unsatisfactory.368 As such, the meaning of the term has often been unclear to her readers, contributing to misunderstanding of what the AAI actually measured, and fanned the flames of controversy regarding other forms of assessment (Chapter 5). Fonagy and colleagues, for instance, were left at a loss regarding the object of the AAI, wondering whether the coding system could actually be best considered as an indirect assessment of the caregiving system, rather than an assessment of anything attachment-related at all.369 Fonagy and Campbell subsequently concluded that the AAI does assess something related to attachment, but that Main’s own description of ‘state of mind regarding attachment’ in terms of internal working models has been confusing and, ultimately, a ‘reductionist over-simplification’.370 However, Fonagy and Campbell offered no further thoughts on what it is that the AAI does, in fact, measure.
In addressing this question, a first thing to note is that, despite its name, the AAI was not an assessment of individual differences in a unitary entity called ‘attachment’.371 This is despite the fact that, among developmental psychologists, ‘adult attachment’ has come to mean ‘what the AAI assesses’. In fact, the name ‘Adult Attachment Interview’ appears to have originated from the interview’s concern with attachment experiences. True, the demand to attend to and communicate about attachment-relevant experiences might well be alarming, and might therefore also activate the attachment system. But this was not tested or even discussed. In fact, the central focus of the coding system developed by Main, Goldwyn, and Hesse was on the capacity to attend to and communicate about attachment-relevant experiences and the feelings they evoke. Main referred to this, somewhat loosely, as individual differences in ‘adult attachment’, since this capacity seemed to her to have strong commonalities with the Ainsworth classifications at the level of attentional processes. The term ‘adult attachment’ was also a powerful assertion that the ideas of Ainsworth were relevant across the lifespan. However, use of the term generated confusion, as well as acrimony between developmental and social psychology researchers in which ownership over the capacity to measure something called ‘attachment’ became a central stake (Chapter 5). However, Main and Hesse’s basic position is that ‘there exist species-wide abilities that are not part of the attachment system itself, but can, within limits, manipulate (either inhibit or increase) attachment behavior in response to differing environments’.372 In Main and Hesse’s conceptualisation of the AAI, what is measured is the deployment of such abilities as applied to adult attention and communication regarding attachment-related matters, not the ‘attachment system itself’.
Nor was the AAI developed primarily as an assessment of ‘internal working models’, if by this is meant representational content regarding attachment figures. (Admittedly, if ‘internal working model’ is taken to mean any cognitive components recruitable by the attachment system, then the AAI does tap these, but the claim does not tell us much.) Comparison of published with unpublished manuscripts suggests that Main’s use of the concept of ‘internal working model’ in the mid-1980s in discussing the AAI was primarily a matter of tying her new methods to existing theory, rather than actually the basis on which she was conceptualising the measure. It was also a position she would drop already in the 1980s. At the meeting of the Society for Research in Child Development in Baltimore in June 1987, Main spent time with Hinde. Hinde had huge respect for Main’s work, as well as a good deal of personal affection. However, underspecified causal claims about motivational processes seems to have been his personal bête noire. At the Baltimore conference he passionately chided Main for her appeal to the ‘internal working model’ concept.
Main wrote to Bowlby to report that she could see Hinde’s point, and that she was planning to give the matter further thought.373 From 1987 onwards, she ceased equating states of mind regarding attachment and internal working models. And in the late 1990s Main acknowledged in print Hinde’s point that the characterisation of the AAI as an assessment of internal working models was, ultimately, ‘misleading and unwarranted’.374 Fonagy and Target agreed, arguing that Main’s description of the AAI as assessing internal working models had prompted much ‘futile’ research and generally ‘distracted attachment researchers’.375They were glad that she had altered her description of the AAI. However, this shift in Main’s work has generally gone unnoticed and unheeded. Kobak and Esposito described how Main’s appeal to internal working models inadvertently trained researchers’ attention on personality trait-style qualities of individuals and their ‘models’ of others, rather than on attentional or communicative processes.376 The characterisation of the AAI as an assessment of ‘models’ of attachment relationships led to the misleading characterisation of the measure as an assessment of ‘attachment representations’. The AAI was not primarily an assessment of individual differences in representations held of particular parents, or even in representations of relationships in general, though both might influence classifications on the measure. Yet thanks to its initial characterisation in terms of internal working models, the AAI is still commonly described, including by trainers in the measure, as an assessment of ‘attachment representations’ in adults. This occurs even when researchers’ own descriptions of what the AAI assesses—for instance, the processing of attachment-relevant information—actually do not agree with the usual meanings of the noun ‘representations’.377
The term ‘representations’ essentially functions as a placeholder. But its use has unsurprisingly contributed to unwarranted assumptions about the measure. For instance, researchers have often wondered why adults are assigned a single classification, rather than a classification for each attachment relationship, as could be anticipated if the AAI was measuring representations of relationships.378 The use of a single classification makes sense, however, if Main’s focus on attentional and communicative processes in the years of its development is recognised: in her work, the AAI has been conceptualised as an assessment of individual differences prompted by the interview’s demand on speakers to attend to and communicate regarding attachment-relevant experiences. The only researcher to have consistently highlighted the importance of Main’s attentional model of states of mind regarding attachment has been van IJzendoorn. Van IJzendoorn had several advantages over other readers. For instance, in a paper from 1992, and citing his personal copy of the unpublished Developmental Psychology manuscript on the concept of ‘states of mind’, van IJzendoorn observed that generally in Main’s writings ‘the concepts of “internal representation,” “state of mind,” and “internal working model” of attachment are used interchangeably’. Nonetheless, he argued, the ‘concept of “state of mind” ’ is in fact less about representational content and more about ‘the direction and organization of attention and memory’, as well as related affective and behavioural aspects.379 Recognition of the importance assigned by Main to attention has gone on to guide van IJzendoorn’s research using the Strange Situation and AAI, for instance his concern with genetic polymorphisms that influence the dopamine system, given the role of this system in attention and reward processing.380
From early on in their examination of the transcripts, a central concept used by Main and Goldwyn to mark individual differences between the transcripts was ‘coherence’. As van IJzendoorn and colleagues noted, however, the Main and Goldwyn usage was technical, and differed in potentially confusing ways from everyday English.381 Indeed, a recent study by Lind and colleagues found an association of only r = .37 between Main and Goldwyn’s scale for ‘coherence’ and a conventional measure of narrative coherence.382 Main’s first use of the term was in the first edition of the coding manual, in 1986, where they defined coherence as ‘the extent to which the reader finds a unified, yet free-flowing picture of the subject’s experiences, feelings and viewpoints within the interview, such that none of these require the reader to make her own, differing interpretation. In addition, the interview transcript is considered coherent when the subject seems able to easily point to the principle and rationale behind her responses; has thought or else seems ready to think about her past and its influences.’383 In judging coherence, the manual urged the coder to look out for contradictions, especially between episodic and semantic memory, logical contradictions, and factual contradictions.
In the early 1990s, Main and Hesse continued to develop the concept of ‘coherence’ in reflecting on the question of what the coding system was capturing. In this, their thinking was influenced by the ideas of a colleague at Berkeley, the philosopher of language Paul Grice.384 Grice had argued that cooperative discourse usually demands, except when permission is given by the partner, adherence to four conversational maxims:
Quality—‘be truthful and have evidence for what you say’ Quantity—‘be succinct, and yet complete’ Relation—‘be relevant to the topic as presented’ Manner—‘be clear and orderly’
Main partly retained her earlier definition of ‘coherence’, but also partly redefined it in terms of Grice’s maxims.385 Whereas initially the term had been defined as (i) the extent to which the reader finds a unified, yet free-flowing picture by the speaker that agrees with the reader’s own account, the concept was tucked into (ii) the four dimensions identified by Grice. Characteristic of a secure-autonomous (F) transcript was that there was good episodic evidence for semantic generalisations; speakers could flexibly turn their attention to their past and to the interviewer in being succinct and yet complete and relevant: ‘For these speakers, the focus of attention appears to shift fluidly between the interviewer’s queries and the memories that are called upon. In this way, the speaker is able, regardless of the difficulty of the subject matter, to remain both truthful (consistent) and collaborative.’386 Finally, the discourse of a secure-autonomous speaker was characteristically clear and orderly, often with a fresh quality to the speech. These qualities require a capacity to both turn genuine and curious attention to attachment-related memories and let go again in order to attend to the interviewer. Main and colleagues developed subscales to help coders in making a judgement about whether a transcript should be classified as secure, parallel to the function of Ainsworth’s scales for the Strange Situation (Chapter 2). Of particular importance was a scale for the coherence of the transcript.387
Dismissing (Ds) interviewees tended to have poor episodic evidence for their semantic generalisations. They also tended to offer oversuccinct answers lacking the details of concrete experience, since they were not able to resource their responses through access to a richly populated and textured inscape of episodic information. Presumably this information had become defensively excluded. Relevance might also be weakly violated as the speaker deflected the conversation away from sensitive topics. In general, however, what is said is orderly. Preoccupied (E) interviewees tended to have poor alignment between episodic and semantic memory, as they were focused on the negative aspects of episodic memories even as they recounted events that could have other more balanced interpretations. However, the primary violations of coherence in a preoccupied transcript were excessive quantity of speech about topics that were not direct answers to the questions of the interview, often with poor orderliness in the communication of experiences caused by a focus on the memories themselves rather than on the comprehension of the interviewer. Their transcripts are a bit wild, marked by ‘highly entangled, confusing, run-on sentences; failures to use past markers in quoting conversations with the parents; rapid oscillations of viewpoint within or between sentences; unfinished sentences; insertion of extremely general terms into sentence frames (‘sort of thing’, ‘and this and that’), and use of nonsense words or trailers as sentence endings (‘dada-dada-dada’).’388 Main and Hesse also created scales to help coders identify dismissing and preoccupied transcripts. To help distinguish dismissing states of mind regarding attachment, the researchers created scales including for idealisation of the parent and insistence on lack of recall. To help distinguish preoccupied states of mind regarding attachment, they created scales for involving/preoccupied anger and for passivity or vagueness of discourse. Like the Ainsworth scales, the scales for the AAI were initially given the primary role of supporting placement of cases within categories. As with the Strange Situation, coders were enjoined to record their scale scores, resulting in an archive of largely unpublished findings. These have become the target of great interest recently, especially in the context of Individual Participant Data meta-analysis (Chapter 6). However, from the 1990s onwards, the coherence score came to see use as a dimensional alternative, or complement, to the category-based coding system.389
Hesse reflected upon the meaning of coherence at an interpersonal level.390 He conceptualised the AAI as constituting two tasks: to reflect on memories of attachment-relevant experiences, and to communicate about these in a way that holds in mind the interviewer’s specific questions. Transcripts classified as dismissing violate the first of these tasks; transcripts classified as preoccupied violate the second. In the same period, Main was reflecting on the meaning of coherence at a cognitive level. She proposed that the coherence of their discourse suggests that secure-autonomous speakers have a relatively unitary model of their experiences and how these influence their behaviour. She perceived that this integration of different sources of information and the lack of need to divert attention to the implementation of a conditional strategy might support a capacity to retain balance in the evaluation of relationships, to perceive how the same reality could be seen in different ways, and to allocate attention as needed to different tasks.391
Main was also impressed by apparent differences in epistemology between speakers: secure-autonomous speakers ‘adopted a more thoroughly constructivist view of their own knowledge-base than less secure adults’, with more subtle forms of awareness of the appearance-reality distinction. Secure-autonomous speakers may acknowledge that their own perspective on an event may differ to that of a family member, or that their recall may be distorted by their regrets about the occurrence.392 She believed that this constructivist view of knowledge served to support a speaker’s capacity to examine their own experiences (‘metacognition’) when prompted by the environment, since there is less segregated information: ‘more epistemic ‘space’’ is available for such an individual ‘because her thinking processes are not compartmentalised’.393 Though Main acknowledged that it could be, in theory, that metacognitive capacities were facilitating secure-autonomous speech, or that there were bidirectional effects, her view was that secure attachment was the cause of metacognitive strengths such as clearsightedness in self-evaluation. A metacognitive monitoring scale was added to the AAI coding manual in 1991 to support coders in making a classification of secure-autonomous state of mind regarding attachment. (Nearly three decades later, however, the scale is still identified as unfinished in the AAI manual. The measurement of metacognitive monitoring would not be pursued further by Main, and would instead be taken up by Fonagy, Steele, and Steele in their work on reflective functioning and mentalisation.394 Main’s lack of further work on the metacognition scale has generally been interpreted as tacit endorsement of the approach of Fonagy and colleagues.)
Main conceptualised the reduced coherence of dismissing and preoccupied speakers as reflecting the conflict between segregated information, and thus conflict between different interpretations by speakers of their own past. Everyone has multiple models of reality, Main observed, but dismissing and preoccupied speakers seemed to have ‘implicitly contradictory models of the same aspects of reality’, segregated from one another perhaps without the speaker’s awareness, but available to someone reading the transcript of their interview.395 On this basis, Main criticised Bowlby’s notion of the internal working model as too calm and settled a metaphor, to the point of actually being ‘somewhat misleading’. Whilst perhaps applicable to secure-autonomous speakers, when asked to describe and evaluate their attachment experiences and relations, dismissing and especially preoccupied speakers present an array of ‘contradictory thoughts, feelings, and intentions which can only loosely be described as a “model” ’.396 These speakers ‘evidence difficulties in obtaining access to attachment-related information; in maintaining organisation in attachment-related information; and in preventing attachment-related information from undergoing distortion’.397 Whereas psychoanalytic theory might conceptualise such effects as evidencing the repression of mental contents relevant to past experiences, Main argued that what is ‘defensively excluded in this conceptualisation is not the memory … but rather an alternative attentional and behavioural patterning’.398
Just as Main conceptualised the infant showing avoidant behaviour in the Strange Situation as primarily avoiding disorganisation, and only secondarily the attachment figure, so dismissing speakers were conceptualised as primarily avoiding a breakdown in their conditional strategy and its attentional and behavioural architecture. And like the ambivalent/resistant infant, the preoccupied speaker was conceptualised as keeping attention and behaviour trained on attachment-relevant information, avoiding cues from the environment that might suggest other priorities for attention and behaviour. Whether in childhood or in adulthood, what conditional strategies have in common in Main and Hesse’s thinking is that they ‘follow upon alterations in the focus of attention’.399 For it is attention that orchestrates what attachment-relevant information and cues are received from the environment or from memory. As such, these alterations in attachment-relevant information are topic specific. They were not anticipated to be carried over into the processing of matters irrelevant to secure base/safe haven availability.400
Yet alterations in attention to these specific matters were nonetheless anticipated by Main and colleagues to have wide-ranging implications for mental health. In line with this account, both dismissing and preoccupied states of mind have been found to be transdiagnostic. Though more prevalent in clinical samples, dismissing and preoccupied states of mind regarding attachment are not characteristic simply of any one form of mental illness (with the partial exception of a link between preoccupied states of mind and personality disorders). They are also remarkably stable during adulthood, though periods of developmental transition or changes in circumstances will often result in changes in states of mind regarding attachment.401 However, contrary to Main’s expectation, there has only been evidence of weak or moderate continuity from infant attachment classifications to AAI classifications, though infant disorganised attachment does generally predict a later insecure classification on the AAI.402 The question of the relationship between actual childhood experiences in attachment relationships and retrospective discourse in the AAI is one that remains debated, as the case of discussions of ‘earned security’ reveals especially clearly.
In her use of ‘guess and uncover’ to identify forms of adult discourse associated with infant secure attachment, Goldwyn identified the importance of the speaker’s ability, in the present, ‘to take a balanced view of relationships’. However, the personal history recounted by such speakers seemed to have one of two forms. A first was ‘a believable picture of one or both parents serving as a secure base or haven of safety in childhood, a picture which is not contradicted within the interview and which may even be illustrated by incidents of parental giving of comfort or support’.403 Narrated in such accounts, it seemed to Main and colleagues, was secure-autonomous speech as a consequence of relatively unbruised and secure attachment relationships in childhood. Yet several transcripts in the development sample revealed speech that was about difficult childhoods, but marked by flexible attention to the good and the bad, to memories of the past and to communication with the interviewer in the present, and to their own perspective and that of others. For example, one mother who had experienced rejection by her family responded to the researchers’ initial query regarding the nature of her early relationships: ‘how many hours do you have? I have one of those families that they should write a whole book about. Okay, well to start with, my mother was not cheerful, and I could tell you right now the reason was that she was overworked.’404 Though this mother had experienced rejecting caregiving, she seemed to Goldwyn and Main nonetheless able to acknowledge negative aspects of the relationship, with insight into both her own and her parents’ experiences. When this mother and her infant had been seen in the Strange Situation five years earlier, they were classified as B3, and the dyad received the lowest possible score on the Ainsworth avoidance scale.405
In the first version of the coding system from 1982–83, Main and Goldwyn identified three trajectories that seemed to be linked, despite a caregiver’s own difficult history of care, to secure-autonomous speech and a secure attachment relationship in the infant–caregiver relationship. A first was that some speakers seemed to have ‘actively engaged in a period of rebellion and escape from the parents’, moving away from these relationships as an opportunity to reorganise how they respond to distress in the context of relationships.406 A second trajectory seemed to be that some speakers had ‘forgiven the parents’, holding in mind both their own perspective and feelings about events but also articulating a sense of the difficulties their own parents had faced.407 A third trajectory, perhaps predisposed by the fact that Main’s sample was from near the University of California, Berkeley campus, was that some had ‘engaged in a period of study undertaken with a view to understanding child–parent relationships and their influence’.408 What all three trajectories seemed to have in common was the effortful achievement of a new perspective. By 1988, Main had come to refer to this as ‘earned security’.
The first study to discuss earned security in print and to study it empirically was conducted by Main and Hesse’s colleagues at the University of California, Carolyn and Philip Cowan, and published in 1994. The Cowans were conducting a longitudinal study of the transition to parenthood, and the AAI was administered to 40 adults when the first-born children in the study were 42 months. In the sample, 10 speakers were classified as insecure, 10 as ‘continuous-secure’, and 20 as ‘earned secure’. Earned security was defined in practice by the Cowans as speakers who were classified as secure-autonomous in speech, but whose transcripts also indicated a high score for neglecting and rejecting behaviour by either caregiver, and a comparatively low score for loving behaviour by at least one caregiver. The Cowans found that the half of their sample who were judged to be continuous-secure speakers reported lower levels of depressive symptoms than the insecure or earned-secure speakers. However, observations of child–caregiver interaction indicated that the earned-secure speakers and continuous-secure speakers offered the same degree of warmth and structure as their children played, in contrast to a lower degree of warmth and structure offered by speakers who had been classified as insecure on the AAI.409 The image of individuals able to ‘earn’ the benefits of secure early relationships through their own efforts in later childhood, adolescence, or adulthood has naturally been one with great appeal, especially in the American clinical community.410 And indeed, researchers found that earned-secure speakers are more likely to have spent time in psychotherapy than continuous-secure, preoccupied, or dismissing speakers.411 The classification also offered attachment researchers a powerful answer to accusations that attachment theory is concerned only with continuity from infancy to adult relationships. Moreover, it supported the application of the AAI as an outcome measure for evaluating psychotherapeutic interventions.412
Roisman and Sroufe, however, felt that there was a need to ‘be cautious about retrospective reports’.413 They cited Main and Goldwyn who had observed that it cannot be ‘presumed that these retrospective interviews can provide a veridical picture of early experience’.414 In a paper published in 2002, Roisman and Sroufe set out to examine the earned-secure classification prospectively, drawing on the Minnesota Longitudinal Study of Risk and Adaptation (Chapter 4); 170 participants in the study completed the AAI at age 19. Since the Cowan’s study, Main had also altered the coding manual to require that participants receive a score of lower than 2.5 on the scale for inferred loving parental behaviour in order to qualify as earned secure.415 Roisman and Sroufe noted, however, that only three of the participants in their whole sample met this stringent standard, perhaps because they had conducted the AAI at age 19 so there had been only scant opportunity for individuals to achieve secure-autonomous status following such adverse care.416 Roisman and Sroufe adopted an approach that resembled the Cowans, though with a few alterations based on methodological discussions in the intervening years.417 On the basis of this approach, 24 participants were classified as earned secure.418
Roisman and Sroufe replicated the Cowans’ finding that earned-secure speakers reported more current depressive symptoms. Prospective data also revealed that, a decade and a half earlier, their mothers had reported more symptoms of anxiety and depression in their children compared to both the continuous-secure speakers and those with a preoccupied or dismissing AAI classification. Findings were in the same direction in adolescence. Observations of the 19-year-old speakers in interactions with romantic partners revealed that earned-secure speakers scored much the same as continuous-secure speakers, and much more highly than dismissing and preoccupied speakers, on measures such as shared positive affect, provision of the partner with a secure base, and conflict resolution.
Yet, looking back in time, data from the prospective cohort study also revealed that there had been no difference between earned- and continuous-secure speakers in terms of their infant attachment classifications with either fathers or mothers. Additionally, in an observational assessment with their mothers at 24 months, earned-secure speakers had received the most supportive care of any of the children, more than the preoccupied, dismissing, and continuous-secure speakers. At age 13, the mother–child interactions of both earned-secure and continuous-secure speakers were much more supportive and positive than those of preoccupied and dismissing speakers. Roisman and Sroufe concluded that the ‘earned-secures were the beneficiaries of among the most supportive maternal care in a high-risk sample’ and, as a consequence, ‘we cannot rule out the possibility that self-described differences in early experience between retrospectively defined earned- and continuous-secures were primarily a function of positive and/or negative reporting biases (e.g. negative attentional biases associated with depression)’.419
Such findings underline the focus of the AAI on current states of mind regarding attachment: differences identified in the interview between subgroups of secure-autonomous speakers did pick out prospective differences in histories of care and relationships, such as more parent-reported symptoms of anxiety and depression in preschool among the earned-secure speakers. However, the findings seemed in some regards to also put into question the accounts of earned-secure speakers of adverse forms of care in childhood. Roisman and Sroufe interpreted their findings as suggesting that ‘earned-secures did not rise above malevolent parenting through sheer will; rather, their success was scaffolded by caring adults, their security was a natural extension of a supportive (although not necessarily ideal) past’.420 They therefore recommended that an alternative to the term ‘earned security’ might be considered. Hesse and Main subsequently agreed that ‘the term earned-secure is not an ideal taxonomic label’, and proposed the term ‘evolved secure’ as an alternative—though this proposal seems to have been ignored so far, or come too late in the day to change the presiding discourse.421 And indeed, attachment researchers have generally continued to depict the childhoods of earned-evolved secure speakers as adverse, and secure-autonomous speech in the AAI as reflecting an individual achievement, something ‘earned’, in the face of this adversity.422
A further study by Roisman and colleagues has provided additional evidence relevant to appraisal of the ‘earned-secure’ classification. They found that speakers classified as earned/evolved secure, even on the basis of the stringent Main and Goldwyn criteria, experienced levels of maternal sensitivity comparable to dismissing and preoccupied speakers at between 5 and 64 months.423 Their caregivers also had more financial difficulties and depressive symptoms. However, on other measures, these speakers appeared to have had more positive care. From middle childhood they experienced care that was around the sample average for sensitivity, and in adolescence observations of child–caregiver suggested that they received more sensitive care than preoccupied and dismissing speakers. They also received much more positive paternal care than the preoccupied and dismissing speakers, with no differences from those classified continuous-secure. The study by Roisman and colleagues has been subject to theoretical and methodological criticism from Hesse, who remains unconvinced that earned-/evoked-secure speakers really did have more positive care.424 In the context of such debates, the ‘earned/evolved secure’ classification will no doubt be subject to further research over the coming years by Roisman and other third-generation attachment researchers.425 Nonetheless, all parties to the debate about ‘earned security’ agree that it should be emphasised that the AAI solicits a retrospective account: it is qualities in the speaker’s discourse and reasoning about attachment-relevant experiences that form the basis of classification, not the nature of the events described. This is why attachment researchers have been especially intrigued by occasions when discourse or reasoning about attachment-relevant experiences appears to be disrupted.
An early observation made by Goldwyn was that at points some narratives became ‘splintered and incoherent, so that ideas were lost, superficially unconnected ideas invaded one another, and the whole approach to the topic of attachment became disorganised’.426 Goldwyn and Main documented occasions of varying degrees of such disrupted discourse—some extensive, some more momentary—across dismissing, preoccupied, and secure-autonomous transcripts, though they seemed somewhat predominant among preoccupied and dismissing speakers.427 Goldwyn’s application of ‘guess and uncover’ revealed that these transcripts frequently belonged to parents in dyads that had earlier been unclassifiable in the Strange Situation according to the Ainsworth categories. Furthermore, this semi-inductive method revealed that many of these speakers had experienced loss of attachment figures, especially before adolescence.428 However, further examination of the transcripts in 1983 revealed that early loss in itself was not a good predictor of infant attachment classification. Rather, what seemed critical was that speakers seemed to show disruptions in their discourse, to be in a semantic sense ‘at a loss’, when discussing the dead attachment figure. The concept of ‘unresolved grief’ or ‘unresolved mourning’ had been gaining prominence in the clinical literature of the 1970s.429 This development drew on earlier accounts by psychoanalysts, including Bowlby, of the way in which acknowledging and accepting a loss could contribute to mental health symptoms.430 Building from both Bowlby and the contemporary clinical literature, Main and Hesse conceptualised the speakers as ‘unresolved with respect to the mourning of an attachment figure’. These speakers were allocated an Unresolved/disorganised (U/d) classification.
In the late 1980s, Main and Hesse developed a ‘Lack of Resolution for Mourning’ scale, with support from a research assistant Anitra DeMoss. Most often markers of lack of resolution occurred in response to the direct question in the AAI about losses, but they could also be identified in other parts of the transcript that touch upon the relationship with the dead person. Main, Hesse, and Demoss distinguished two species of lapses in the transcripts.431 Sometimes these co-occurred, especially in clinical or forensic samples, but mostly transcripts displayed one or the other as a characteristic form. A first species was lapses in monitoring of reasoning without recognition by the speaker, leading to breaches in ‘coherence’ through interruptions in the plausibility of the speaker’s account.432 For instance, such lapses were implied in a belief that the dead person remains alive and actively involved in the speaker’s life (e.g. Mrs Q, Chapter 1). This includes statements in interview such as ‘So I don’t now what field I’m going to select, but my father [deceased 15 years ago in childhood] says that I should choose law’. Main and colleagues commented on ‘this slip of the tongue to the present tense regarding a lost attachment figure. It is taken as an indication of disbelief that the person is dead and leads to U/d category placement because: a) the speaker is talking about something going on in his immediate surroundings; b) which is currently of vital import to him, and is c) actively bringing the long-deceased into a freshly constructed sentence; d) as though he or she continued to have input. However, slips to the present tense sometimes do not indicate serious disorientation/disorganisation. If, for example, a person says of a deceased father ‘my father has been in banking for a number of years’ the slip may be minor.433
The coding manual positioned lapses in reasoning about whether an attachment figure was dead or not dead as the paradigmatic case of unresolved loss. Other rarer lapses in reasoning around loss included confusion between the dead person and the speaker, and characterisations of events in time or space that are not possible—such as being present at a family tragedy and also, simultaneously, being absent from the event in another country. Lapses in reasoning were ultimately characterised as ‘things which cannot be true in the external world’,434 though, anticipating problems with this definition, Main and colleagues advised coders as best they could to exclude statements that are grounded in a self-aware and integrated religious or cultural viewpoint.435 An additional set of relatively common lapses of reasoning was identified by Main and Hesse as occurring when speakers stated that they have done something that cannot be true psychologically, such as using willpower to erase experience of a past event. These are claims that, as it were, cannot be true of the internal world.
A second type of lapse indicating unresolved states of mind was lapses in the monitoring of discourse without recognition by the speaker. Examples included: invasions of remarks about a death into apparently unrelated discussions; sudden use of eulogistic speech; exceptionally long blank pauses that the speaker does not him-/herself seem to notice; or apparent absorption in particular details surrounding the death.436 Though there was no explicit statement in the AAI coding manual stating that lapses in reasoning are more important in making a classification than lapses in discourse, it is notable that only a few of the examples given of lapses in discourse are afforded a high score on the scale for unresolved loss, whereas many of the lapses in reasoning are given a high score. There may be more ambiguity regarding the cause of a lapse in the monitoring of discourse.437 Additionally or as a consequence, it may be inferred that lapses in discourse are less proximal markers than lapses in reasoning for what Main and colleagues were seeking to capture.
In line with this supposition, by the early 1990s, Main’s work with Solomon on infant disorganisation had led to a re-evaluation of the concept of lack of resolution. In 1991, Main and colleagues wrote that ‘as we have gained an increasing understanding of the nature of the link between the adult’s and infant’s state, unresolved/disorganised/disoriented has come to seem the best descriptor’.438 Main and colleagues concluded that there could be various ways that a loss might be left unresolved over time but without contributing to a disorganised/disoriented state of mind regarding the attachment-relevant cognition: ‘thus, for example, effective dismissal of the import of a loss is certainly indicative of failure of resolution of mourning (and is often referred to as “failed mourning”), but is not considered disorganised/disoriented’.439 In such a case, attention has been effectively directed away from the potentially disorganising loss, much as Main conceptualised the Group A pattern in the Strange Situation as avoidance of conflict and loss of regulation.440 A smooth surface is put in play by effective dismissal, against which grief and sorrow will ricochet.
Likewise, distressed pining after a lost attachment figure is often considered by clinicians as indicating that the work of mourning is not finished. And the term ‘unresolved’ would suggest that such cases would be included and coded. Indeed, several other researchers have advocated for the extension of the unresolved/disorganised classification to include this kind of response.441 Again, however, Main and Hesse claimed that this ‘distress is not considered evidence for a disorganised/disoriented state of mind’, though they acknowledged that the boundary is less clear in this case since weeping does entail ‘a very slight index of disorientation’.442 The distinction between preoccupation with a loss and disorganisation/disorientation is real but somewhat permeable, just like in Main’s thinking about C and D in the infant classification system, essentially since distress can be dysregulating. This implies a subtle but vital qualification. Though the category was named ‘unresolved for loss’, Main and colleagues specified in the manual that coders were not actually being instructed to identify lack of resolution. Rather, the coding protocols were situated as seeking to identify instances of the ‘disorganised/disoriented state of mind’ with regard to attachment that unresolved losses often seemed to occasion.443 Discussions of an unresolved loss in the AAI seemed to be the frequent occasion of disruptions in attentional processes and the processing of information relevant to attachment. In Main and Solomon’s work on the Strange Situation, the term ‘disorganised’ was used to refer to both behaviour and motivation. The term sought to laminate conflicting, confused, or apprehensive observable behaviour with inferences about the functioning of the (invisible) attachment system. So too in the AAI. The concept of ‘unresolved’ speech laminated observable lapses in reasoning or discourse in transcripts with inferences about the (invisible) integration and coherence of attentional processes in the retrieval and communication of information about attachment-relevant experiences.
Main and Hesse developed a nine-point scale for coding unresolved/disorganised/disoriented states of mind regarding attachment, which was based partly on the inferred extent of disruption to the retrieval and communication of information about attachment-relevant experiences, and partly on how certain the coder was that lapses of reasoning or discourse were what they were seeing in the transcript. As with the infant system, scores above 5 warranted a classification. In most instances, unresolved/disorganised/disoriented states of mind appeared as an interruption or splintering within the speaker’s characteristic forms of reasoning and discourse when the topic of loss or abuse was raised, though in some cases it could be more pervasive across the transcript. Main and Hesse conjectured that frightening experiences or circumstances surrounding or following a loss or frightening ideas regarding the relationship with the lost figure would be one pathway to disorganised/disoriented states of mind. This idea was supported by the finding that unresolved states of mind regarding attachment are moderately associated with frightening, frightened, or dissociative behaviours by caregivers towards their infants (r = .28).444 It was also supported by work by Beverung and Jacobvitz, which demonstrated that unresolved loss was twice as common when speakers perceived the loss as sudden.445 However, to the researchers’ surprise, there was no association in their sample between U/d following bereavement and age at loss, the nature of the relationship, the extent of emotional support, or the cause of death.
A fundamental study for the establishment of the U/d classification was conducted by Ainsworth and Eichberg. This was Ainsworth’s final direct involvement in empirical research. The AAI and the Strange Situation were conducted with a sample of 45 Charlottesville mother–infant dyads. Ainsworth and Eichberg found that 30 dyads of the sample had experienced loss of an attachment figure, but 20 displayed few or no markers of unresolved/disorganised/disoriented state of mind regarding attachment. A qualitative review of these transcripts revealed that these speakers had experienced various forms of social connectedness following the bereavement, whether in terms of the availability of comfort from others or in terms of their own responsibility for other family members. Only two of the mothers classified as resolved for their loss were part of dyads classified as disorganised in the Strange Situation, showing clearly that loss in itself was not a powerful predictor of infant disorganised attachment.446 In the final count, ten bereaved mothers received a U/d classification, and of these all were members of dyads classified by blind coders as disorganised/disoriented in the Strange Situation. The other five dyads who received a D classification had experienced frightening occurrences. Many of these were quite recent, such as a near-death experiences or a partner’s severe drug dependency.447 Half the mothers whose discourse received a U/d classification had a secondary classification as preoccupied. Ainsworth cited conversations with Main and Hesse that there can be a close relationship between preoccupation and unresolved states of mind. For instance, ‘In the case of a mother who is preoccupied with her early attachments, it is reasonable to suppose that the memory of her fear of the parent might sometimes intrude into everyday life, and re-evoke the anxiety; as such times the infant might find his mother’s behaviour especially frightening since there was no apparent occasion for it.’448 Ainsworth and Eichberg also found a 90% match between the other classifications: secure-autonomous state of mind in the AAI was highly associated with infant secure attachment, and dismissing state of mind highly associated with avoidant attachment. There was no association between preoccupied state of mind and ambivalent/resistant attachment.
The strong association between the Strange Situation and the AAI, evidenced in a study coded by no less than Ainsworth herself, provided a powerful consecration of the later measure. However, the results for unresolved/disorganised states of mind from the Ainsworth and Eichberg study have subsequently had to be excluded from meta-analyses, since the study has been such an outlier in terms of the strength of the association between the AAI and the Strange Situation procedure.449 Van IJzendoorn observed that, though generally regarded as a replication, the study should frankly be recognised as another exploratory work.450 In fact, this was in no way masked by the authors. Ainsworth clearly stated in the paper that, after coding the sample a first time, she repeated her coding on the basis of changes made by Main to the coding manual, which led to the perfect correlation between unresolved loss and infant disorganised attachment.451 However, there were no changes mentioned in the classification of secure-autonomous or dismissing states of mind regarding attachment. It is therefore likely that Ainsworth’s remarkable powers as an observer also contributed to the strength of the association.
In contrast to the 100% agreement reported by Ainsworth and Eichberg, the latest meta-analytic finding regarding the association between U/d parental discourse and the D classification in the Strange Situation based on three decades of research is r = .21. This is weaker than the associations between the other paired categories: secure-autonomous (F) discourse and secure (B) infant attachment have an association of r = .31; dismissing (Ds) discourse and avoidant (A) infant attachment have an association of r = .29; and preoccupied (E) discourse and ambivalent/resistant (C) attachment have an association of r = .22.452 There has been much discussion of factors that may ‘close the transmission gap’ by mediating between a caregiver’s state of mind regarding attachment and the classification of the caregiver–infant dyad in the Strange Situation.453 A variety of proposals have been made including: caregiver perception of the child’s behaviour in terms of intentional mental states (mentalisation); gene × environment interactions; caregivers’ support for their child’s exploration; caregivers’ use of effective limit-setting; and the nature of the repair offered following mismatches between children’s signals and their caregiver’s responses.454 However, a further proposal for the source of the transmission gap has had to do with the operation of the U/d classification, and especially in the way that unresolved traumatic experiences have been operationalised.
Between 1987 and 1989, Main and Hesse ran three training institutes for the AAI: in London (organised by John Bowlby and John Byng-Hall), in Virginia (organised by Mary Ainsworth), and in Rome (organised by Nino Dazzi and Massimo Ammaniti).455 During this time, Main and Hesse saw many new transcripts, especially those collected by clinical colleagues. These new transcripts clearly showed that disruptions of discourse suggestive of disorganised/disoriented states of mind regarding attachment could occur when speakers discussed memories besides bereavement. Besides bereavement, the other frequent occasion for lapses in discourse or reasoning seemed to be experiences of traumatic abuse. Main and Hesse’s growing attention to traumatic abuse may also be placed in the context of the increased prominence of this topic within academic psychology and wider American cultural discourses by the late 1980s.456 Until that point, Main and Hesse had advised coders to extend use of the ‘unresolved loss’ classification to encompass disorganised/disoriented states of mind about abuse: ‘researchers were advised to use the indices of disorganisation and disorientation in thought processes during discussions of a loss in order to identify unresolved trauma of other kinds’.457 One problem with this extension was that the standard AAI questions were not well adapted to exploring abuse, since the topic is only raised in quite a general way, in contrast to loss experiences which are extensively probed.458 Nonetheless, the interview questions regularly elicited lapses in monitoring of reasoning and discourse in higher-risk samples, and occasionally in lower-risk samples too.459 By the end of the 1980s, Main and Hesse had ‘now completed a draft of a separate scale for assessing unresolved experiences of physical abuse’.460 That bereavement and traumatic abuse could have the same kinds of consequences for attention and information-processing, and hence on discourse, seemed intuitive to Main and Hesse in light of their idea that the intrusion of fear in relation to an attachment figure might be a cause of both infant disorganised attachment and disorganised/disoriented states of mind. Traumatic abuse is, definitionally, frightening.
In fact the traumatic abuse scale, completed at the start of the 1990s, was not just for physical abuse but also included any experience where an attachment figure may have traumatised the child through frightening behaviour. Sexual abuse and threats to kill the child were therefore also included. Main and Hesse excluded cases where a ‘parent was described as hostile or mean at times, but without being clearly highly frightening in the view of the judge’. They also advised that traumatic abuse should not be coded if the event was a one-off, even if a child was for instance hit in the face by a parent or touched sexually by a drunk parent through clothing, if this was not ‘overwhelmingly frightening’ or ‘expected to escalate’. The coding manual suggested that an important factor in making this differentiation is the extent to which the speaker did or ‘did not fear that the parent would go out of control’.461 Main and colleagues advised that if the child was very frightened in the situation and/or afterwards, then it should be coded as traumatic abuse, even if the speaker does not now regard the behaviour as abusive.
Nonetheless, an unresolved/disorganised/disoriented (U/d) classification depended in part on how events were spoken about. Sexual or physical abuse by a parent, spoken about with ‘continuing pain and regret’ but without lapses in reasoning or discourse would not receive an unresolved/disorganised/disoriented classification.462 An important example given in the coding system was ‘alternating clear report of abuse with denial that it was abuse’. This had an important paradigmatic status in the manual, much like statements that imply that a dead person is still alive in scale for unresolved loss, since it suggested direct conflict between segregated accounts of reality.463 Other markers of unresolved traumatic abuse given in the coding system included: feelings of having personally deserved abusive treatment by an attachment figure; confusion by speakers between themselves and the perpetrator; as well as other signs similar to unresolved mourning such as invasion of the topic of abuse into discussion of other matters.
In both the case of loss and abuse, Main and Hesse specified that the classification cannot be given if there is no specific and identifiable historical event mentioned in the transcript.464 Additionally, in the case of abuse but not loss, the abuse described must be established as sufficiently severe and overwhelmingly frightening. Hence there may be experiences that, despite appearances, may not be coded as unresolved even if relevant disturbances of reasoning or discourse are present. These include: a parent leaving and never being seen again (but presumably remaining alive); anticipated bereavement in the case of a terminally ill attachment figure; experiences of chronic neglect in childhood; exposure to an atmosphere of emotional abuse between parents during childhood but without there being a stand-out ‘event’; and fear of an intimidating current partner or ex-partner.465 Similarly, it is not clear what status repeated hospitalisations has within the AAI coding system in terms of making an unresolved classification, even though hospitalisation was the foundational experience of trauma and loss in the emergence of attachment theory (Chapter 1). Pervasively frightening, frightened, or dissociative caregiving may contribute to a disorganised attachment classification in the Strange Situation and controlling behaviour on reunion at age six; but if the child from this dyad grows up without a specific bereavement or identifiable subsequent trauma that can be relayed discretely to the interviewer, it would be impossible for them to receive an unresolved classification on the AAI in adolescence or adulthood.466
In the coding manual, Main and colleagues advised coders to scale unresolved traumatic abuse and unresolved loss separately, and then to draw on both ratings in making a judgement regarding whether a transcript should be placed in the U/d category. Seen in wider context, this would appear a very surprising decision: abuse and bereavement are generally not treated as psychological experiences of a kind and of equivalent significance. However, from Main and Hesse’s perspective, what abuse and loss have in common is that an experience of the attachment figure may make thinking about the attachment figure alarming and overwhelming, causing both the aversion and the intensification of attention. The scales for unresolved traumatic abuse and unresolved traumatic loss are therefore situated in the manual as means to the end of a categorical judgement regarding whether an unresolved/disorganised/disoriented state of mind regarding attachment is present to a marked degree. As such, following Main and Hesse’s lead, the relative distribution of unresolved loss and unresolved traumatic abuse in studies is rarely available in published studies; only the overall U/d category is usually reported.467
However, a study by Fonagy and colleagues found that the association between unresolved traumatic abuse in a parent’s AAI and infant disorganised attachment classification in the Strange Situation was higher than the usual association between U/d on the AAI and disorganised attachment in the Strange Situation.468 Though acknowledging the need for further replication, on the basis of these findings Fonagy and colleagues suggest that the intrusion of unresolved traumatic experiences into everyday functioning may be a more potent cause of frightening or dissociative behaviours by caregivers towards their child than an unresolved loss.469 These conclusions are in line with a little-discussed finding reported by Main and Hesse from their Berkeley sample that lapses in reasoning and discourse in the AAI occurred far more frequently in discussions of abuse experiences than in discussions of loss.470 In 1995 Main also offered the provocative hypothesis that infant disorganised attachment will usually resolve by adulthood unless the basis of the disorganisation lies in traumatic abuse: ‘So long as direct maltreatment is not involved, many, perhaps most, are expected to have become ‘organised’ by adulthood, being either secure, dismissing or preoccupied’.471 A decade later, a follow-up conducted by Main and Hesse with 44 of the Berkeley sample at age 19 was consistent with this hypothesis. There was no association between infant disorganised attachment and an unresolved/disorganised classification on the AAI in this low-risk sample.472 However, a more adequate appraisal of Main’s hypothesis would require a cohort study including participants with abuse experiences.
Several criticisms of the operationalisation of unresolved traumatic abuse have been raised. George and Solomon, among others, criticised the coding system for unresolved/disorganised traumatic abuse as too limited. In their sample, most children who displayed controlling-punitive and controlling-caregiving behaviours on reunion had caregivers who were classified U/d on the AAI. However, examination of the cases where caregivers received a different classification revealed transcripts showing lapses in reasoning or discourse—but about kinds of events not included by the system, such as frightening occurrences in the immediate life of the speaker rather than in childhood. This was also observed by Ainsworth and Eichberg. George and Solomon concluded that ‘our findings suggested that trauma should be defined as events that leave the individual feeling helpless and out of control, including current/recent traumatic events in the caregiving relationship’.473
Levinson and Fonagy also criticised the boundaries placed by Main and Hesse around the definition of traumatic abuse.474 In a study in the 1990s of individuals incarcerated for violent crimes, they found that the prisoners reported histories of severe and appalling abuse in childhood. However, in the AAI their participants systematically dismissed the importance of these experiences, and so could not be coded as unresolved according to the coding protocols. Levinson and Fonagy argued that these experiences are, in fact, best regarded as indicating unresolved/disorganised states of mind regarding attachment, and that both unresolved and dismissed distress had contributed to the capacity of these speakers for callous and violent behaviour.475 (In practice, contemporary coding norms would now likely place the Levinson and Fonagy participants as ‘Cannot Classify’ rather than dismissing, since a truly dismissing transcript would not report abuse in a way that would seem severe and appalling. But this is an evolution in the culture of coding stemming from Hesse’s work on the ‘Cannot Classify’ category, rather than reflecting a change to the manual.) Like Levinson and Fonagy, Lyons-Ruth and colleagues criticised the boundaries of the unresolved classification. They proposed that it should be extended to encompass discourse suggesting unresolved/disorganised states of mind where no bereavement or specific trauma is identifiable. They developed an additional ‘Hostile/Helpless’ coding system to identify unresolved/disorganised states apparent especially in the derogation of attachment figures or of speakers themselves, or in strong identification with a hostile or a helpless caregiver.476 Fonagy, Target, Steele, and Steele likewise expanded the boundaries of the unresolved classification in their Reflective Functioning Scale, which includes assessment of ‘unintegrated or bizarre statements, suggesting a lapse in reasoning, without this being around the topic of a bereavement or trauma’.477
However, Main and Hesse have been reluctant to alter the AAI coding system much in the past two decades. There are likely a few reasons for this. They perceived such changes as risking the capacity to commensurate empirical studies over time using the AAI. Furthermore, they clearly felt that it would be better to be overcautious in defining traumatic experiences, given that the concept of ‘trauma’ has seen such widespread and diffuse use, and their goal has ultimately been to pin down experiences that form the basis specifically for unresolved/disorganised/disoriented states of mind.478 Bowlby’s emphasis on the need to attend to ‘actual historical events’ locatable in time and space (Chapter 1) also remained an influence in the background for Main and Hesse despite their assertion that the AAI is not a veridical representation of imputed historical experiences.479 In practice, however, many but not all coders have circumvented the problem by coding cases with apparent unresolved states of mind but no locatable traumatic experiences as ‘Cannot Classify’. This then allows classification ‘by the back door’ since, by convention, ‘Cannot Classify’ cases are folded in with Unresolved cases in statistical analyses.
Yet there has been some movement in the definition of the classification. For instance, throughout the 1990s unresolved/disorganised speech regarding miscarriage and stillbirth was classified with ‘loss of pets’ as not indicating a true trauma or bereavement. However, studies by Bakermans-Kranenburg and colleagues and by Hughes and colleagues found that infant disorganised attachment was predicted by lapses in reasoning or discourse relating to experiences of miscarriage and stillbirth, when this was probed in the interview.480 Main and colleagues therefore acknowledged that these experiences ‘may be rated using the same principles, although the nature of the loss should be marked on the coding form. At present such cases should be analysed separately.’481 As a consequence, there has been some variation between laboratories in how such cases are handled, with some groups probing in interview and including unresolved trauma/loss regarding miscarriage and stillbirth as sufficient basis for a U/d classification, and others following the letter of the manual and not probing or including these experiences with the other cases in the U/d category.482 In most samples, the number of cases is small enough that these variations are not critical. Rather, the instability reflects structural tensions faced by the AAI as an instrument in attempting to pin down the diffuse concept of trauma. It also reflects differences between Main/Hesse and their critics in conceptualising the idea of ‘unresolved/disorganised/disoriented states of mind’.
Some attachment researchers, for instance Mary Target and several Italian colleagues, have accused Main and Hesse of self-contradiction and incoherence in theorising dissociation and fear.483 They urge that there are significant outstanding questions for Main and Hesse in this area. This latter point is undoubtedly true. However, the accusation of self-contradiction and incoherence is overstated: fear, trauma, dissociation, and disorganisation have quite distinct and coherent places in Main and Hesse’s theory. In 1992, Main and Hesse published a long and finely etched chapter discussing the mechanism that they saw as the basis for both infant disorganised attachment and unresolved/disorganised/disoriented states of mind regarding attachment. Central to their concerns in the chapter was the idea of ‘dissociation’, which had been gaining prominence in clinical and academic psychology through the previous decade, as well as public debates about ‘recovered memories’.484 Main and Hesse had their attention drawn to the concept by Giovanni Liotti, who argued for the potential value of the concept for interpreting seemingly contradictory, incomplete, or disrupted sequences of behaviour or speech.485 The 1992 chapter by Main and Hesse offered an account of how dissociation related to other key concepts in their theory: fear, trauma, and attention. However, the work was only ever published in Italian.486 Without an English translation, the chapter has not been widely known or discussed, contributing to a tendency for subsequent interpreters to treat fear, trauma, dissociation, and disorganisation as confused or as interchangeable elements in Main and Hesse’s theory.487
In the model put forward in the 1992 chapter, the attachment system in childhood, the caregiving system in adulthood, and the retrieval and communication of attachment-related experiences in adulthood have something important in common. All three are underpinned by the coordination of attention, and individual differences ‘follow upon alterations in the focus of attention’ with respect to attachment-relevant information, including external perceptions and memories.488 Attachment behaviour in infancy reflects these alterations in the focus of attention most directly, since the threshold for activation of the attachment system may be raised or lowered ‘by focusing attention either away from or toward 1) the attachment figure and 2) any cues to danger implicit in the situation’.489 Retrieval of information and communication with the interviewer in the AAI also reflects individual differences in the alteration of the focus of attention with respect to attachment-relevant information. Dismissing states of mind are underpinned by a tendency to direct attention away from attachment-relevant memories and perceptions in the past and in interaction with the interviewer. Preoccupied states of mind are underpinned by an intense focus on attachment-relevant memories and perceptions. Finally, the caregiving system is distinct from the attachment system. But the assemblage of the caregiving system incorporates some component elements from the attachment system, such as interpretations of the meanings of physical touch (Chapter 1). As such, individual differences in the alteration of the focus of attention with respect to attachment-relevant information may have some effect on the functioning of the caregiving system.
In 1990, Main and Hesse argued that an approach/avoidance conflict is produced in the Strange Situation for an infant with past experiences of a caregiver who displays alarming behaviour, since the same figure will elicit a disposition to both withdraw and approach as a safe haven, dispositions that are incompatible and mutually exacerbating. However, the behavioural account of an approach/avoidance conflict was the external face of a hypothesised mechanism occurring at the level of attention. In their 1992 chapter, the theory of this attentional mechanism was extended as a model of adult caregiving behaviour and autobiographical discourse. In all three cases, where attachment-relevant perceptions and memories are also experienced as alarming, Main and Hesse argued that attention cannot simply be turned away or turned to them since the other response intrudes. The result is a ‘looping’ of attention, and, if sustained or repeated over time, potential damage to the behavioural system itself is expectable.490 This damage might be seen as weakened regulatory capacities, localised holes or blockages in the functioning of the system, or potentially even the development of relatively independent ‘nets’ of dispositional responses ‘potentially organised with respect to one of the competing and incompatible goals’.491 In the chapter, Main and Hesse appeared ambivalent as to whether dissociation represented a kind of defence or adaptation more extreme in kind than the conditional strategies (as for Bowlby), or essentially a kind of breakdown. In any case, in reducing environmental responsiveness and the integration of information, dissociation was anticipated to have repercussions for functioning and for mental health.
In the 1992 chapter, Main and Hesse proposed a new ‘understanding of the qualitative structure of trauma’ in the context of attachment.492 Where an attachment-relevant experience is itself alarming and the looping of attention occurs, there are consequences for the encoding of the memory. The effective tagging and encoding of embodied memory, Main and Hesse supposed, requires the attentional process lost to the loop. The looping of attention inhibits the integration and semantic extraction of experiences, so that these memories may be accompanied by associations based on episodic rather than semantic resonances and may lack important contextual markers about time and place. This accounts for the unhoused, invasively intense quality of traumatic memories, and of the lapses in discourse and reasoning seen in the AAI. The common mechanism is also proposed as accounting for the fact that dissociation can be one consequence of such a wide variety of forms of trauma.
As van IJzendoorn and Schuengel among others have observed, ‘the construct of dissociation can easily be overstretched to include almost every defense mechanism’, and indeed Main and Hesse were rather unclear in their use of the term, sometimes intending a narrower ‘prototypical’ meaning and sometimes using the term loosely as a synonym for any form of mental segregation.493 Nonetheless, it is quite possible to identify their key claims about dissociation through a close reading. Main and Hesse argued that the looping of attention can cause the most prototypically dissociated responses such as fugue states where a child or an adult is unresponsive to the environment, perhaps accompanied by ‘blank, unseeing eyes’ or ‘upward rolls of the eyes’ or a startle back to environmental alertness.494 However, problems in the encoding of experience caused by past looping of attention can also produce segregation within behavioural systems, such that two incompatible responses may be activated without coordination in response to the same cue from memory or from the environment. This segregation was regarded by Main and Hesse as also supported by dissociative processes, even if in itself it is not reducible to dissociation and less prototypically dissociative than a fugue state.
In the chapter, Main and Hesse then applied their model one by one to frightening/frightened caregiver behaviour, to the infant Strange Situation, and to lapses in reasoning or discourse in the AAI. In relation to frightening/frightened caregiver behaviour, one of the three main categories is dissociative behaviours. Main and Hesse speculated that dissociative behaviours may occur when working memory—short-term perceptual and linguistic processing—is overwhelmed by the looping of attention around loss or abusive experiences relating to attachment figures. One potent cause of such looping is anticipated to be alarming memories of or associations with these figures. It may be prompted by unlikely objects because it has been poorly encoded, so it takes an overexpansive field of reference. This then increases the circumstances that grant uncomfortable freedom to experiences of the past, and place memory’s sharp edges up against the throat of the present. For instance, the touch of an infant may evoke poorly encoded and frightening memories for a caregiver of abusive touch by an attachment figure in childhood or adulthood.495 This may then elicit dissociative, frightening, or frightened responses by the caregiver towards the child. Or again, the features of a child may recall those of a dead attachment figure, leading to loops of attention or activation of the fear behavioural system if the deceased attachment figure and/or their passing was in some way alarming.496
Even though they may co-occur, Main and Hesse were adamant, however, that not all frightening/frightened caregiver behaviours should be reduced to the effects of dissociation. There may be dissociative processes implicated in some or many of them, but not necessarily (i) to the same degree or (ii) in the same way. Though Main and Hesse do not elaborate the point, reasons for frightening behaviour by caregivers that do not require dissociation can readily be identified. One process is acknowledged by Dozier and Bernard, who observed in their work with at-risk dyads that ‘frightening behaviours can be rewarding to parents because they are so powerful in eliciting reactions from children’.497 Another case could be when a child’s safe haven is regularly under attack in the context of domestic violence. These children may experience their caregiver’s fear as frightening, without the caregiver’s behaviour being dissociative. Main and Hesse’s claim that dissociation may not always operate in the same way is also highlighted by their acknowledgement that other affects besides fear may be implicated in looping attention and anomalous caregiving behaviours. For instance, they later noted that if a loss has been profoundly confusing, the confusion may act in ways analogous to alarm in relation to the caregiver in producing attentional loops and overwhelming working memory.498 Fonagy and colleagues suggested that other difficult feelings such as shame, guilt, anger, and disgust may be implicated in disrupting ordinary states of mind following trauma.499 However, these are not matters considered by Main and Hesse.
Appraising disorganised/disoriented attachment behaviour in the Strange Situation in light of their concern with dissociation, Main and Hesse observed that some of the behaviours used for coding infant disorganised attachment appear dissociated ‘at a phenotypic level’. One category is ‘freezing/stilling’. This, they suspected, directly reflects a lapse in serial processing in the context of looping attention. Another category Main and Hesse identified as phenotypically dissociative was ‘direct indices of disorientation’. They regarded this kind of behaviour as caused by ‘incompatible perceptions, experiences and impulses in which independent “nets” have developed and momentarily control behaviour’.500 Some undirected behaviours could also have this basis, for instance when children approach the stranger with arms raised directly on reunion with their parent. Main and Hesse also argued that the most extreme and sharply defined ‘sequential contradictory’ and ‘simultaneous contradictory’ behaviours shown towards the caregiver on reunion may represent an expression of such independent ‘nets’ of responses. For instance, ‘the infant may simultaneously scream for the parent and stretch as far out of the parent’s arms as possible with eyes cast to the side’—this seems to entail fully developed behavioural dispositions in contradiction.501 However, they identified two further kinds of behaviour listed in the Main and Solomon indices where a ‘dissociative state need not be implied’: direct apprehension of the caregiver, and forms of conflict about approach that remain environmentally responsive. They argued that ‘not all disorganised-appearing behaviour listed by Main and Solomon need imply more than momentary experiences of conflict, and expressions of conflict between approach and avoidance behaviour towards the parent need not imply the intrusion of a dissociated secondary plan or system’.502
Having considered caregiving behaviour and the Strange Situation, Main and Hesse examined the potential relevance of dissociative processes in lapses of monitoring of reasoning or discourse in the AAI. In this case, they proposed that dissociation is the proximal mechanism of most, and perhaps ‘virtually all’, lapses in reasoning and discourse in the AAI.503 In some cases, the lapses are minor and suggest merely absorption of attention during speech, conceptualised as a minor form of dissociation.504 In other cases, the lapses are more major and suggest the operation of segregated processes, as when a speaker suffers from an ‘intrusion of dissociated ideas, or holds two incompatible ideas regarding a loss or abuse experience in parallel’.505 They gave out-of-context eulogistic speech about a lost attachment figure in the AAI as an example of absorption. Such discourse suggests that the question about losses evoked a memory which has been partially or wholly processed as an immediate perception.506 The memory may have been encoded in such a way that it lacks cues for context, and/or frightening aspects of the information about the attachment figure are producing attentional loops that disturb working memory and the integration of remembering with the interpersonal demands of the present interview. By contrast, a more intense and potentially a qualitatively different form of dissociative processing may be seen in lapses in reasoning, when incompatible ideas regarding a loss or abuse experience appear to be held in parallel (e.g. dead/not dead). Usually such contradictions will be monitored and identified before they can appear in speech, or will be corrected after they occur. However, if attentional resources are tied up in loops, then monitoring can suffer as a result.507
In a 2006 article, Hesse and Main argued that future investigators should attempt to discriminate between the degree of ‘dissociative components’ in lapses of discourse and reasoning, and examine their distinct antecedents and associations with anomalous caregiving behaviours.508 More of the ‘transmission gap’ between caregiver states of mind regarding attachment and infant disorganised attachment might be closed, they argued, if the specific contribution of dissociative processes was unpicked. Furthermore, the prediction of sequelae such as later dissociative symptoms (Chapter 4) might be sharpened ‘through an examination of subtypes of disorganisation and disorientation. Among the likeliest candidates to be predictive of the dissociative disorders are trance-like stilling and freezing, dissociated actions, and simultaneous or rapid alternation of avoidance and resistance.’509 However, this call has not been noted by researchers in part because, in the absence of the 1992 chapter, the conceptual relationship between the U/d classification, dissociation, and fear has remained blurry. Main and Hesse hoped that the 2006 article would make their position clear, and regret that there appears to be little awareness of their account of how exactly U/d, dissociation, and fear interrelate.510 In particular, few attachment researchers seem to know that Main and Hesse argued that lapses in reasoning or discourse in the AAI have varying degrees of a dissociative basis, but that only some forms of frightening/frightened caregiver behaviour are regarded as especially dissociative. Likewise, few know that Main and Hesse signal distinctions among more dissociative, apprehensive, and conflict behaviours (and a potential difference with stereotypic behaviours) within infant disorganised attachment. As a consequence, too few studies of attachment have included measures of dissociation, in part due to lack of recognition that the relationship between dissociation, trauma, and unresolved/disorganised states rather urgently requires empirical disentanglement.
One major exception has been van IJzendoorn, Bakermans-Kranenburg, and colleagues. Hesse’s doctorate was completed under van IJzendoorn’s supervision during the 1990s, and there were regular reciprocal visits by the researchers between Berkeley and Leiden. A study by Schuengel, van IJzendoorn, and Bakermans-Kranenburg provided data to examine Main and Hesse’s proposal that the unresolved/disorganised/disoriented classification on the AAI would have a more intimate and consistent relationship with adult dissociative processes than frightening/frightened behaviour or the dyad’s classification in the Strange Situation procedure. Following a review of measures of dissociation, the researchers selected a self-report measure—the Dissociative Experiences Scale—with strong evidence of validity and reliability.511 The measure captures dimensions of depersonalisation, derealisation, and selective amnesia, in addition to absorption. In agreement with Main and Hesse’s proposal, they found that reports of dissociative experiences are more common in speakers classified as unresolved/disorganised/disoriented in the AAI.
Later research has in general terms supported this link between the Dissociative Experiences Scale and unresolved/disorganised/disoriented speech in the AAI, but found that it may hold only for unresolved traumatic abuse, not unresolved loss.512 On the other hand, Schuengel and colleagues found no significant association between the self-report measure of dissociation and either frightening/frightened behaviours or with infant disorganised attachment.513 Such findings suggest that more heterogeneous processes are in play in these latter two assessments, or that self-report cannot capture the forms of dissociation relevant to frightening/frightened behaviours or the kinds of caregiving linked to infant disorganised attachment. With so few trained coders of the frightening/frightened (FR) coding system, it is unsurprising there has been no later study to have used both this coding system and a measure of dissociation. However, it is a mark of the poor reception of Main and Hesse’s ideas about fear, trauma, dissociation, and disorganisation that no later study has used a measure of caregiver dissociation alongside the Strange Situation.514
Further interrogation of the relationship between unresolved/disorganised/disoriented states of mind regarding attachment and the construct of trauma has been pursued by van IJzendoorn and colleagues. They conducted the AAI and a clinical assessment for post-traumatic stress disorder (PTSD) with 31 combat veterans in treatment for PTSD and 29 veterans not in treatment for PTSD. The researchers added a specific probe about combat experiences to the AAI schedule, and coded both combat-related lapses in reasoning or discourse and the non-combat-related lapses that are usually coded in other samples. Rates of secure-autonomous attachment did not differ between the groups. However, lapses in reasoning or discourse on the AAI in discussions of combat were found to be so strongly associated with PTSD symptoms that it was almost as if they were the same construct (r = .80):
Bakermans-Kranenburg and van IJzendoorn also reported meta-analytic findings that almost all adults with PTSD across different samples are classified as unresolved (U/d).516 They aligned these findings with Main and Hesse’s proposal that some forms of unresolved/disorganised discourse are caused by the intrusion of poorly processed perceptions which ‘may disrupt attention … in the form of absorption and unmonitored intrusions of memories, affects and sensory perceptions concerning the trauma’.517 However, they warned that ‘there may be an asymmetric relation in the sense that not all AAI unresolved trauma involves PTSD, while PTSD would almost always involve AAI unresolved trauma’, at least when the trauma is probed for and coded.518
Van IJzendoorn and colleagues also examined lapses of reasoning or discourse appearing in discussion of experiences unrelated to combat. Whereas in the control group 7% displayed lapses in reasoning or discourse relating to these other experiences, 42% of the group in treatment for PTSD displayed U/d markers.519 This finding was interpreted as suggesting that an unresolved state of mind regarding non-combat traumas may have predisposed vulnerability to lack of integration in mental processing of combat experiences. Traumas and losses may layer on top of one another, with each contributing to greater vulnerability to unresolved/disorganised states of mind. However, the researchers were led to reflect further on the image of segregated processing or dissociation solely as a risk factor by their later research. In a study of Holocaust survivors and a group of rigorously matched control participants, they found, as expected, that dissociative symptoms were more common among the Holocaust survivors. Yet there were no differences in the physical, psychological, or cognitive functioning in the children of the survivors compared to the control. Van IJzendoorn and colleagues proposed that some form of dissociation may here be serving, actually, as a protective factor: ‘Removing traumatic memories from one’s mind may result in reduced hyper vigilance, normal cortisol levels, and reduced fight or flight responses, all of which might be adaptive’, and indeed may reduce the display of frightened/frightening behaviours towards children.520 This conclusion is supported by recent work on the Dozier and colleagues Attachment and Biobehavioural Catch-up intervention, in which dissociative behaviour by caregivers towards their infants was strongly negatively associated with emotional dysregulation in response to frustrating tasks when the children were aged three and four.521
In another later paper, van IJzendoorn and collaborators added a further qualification to the image of early unresolved loss or trauma predisposing U/d for combat experiences. A study of 184 twins revealed that genetic factors accounted for around half of variance in dissociative symptoms, suggesting a role for genetic factors; however, the contribution of genes associated with dissociation was intensified for individuals who had experienced trauma.522 This suggests a complex gene × environment interplay in the activation of dissociative processes. Genetic contributions to dissociation may then be implicated in the higher rates of lapses of reasoning and discourse among both combat and non-combat experiences in veterans with PTSD. However, an alternative/additional explanation could be that the traumatic combat experiences activated a latent predisposition towards dissociation, undermining the retrieval of and communication about earlier non-combat-related experiences in the AAI.
Van IJzendoorn and Bakermans-Kranenburg punned that the relationship between trauma, dissociation, and the unresolved classification is an ‘unresolved issue’ for the field.523 Like an unresolved loss or trauma, the issue is barely recognised among researchers and the audiences of attachment research, if at all, and reflects a lack of integration of different kinds of information. Yet despite their punning, van IJzendoorn and Bakermans-Kranenburg clearly regarded this as a serious problem. Unless matters are clarified by further research, they even wondered whether lack of resolution for trauma on the AAI ‘shows sufficient incremental validity beyond established measures for posttraumatic stress symptomatology’ to be worth continuing to use.524 The appearance of unresolved discourse in the AAI could be an expression of PTSD, or—more likely—of some part of this somewhat heterogenous phenomenon.525 Furthermore, van IJzendoorn and Bakermans-Kranenburg anticipate that clarification of the relationship between the AAI and PTSD will be hindered by the fact that the AAI protocol does not adequately probe the topic of trauma. This is a point likewise made by Riber, and by Bailey, Moran, and Pederson.526 Van IJzendoorn and Bakermans-Kranenburg, Riber, and Bailey and colleagues have all responded to this predicament by making adaptations to the questions asked in the AAI. Other researchers have not, or at least have not reported doing so. This is likely to have contributed to variability in reports of the association between the AAI and PTSD symptoms.
Another finding relevant here is that in a study of patients with a personality disorder diagnosis, Fonagy and colleagues documented that whilst participants have a more difficult time identifying their own unresolved losses, 87% of participants who regarded themselves as traumatised were rated as unresolved/disorganised/disoriented on the AAI.527 If replicated by other studies, this high level of agreement between self-report of post-traumatic stress and unresolved trauma in the AAI would align with van IJzendoorn and Bakermans-Kranenburg’s question about incremental validity.528 One forthcoming study offers relevant findings. Whereas childhood PTSD is usually associated with lower hippocampal volume, Cortes Hidalgo and colleagues found that infant disorganised attachment was associated with larger hippocampal volume (d = .21) compared to other participants.529 However, this is as yet only limited and preliminary evidence. The questions raised by van IJzendoorn, Bakermans-Kranenburg, Pederson, and others regarding how the U/d classification for the AAI relates or adds to existing assessments of PTSD remain an area for future research.
Another outstanding question left by the work of the Berkeley group has been the status of depression in relation to the Berkeley assessments of attachment. Depression has mostly been considered in relation to the AAI or in the mothers in dyads seen in the Strange Situation. Given that few infants or toddlers receive a diagnosis of depression, there has been little pressure for Main and colleagues or the wider field of attachment research to consider how depression in a young child might interact with or reflect disruption to the attachment system. In general, apparent symptoms of depression in children under three have been regarded by the scientific community as reflecting parental mental health rather than a quality or property of the child.530 The Berkeley six-year measures have not been in wide circulation, so there has been almost no study of their relationship with depression in middle childhood.531 One of the few such studies was conducted by Gullone and colleagues, who used the family drawing system with 326 children aged eight to ten. The researchers found a weak but material association between attachment insecurity and the child’s report of depressive symptoms (r = .25).532
Whereas the contribution of child depression to disruption of the child’s attachment system has been understudied, there have, by contrast, been numerous studies examining the association between maternal depression and infant attachment in the Strange Situation. This was a topic of special and widespread interest, since Bowlby’s emphasis on caregiver unavailability was readily extrapolated to maternal depression as a potential source of major disruption to the infant–caregiver attachment relationship.533 Several attachment researchers from diverse traditions—ranging from Hazan and Shaver to Cicchetti, Toth, and Rogosch—claimed that disorganised attachment was an expectable result of caregiver depression.534 Studies set out to examine the question empirically. By 1999, van IJzendoorn and colleagues could report findings from a meta-analysis of 1053 dyads seen in the Strange Situation where the mother had been identified, either by clinicians or by the researchers, as showing symptoms of depression. To the surprise of the researchers, the percentage of dyads receiving a disorganised attachment classification was only 21%, not significantly different to incidence in the general population.535 Van IJzendoorn and colleagues tried separating the seven studies of caregivers recruited from community samples from the nine studies recruited specifically on the basis of a diagnosis of clinical depression. In the former group there was no association with disorganised attachment, even at trend level. In the latter group there was a very weak (r = .13) association with disorganised attachment.536
One interpretation of these findings was that the severity of depressive symptoms may play a role. However, severity of depressive symptoms has not generally been found to moderate the relationship between maternal depression and infant disorganised attachment,537 though one study reported that intermittent symptoms have a weaker effect than continuous symptoms of depression.538 Comorbidity with other mental health issues has likewise generally not been found to strengthen the link between depression and disorganised attachment.539 Tharner and colleagues hypothesised that the association between maternal depression and child disorganised attachment seen in clinical samples may be due to an interaction with environmental adversities such as poverty that might contribute to feelings of despair or overloading of demands on cognitive resources.540 Equally, however, it is possible to use the same argument to suggest that clinical status is here serving as a weak index of other factors—for instance Hofer’s ‘hidden regulators’ (Chapter 1)—with relevance to differences in infant–caregiver attachment, without depression itself making an independent contribution. Bigelow, Beebe, and colleagues have provided evidence suggesting an additional hypothesis: that the caregiver’s capacity to hold the child in mind may account for what association there is between parental depression and infant disorganised attachment.541
As mentioned earlier, a meta-analysis by Groh and colleagues found that avoidant attachment in infancy predicted later depression and anxiety symptoms. But contrary to previous narrative reviews and the expectations of attachment researchers in general, there was no association between infant disorganised attachment and later symptoms of depression.542Groh and colleagues identified a critical mismatch between theory and empirical findings. Bowlby’s primary discussions of depression, at least in print, suggested that depression occurs when a behavioural system cannot be terminated. From this account it would be expectable for unresolved loss and, probably, unresolved trauma to be associated with depression. Since the 1990s, Main’s ideas have served as the field’s dominant source of theory; though her expressive range is astonishing for a research scientist, she barely ever mentions depression. The result has been a gap in explanatory resources for considering why disorganised attachment has so little relationship with depressive symptoms, but has well-replicated associations with externalising symptoms.543 Groh and colleagues glumly conclude that ‘given the current state of the literature on attachment and internalizing symptoms, relatively little can currently be concluded with confidence’.544 Perhaps something about the attachment system or attachment-relevant information can overcome many aspects of depression and nonetheless function sufficiently effectively. Attachment researchers have generally conceptualised depression as a kind of parental unavailability, but maybe this assumption is too quick or oversimple in some way. Or perhaps only some aspects of depression interact with the attachment system and the processing of attachment-relevant information; this might be suggested by Bowlby’s distinctions between levels of defensive exclusion (Chapter 1). It seems likely that a requisite of further theory development in this area will be close attention to the constructs of disorganised attachment and unresolved/disorganised/disoriented states of mind, in order to better understand why they would have little association with depression.
See Table 3.1 in the Appendix to this chapter for consideration of key concepts discussed here.
The theory, methodology, and even the basic conceptualisation of the research object of attachment research entered a new era with the innovations introduced by Main and colleagues in the 1980s. Yet criticisms of attachment theory and research, for example by anthropologists (Chapter 2), take Bowlby and Ainsworth as their targets and rarely demonstrate direct knowledge of the work of Main and colleagues. Indeed, the ideas and approaches of the Berkeley group are surprisingly unknown in the wider academic community and among the general public, and much of what is circulated has been cut-price, simplified versions mistaken for the technical form (‘allodoxia’; Chapter 1).
One contributing factor has been Main’s approach to disseminating her work. Main has pursued little public engagement. Though she has sometimes delivered public lectures, she has never given interviews to the popular media or written articles for popular venues.545 Key ideas in print are often to be found as book chapters in now-obscure volumes (e.g. ‘Avoidance in the service of proximity’), or were not published in English (e.g. ‘Attaccamento disorganizzato/disorientato nell’infanzia e stati mentali dissociati dei genitori’, ‘Sicherheit und Wissen’). Main’s two most comprehensive statements of her perspective and methodological innovations were not published at all (Behaviour and the Development of Representational Models of Attachment, 1986; Four Patterns of Attachment Seen in Behaviour, Discourse and Narrative, 1995). Pivotal statements on the meaning of the ‘state of mind regarding attachment’ construct also remained unpublished (e.g. the AAI coding manual; ‘Interview based adult attachment classifications: Related to infant–mother and infant–father attachment’). The Berkeley–Leiden Adult Attachment Questionnaire Unresolved States of Mind scale and several empirical studies using it (Chapter 5) all remained unpublished. Main’s amendments to the protocol manual for the Strange Situation are only available in manuscript to those attending training institutes. The manuals for the FR coding system and the six-year systems remain unpublished. In accepting the criticisms raised in the present chapter, Main and Hesse have pledged that they will attempt to get more of these texts into print, to reduce their contribution to misapprehensions about their work.546 This will include the AAI coding manual, which is currently 304 pages (though no doubt will be longer by the time it hits print).
Over the past 30 years or so, the restricted circulation of the work of Main and colleagues has made it difficult for the developmental research community drawing on the theory and measures originating in Berkeley to access, understand, and scrutinise their own constructs. Over the decades, the most direct access to the ideas and methods of developmental attachment research as oriented by Main has not been through published papers, but through fortnight-long in-person trainings in the AAI and Strange Situation which serve as opportunities for enculturation in an oral culture and its interpretive framework.547 As we saw in Chapter 2, van IJzendoorn had already identified the potential risks of inaccessibility and insularity associated with an oral culture as one of the legacies of Ainsworth. One of the legacies of Main and colleagues has been an intensified reliance on an oral culture among researchers in the developmental tradition, such that full social citizenship of the community has been partly based on graduation from methods training institutes. This worked against integration between attachment researchers and the wider discipline of psychology: publications in journals such as Child Development or Psychological Bulletin by attachment researchers time and again reported results from Main’s measures, but without the space in methods sections and discussion to elaborate fully what these actually meant. Even the basic explanation of measures like the AAI already requires an unusual amount of space for empirical science journals. It has been difficult to justify yet more exposition of technical detail in order to convey the concepts and theory held within the baroque coding systems.
The limited circulation of detail about the Berkeley measures reduced—though certainly did not halt—their exposure to technical criticism, and to attempts at psychometric appraisal or modification, by researchers outside of Main’s personal network.548 It also left concepts such as ‘state of mind regarding attachment’ and ‘disorganised attachment’ circulating widely through the discipline of psychology severed from the context that would have clarified their sense.549 Neither concept, for instance, has been well understood by social psychological attachment researchers (Chapter 5), since few—if any—social psychologists have received socialisation into the oral culture of the developmentalists or gained reliability in coding their measures. The restricted circulation of Main’s texts and full detail on her ideas and measures also limited their effective reception beyond psychology. In particular, it has made it difficult for attachment assessments in the developmental tradition to enter cognate and applied areas, such as social work research and empirical inquiry in family law.550 The measures were not built or adapted for travel, and so therefore generally did not, except in simplified, summary form.
There was one significant exception to the restricted circulation of the ideas of Main and colleagues. Main published several of her important publications in clinical journals, contributing to an exceptionally warm reception for her work, especially among psychoanalytic clinicians, and more recently among social workers. Her work was also advocated to the clinical community by popularisers such as Dan Siegel and David Shemmings.551 The narrative of infant disorganised attachment as caused by frightening caregiver behaviour, and as leading to controlling-caregiving or controlling-punitive behaviour and to a wide range of mental health problems, has circulated very widely.552 The AAI has also seen some clinical application, especially among clinicians with a background in psychoanalytic practice.553 The contributions of Main and her colleagues have played a significant role, alongside the contribution of others such as Fonagy and Holmes,554 in persuading the clinical community to attend to attachment research, something Bowlby himself felt that he had failed to achieve. Chefetz, for instance, described Main and Hesse’s work as an ‘explanatory godsend’ for clinicians, offering a ‘brilliant glow’ to difficult clinical phenomena.555
Among the second generation of developmental attachment research, the work of Main and colleagues has had as much influence as that of Bowlby and Ainsworth. The Berkeley group have been described as having ‘unprecedented resonance and influence’.556 In his final years, Bowlby described Main’s contributions to attachment theory as ‘impressive’, ‘clinically sophisticated’, and ‘striking’.557 When a critic spoke out at a conference against the complexity of the descriptions of children and adults that Main was compacting within categories of infant, child, and adult attachment, Bowlby replied ‘it is a language of its own, and one well-worth learning’.558 Alongside such support from Bowlby, Main has widely been regarded as receiving the baton of method-giver from Ainsworth, who used the Berkeley innovations in theory and methodology in her final publications (Chapter 2). As such, positions assumed to be held by Main have had a powerful authority for the second generation of attachment researchers, delimiting the parameters of methodology, and therefore to an extent the parameters of theory.559 Unfortunately, the stances ascribed to Main frequently differ from, flatten, or petrify into cliché the beliefs of an actual individual, Mary Main, whose ideas have an underpinning architecture that is not widely understood.
Michael Rutter ascribed to Main and colleagues two of the five great advances in psychology contributed by research in attachment on the basis of their introduction of the infant disorganised attachment classification for the Strange Situation procedure, and the AAI.560 However, these two methodological innovations were not wholly distinct. They both reflected Main’s universal model of human emotion regulation, with special primacy given to the role of attentional processes in the formation of individual differences. As discussed in Chapter 2, the idea of behavioural systems was the basis for the Strange Situation procedure but then became interred within it. Likewise, the theory of attentional processes, the ‘guess and uncover’ approach, and the approach to evolutionary theory which formed the platform for Main’s methodological innovations were all buried inside them. It might be thought that the notion of minimising and maximising strategies has been enormously popular among attachment researchers. But—seemingly without realising that there is a distinction—the version that has come into use has been Cassidy’s notion as minimising or maximising of attachment as individual strategies, rather than Main’s version as minimising or maximising attention to attachment-relevant information as part of a repertoire of behaviours granted by human evolutionary history. Main’s attentional theory has been little recognised, and has seen almost no use, no critical scrutiny, and no operationalisation.
A long-time collaborator of Cassidy’s, Yair Ziv has considered the legacy of Ainsworth and Main, and appraised the overarching authority given to the four-category Strange Situation and AAI within the developmental tradition of attachment research. Ziv and Hotam argued that in the wider context of academic psychology, overarching theories have become rare, increasingly looked down on as departing or even distracting from the empirical task of psychological research.561 They observed that, within the field of attachment research, ‘scholars use the research tool as their guide, and, probably unintentionally, as the theory’s ‘surrogate’, authorised by the predictive power of the methods.562 Ziv and Hotam alleged that attachment research has become largely a hunt for the correlates of the Strange Situation and AAI: further theoretical developments have been blocked by the identification of methodological tools with theoretical constructs, combined with the conviction that the methodology cannot be refined without undermining the basis for a cumulative research paradigm.563 They express scepticism that methodological innovation—they give the example of the use of Ainsworth’s dimensional scales advocated by Fraley and Spieker—would really represent a threat to the possibility of cumulative attachment research, though they know that others would disagree.564
It is plausible, as Ziv and Hotam argued, that attachment research has partially circumvented the decline of theory in academic psychology by embodying theory within methodology, protecting theoretical propositions whilst supplanting the need to explicitly acknowledge or discuss them.565 However, other factors may be mentioned as having obstructed understanding of the position of Main and colleagues. These include confusions regarding the use of terminology, the complexity and gestalt-like quality to the coding of observational measures of attachment, and restrictions on the accessibility of Main and colleagues’ texts containing details of methods and theory.
The movement of Main’s ideas across disciplinary spaces has occurred at the price of a mangled and simplified image of the methods and theory of the Berkeley group. Much like the Strange Situation was misadministered in Japan by researchers without the social links or deep theoretical knowledge to scaffold its appropriate use (Chapter 2), Main and colleagues have seen misapplications of their work in child welfare contexts.566 This predicament can be regarded at least in part as resulting from a combination of the compelling, magnetic, and absorbative quality of the ideas such as attachment, fear, disorganisation, trauma, and dissociation—combined with problems in the effective circulation of the technical ideas and methods of Main and colleagues outside the communities of academic and psychoanalytic psychology. The characterisation of disorganisation as a category may also have inadvertently helped it meld with the diagnosis-focused infrastructures of clinical and welfare investigations of families. Third-generation attachment researchers have tended to avoid categories in their adaptations of attachment measures for use by practitioners, not only because dimensions often have better psychometric properties, but also to try to avoid the problems they have seen with the reception of Main and Solomon’s disorganised classification.567
Main has had health difficulties at times which have reduced her capacity to write, to respond to questions and clarify ambiguities, as well as—until recently—to offer trainings (or train a trainer) in the six-year systems or the frightening/frightened coding system. This has been compounded by her perfectionism and concern with detail, which has made it extremely difficult for Main to judge a coding system or a piece of writing finished enough to go into circulation. As a result, over the decades vast swathes of her writing has remained unpublished, misunderstandings have abounded, and there have been barely a handful of accredited coders in any of the six-year systems or the frightening/frightened coding system. The infant disorganised coding system was also left in an underdeveloped state that contributes to it being difficult to learn; again, there are now remarkably few active, accredited coders.568 Though dedicated researchers have found workarounds, such as the use of alternative assessments, the lack of training and reliable coders have obstructed lines of empirical inquiry in the developmental tradition of attachment research. For example, though Main and Hesse’s work is cited as an inspiration in many attachment-based interventions, none has used frightened/frightening caregiving as an outcome measure.569 (Main and Hesse have said that they will soon be certifying four trainers, contributing to the future availability of the FR coding system.)
Restricted circulation of knowledge of the Berkeley measures has fuelled misunderstandings of Main’s theoretical position, which is closely tied to understanding the meaning of the assessments she and colleagues developed. There are vast discrepancies between the assumed positions generally ascribed to ‘Mary Main’ regarding the four-category mode of individual differences, and the theoretical positions available on a close reading of Main’s texts and coding manuals. These discrepancies are especially located around the unrecognised technical meanings given to key elements of Main’s conceptual vocabulary, including ‘disorganisation’, ‘fear’, ‘coherence’, ‘preoccupation’, ‘unresolved’, and ‘dissociation’. Yet there have been some researchers aware of Main’s position and trained in use of her measures. One set of researchers with strong understanding and close personal ties to Main and Hesse has been the Minnesota group. The work of Sroufe, Egeland, and colleagues with the Minnesota Longitudinal Study of Risk and Adaptation would prove just as influential as the Berkeley study for shaping the second generation of attachment research, providing independent validation of the measures developed in Baltimore and Berkeley, and reporting findings to support and to qualify the claims by Bowlby, Ainsworth, Main, and Hesse.
Table 3.1 Some key concepts in Main and Hesse’s writings
|
Which of the following was found to be the least satisfying stage of life by the longitudinal Berkeley older generation study?
Pos Terkait
Copyright © 2024 ketajaman Inc.
