Demystifying attachment

Robbie Duschinsky with part of an untold story.

John Bowlby, the founder of attachment theory, was a prolific letter writer, corresponding regularly with leading figures across disciplines including psychiatry, developmental psychology, zoology, and cybernetics. His letters and the replies form a core part of the Wellcome Trust’s John Bowlby Archive. For the past six years I have been drawing on the extensive unpublished materials from such archives while working on a book on the untold story of attachment research…

By the 1980s, Bowlby had become quite frustrated with the concept of ‘attachment’.

He had drawn the term from the work of Anna Freud. Burlingham and Freud (1942, p.10, 47) observed that children ‘will cling even to mothers who are continually cross and sometimes cruel to them’, concluding that some form of ‘attachment of the small child to his mother’ seemed ‘to a large degree independent of her personal qualities’.

These observations appealed to Bowlby in the 1950s since they emphasised the child’s disposition to seek proximity with their familiar caregiver when alarmed – that is to say, use the caregiver as a ‘safe haven’. Bowlby was becoming convinced that this disposition was an evolutionary-based behavioural response. He was engaged in intense discussions with researchers on animal behaviour, above all Robert Hinde at Cambridge, who had observed that the young of many kinds of animals monitor the availability of their caregiver or caregivers, and cling to them when alarmed. Bowlby had previously written about the child’s ‘love’ for the caregiver or the child’s ‘tie’ to their caregiver. However, he wanted a more technical term, to capture the clinging behaviour human children showed when alarmed. So from 1961 onwards he adopted Burlingham and Freud’s term ‘attachment’.

This background created a fundamental problem for Bowlby: during the 1960s, he used the term ‘attachment’ in two different ways. On the one hand, he drew on Hinde and animal behaviour to cover an evolutionary-based disposition for infants to monitor access to and, especially when alarmed, seek proximity with their familiar caregiver or caregivers. On the other hand, Bowlby wanted to draw on the work of Anna Freud and the psychoanalytic tradition, as well as the connotations of the word ‘attachment’ in ordinary language, in appealing to clinical and popular audiences. He used also the term more broadly, to mean an emotionally-invested relationship that serves a symbolic source of comfort and protection.

Essentially, Bowlby wanted what he saw as the best of both worlds: the scientific motivational concepts available from the study of animal behaviour, and the clinical relevance and emotive resonance offered by psychoanalytic discourse. His struggle with the ambiguities produced for the concept of attachment are evident in his unpublished writings and correspondence. But he rarely addressed the issue in print, as he saw no way to resolve the ambiguity without sacrificing one side or the other.

‘Attachment has become a bandwagon’
During the mid-1960s, Bowlby’s colleague Mary Ainsworth, based in the United States, developed a standardised laboratory procedure. In the ‘Strange Situation’ infants face two brief separations and reunions with their familiar caregiver. These serve as prompts to elicit the child’s expectations about the availability of this caregiver under conditions of alarm, based on the child’s previous experiences of the relationship. Ainsworth identified ‘secure attachment’ as behaviour suggesting the infant’s confidence in the caregiver’s availability when alarmed, and ‘insecure attachment’ as behaviour suggesting a lack of confidence in the caregiver’s availability under such circumstances, whether because the child did not directly seek the availability of their caregiver (avoidant attachment) or because they didn’t appear comforted by their caregiver’s presence (resistant attachment).

However, already by the late 1960s, Ainsworth (1968) was writing to Bowlby with concern: ‘attachment has become a bandwagon’. She felt that the Strange Situation was a specific assessment of the infant’s expectations about their caregiver’s availability. This could be anticipated to predict aspects of the child’s later socioemotional development. But the Strange Situation was coming to be interpreted as an assessment of the whole child-parent relationship, with ‘attachment’ itself treated as residing somehow inside the child rather than being a quality of the child-caregiver relationship. Ainsworth worried, entirely correctly, that the result would be over-stated expectations for prediction across infinite domains based on the child’s behaviour in the Strange Situation.

Over time Bowlby came to ruefully acknowledge in correspondence that he kept using the word attachment ‘for purely historical reasons’ (Bowlby, 1983). This was despite the fact that it necessitated frustrating work, time and again, to clarify the distinction between the technical usage and the various connotations of the term, including adjudication between his own earlier multiple uses. He came increasingly to refer to ‘care-seeking’ rather than ‘attachment behaviour’, to distinguish the inner disposition to seek the caregiver when alarmed from its observable expression. Care-seeking behaviour was paired with and distinguished from ‘secure base’ behaviour, as observable signs of confidence in the availability of the caregiver when not alarmed.

However, Bowlby ultimately retained the term ‘attachment’ to refer to the inner disposition for both care-seeking and secure base behaviour in response to environmental cues. No doubt in part this was a result of the capacity of the term ‘attachment’ to resonate with different audiences – popular, clinical, and academic – even if there was only partial overlap in what these audiences actually understood by Bowlby’s use of it.

Attachment and evolutionary theory
For his part, Hinde felt from early on that Bowlby was making a mistake. He wrote to Bowlby directly: ‘I think that you sometimes reify the concepts that you are using for explanation as though they were mechanisms’ (Hinde, 1967). Hinde repeated this criticism on several subsequent occasions (e.g. Hinde, 1991).

The Hinde-Bowlby correspondence reveals that he had two concerns. A first was that readers would misunderstand Bowlby’s meaning. He was absolutely right. Confusion between the narrow and broad meanings has made it seem like attachment is the child-caregiver relationship, when in fact it is just one component. This has led to all kinds of problems, for example when family courts use assessments or assumptions about children’s attachment for deciding custody or care proceedings (Forslund et al., 2021). Where reference to ‘attachment’ appears within social policy, it is common to find early attachment equated with the child-caregiver relationship, which in turn is characterised as determining later development. So, for instance, in the UK, over the past decade there have been consistent appeals to Bowlby and the idea of attachment by the political right, who have argued that a policy focus on the early years then justifies cuts to other public services, with attachment security presented as an alternative to social security (Duschinsky et al., 2015).

Yet Hinde also had a second concern: the ambiguous way Bowlby was using the concept of attachment was not only setting up misrepresentations of his work, but also contributing to errors in Bowlby’s own theory, especially in conjunction with Bowlby’s weak understanding of developments in evolutionary theory subsequent to the early 1960s. This combination, Hinde believed, had led Bowlby to conclude that evolution had wired human infants to always seek proximity with their caregiver when alarmed. This seemed implausible to Hinde. He would raise the concern in print in the early 1980s, writing that natural selection would likely ‘favour individuals with a range of potential styles from which they select appropriately’ (1982, p.71). Whilst direct proximity-seeking might be regarded as the desirable response in many circumstances, Hinde emphasised that infant survival would have been more likely if they could adapt to the conditions of care in which they found themselves. They therefore needed alternative care-seeking strategies for other conditions. Hinde therefore anticipated that evolution would have given humans a repertoire of ‘conditional strategies’ for responding to caregiving environments where direct proximity-seeking was not possible or effective. The availability of conditional strategies could be anticipated to contribute to survival under such circumstances.

Avoidance and resistance
In correspondence and conversation with Hinde, ideas about conditional strategies were also developed by Mary Main, one of Ainsworth’s doctoral students. Main made the proposal that there can only be three basic ways that the attachment response can be expressed, and ultimately managed to convince Bowlby and Ainsworth of this. At first sight this claim seems, frankly, completely wild. And it is generally misunderstood as a kind of essentialism about ‘kinds of children’, as if security and insecurity were discrete boxes for categorising lives (e.g. Gaskins, 2013). This misunderstanding has been helped along by the fact that Main has left many of her key works unpublished, circulating them only in manuscript form to colleagues.

Much like for Bowlby, there is a massive disconnect between Main’s ideas and how they have generally been understood. Piecing together Main’s published and unpublished works, including two major as-yet-unpublished books, her account of attachment becomes clear, and is highly intriguing. Main suggested that both secure and insecure responses by infants in the Strange Situation could be conceptualised as ‘conditional strategies’, behavioural repertoires made available by evolutionary processes for responding to different forms of caregiving environment. An infant in a resistant dyad directs attention away from potential information that might suggest that the environment is unthreatening and that the caregiver is available. By contrast, an infant in an avoidant dyad directs attention away from potential information that might elicit alarm, distress or a tendency to approach familiar adults for comfort. Like the attachment response itself, these two strategies for using attention to modulate the attachment response were conceptualised by Main as part of our evolutionary endowment.

Main’s model of avoidance and resistance as conditional strategies situated attachment theory as a global account of human emotion and relationships. For Main two basic conditional strategies would exist for attachment, as well as other evolutionary-based behavioural responses common to humans and other animals (e.g. caregiving, sex, dominance). The output of an evolutionary-based behavioural system could be minimised or intensified, underpinned by the allocation of attention. The three Ainsworth categories were identified using 23 middle-class infants in the mid-1960s in Baltimore. But for Main they offer a microcosm of the three fundamental ways that humans, in general, may 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. This model led her to discover striking parallels to infant behaviour in the Strange Situation in the autobiographical speech of adults, in what became known as the ‘Adult Attachment Interview’.

Into adulthood
The Adult Attachment Interview is an observational assessment of adult discourse in response to questions from an interviewer about the speaker’s experiences in childhood. The measure is coded not on the basis of the speaker’s stated opinions, nor on the individual’s inferred actual history, but on the manner in which the speaker attends to and communicates about attachment-related experiences. When Main and colleagues first put forward the Adult Attachment Interview, this was understandably assumed to be an attempt to measure ‘attachment’ in adulthood. With three categories of infant attachment and three categories for the Adult Attachment Interview, it appeared to many that Main was implying that attachment could be expected to be stable across the life course, and that anything outside of these categories represented attachment pathology. In fact the manuscript history shows clearly that the name stems from the fact that it was an interview with adults about their attachment experiences, not that it intended to measure ‘adult attachment’

The Adult Attachment Interview assesses the role of attentional processes in presentation of biographical information: whether this information could be retrieved and reported in a coherent way, or whether there appeared to be minimisation or intensification of attention to attachment-relevant concerns. For Main, as for Hinde, ‘attachment’ means an evolutionary-based disposition for infants to monitor access to and, especially when alarmed, seek proximity with their familiar caregiver or caregivers. In adulthood this disposition shapes how memories of caregivers are recalled and described. Available to both infants and adults are two basic conditional strategies that can, within limits, inhibit or increase attention to attachment-related information in response to expectations about the environment. This does not imply, however, that the predominant conditional strategy used by an infant is then fixed for life. Conditional strategies appear to predicably shift over time and especially when an individual’s circumstances change, though there may be undertow from well-established patterns (Opie et al., 2020; Van Ryzin et al., 2011).

In recent work, Harriet Waters and her colleagues (Waters & Waters, 2006; Waters et al., 2020) have looked to cognitive psychology for a way to gain footing amidst the ambiguities in the concept of ‘attachment’ left by Bowlby. This approach seeks, more directly than the Adult Attachment Interview, to examine individuals’ expectations, considered as kinds of ‘secure base scripts’, about the availability of familiar caregivers or partners (e.g. Waters & Facompre, 2021). These expectations can be assessed in various ways. So, for instance, children may be presented with scenarios in which someone like them is confronted with an alarming situation, and they are asked what they would expect to happen then. Or Adult Attachment Interview transcripts can be re-coded to focus on what the speaker’s expectations appear to be about the availability of others when they need or have needed help or support.

The ‘secure base script’ approach does not claim to be a measure of ‘attachment’. Instead it represents a concerted attempt to respond to the problems that have bedevilled the concept of ‘attachment’, and return to the issue of expectations about the availability of help or support which were central to, but have been readily and frequently lost within, the work of Bowlby, Ainsworth and Main. It is also an approach more easily adapted than the Adult Attachment Interview to integration with experimental methods.

Important insight remains
In sum: problems with the concept of attachment in circulation have led to perceptions of attachment as ‘overrated’ (Elizabeth Meins in this magazine in January 2017), or even as fundamentally invalid or irrelevant. In the context of these discussions, it may be of interest that frustrations with the concept of attachment go back to Bowlby and Ainsworth and their collaborators, based in Bowlby’s attempts to bridge psychoanalysis and animal behaviour research.

The image of attachment researchers concerned to put infants into lifelong boxes stems in part from a misunderstanding of Main’s claims about the role of attentional processes. Development is a dynamic process with opportunity for change across the life course. Early relationship experiences are not viewed by attachment researchers in a deterministic way, but as providing some of the foundation for psychological and social development across various domains. Many attachment researchers today are therefore refocusing on script-like expectations about the availability of others to provide support in times of need.

Whilst caution about the term ‘attachment’ may be warranted, given its ambiguities and at times troubled history, the central idea of expectations about the availability of others as a safe haven remains an important insight and a lively area of research.

- Robbie Duschinsky is Senior University Lecturer in Social Sciences in the School of Clinical Medicine, University of Cambridge. Reporting from his study of the history of attachment research, he is the author of Cornerstones of Attachment Research (2020) and Mentalisation and Epistemic Trust (2021), both with Oxford University Press. He has also worked to make available hard-to-obtain but significant papers identified through his research, as co-editor of Trauma and Loss: Key Texts from the John Bowlby Archive (2019, Routledge) and Attachment Theory and Research: A Reader (2021, Wiley). [email protected]

Cornerstones of Attachment Research is published by Oxford University Press and is free to download from

Key sources
Ainsworth, M. (1968). Letter to John Bowlby, 27 April 1968. PP/Bow/K.4/12
Bowlby, J. (1983). Letter to Helen Block Lewis, 12 January 1983. PP/ Bow/J.9/123
Burlingham, D. & Freud, A. (1942). Young Children in War-Time. Allen & Unwin
Duschinsky, R., Greco, M. & Solomon, J. (2015). Wait up!: Attachment and sovereign power. International journal of politics, culture, and society, 28(3), 223-242.
Forslund, T., Granqvist, P., van IJzendoorn, M.H. et al. (2020). Attachment goes to court: Child protection and custody issues. Attachment & Human Development, Early View, doi:10.1080/14616734.2020.1840762
Gaskins, S. (2013). The puzzle of attachment. In N. Quinn & J.M. Mageo (Eds.) Attachment Reconsidered (pp.33-64). Palgrave Macmillan.
Hinde, R.A. (1982). Attachment: Some conceptual and biological issues. In C.M. Parkes & J. Stevenson-Hinde (Eds.) The Place of Attachment in Human Behavior (pp.60-76). Tavistock.
Hinde, R. (1967). Letter to John Bowlby, 28 June 1967. PP/Bow/K.4/11.
Hinde, R. (1991). Relationships, Attachment and Culture: A Tribute to John Bowlby. Infant Mental Health, 12(3), 154-163.
Opie, J.E., McIntosh, J.E., Esler, T.B. et al. (2020). Early childhood attachment stability and change: a meta-analysis. Attachment & Human Development, Early View, doi:10.1080/14616734.2020.1800769
Van Ryzin, M.J., Carlson, E.A. & Sroufe, L.A. (2011). Attachment discontinuity in a high-risk sample. Attachment & Human Development, 13(4), 381-401.
Waters, E., Vaughn, B. & Waters, H. (2021). Measuring Attachment: Developmental Assessment Across the Lifespan. Guilford.
Waters, H. & Waters, E. (2006). The attachment working models concept. Attachment and Human Development, 8, 185-197.
Waters, T.E.A. & Facompre, C. (2021). Measuring secure base script knowledge in the Adult Attachment Interview. In E. Waters., B.E. Vaughn & H.S. Waters (Eds.) Measuring attachment. Guilford.
Waters, H., Waters, T. & Waters, E. (2020). From internal working models to script-like attachment representations. In R.A. Thompson, J.A. Simpson & L. Berlin (Eds.) Attachment: The fundamental questions (pp.111-119). Guilford.

BPS Members can discuss this article

Already a member? Or Create an account

Not a member? Find out about becoming a member or subscriber