Attachment Was Never a Test

I knew about attachment theory before I became a mother. I had spent two years studying psychology, and I was already working as a doula. Attachment was not a new idea to me.

But becoming a mother changed my relationship to what I had learned.

Six years ago, when my son was a newborn, I began publicly questioning some of what I had learned about attachment and the way those ideas were being translated to mothers.

Four months postpartum, I left school. I was struggling to read and write the way I had before while adjusting to the cognitive, physical, and emotional demands of caring for a newborn. But I never stopped reading. I would nurse my son to sleep, put him down, and use those quiet stretches to read research papers about mothers.

I wasn’t trying to disprove attachment theory. I was trying to understand it from a position I had never occupied before.

I was the mother now.

Responsiveness. Sensitivity. Presence. Attachment. Development. These were no longer concepts I encountered only academically. They had entered my home. And I began noticing how easily knowledge about children’s development could become rules for maternal behavior.

So I started asking:

Where does the research end, and where do the rules we have built for mothers begin?

I followed that question backward: through popular parenting advice, through the theories beneath it, through the original researchers, and eventually into the critiques, revisions, cross-cultural research, and assumptions that accumulated as developmental knowledge traveled into parenting culture.

Six years later, I am still following it.

What follows is not an argument that attachment does not matter. It is an examination of what the research supports, what it does not, and what happened when a theory about children’s relationships with caregivers became something mothers could experience as a judgment of themselves.

This essay is part of a question I have been asking since I became a mother.

For many mothers, it starts somewhere like this. Not in a classroom. In a hallway.

The baby is in the crib, crying, and she is standing in the hallway outside the door with her hand flat against the wood, not opening it yet. It’s been ninety seconds. She’s counting. Somewhere in the last year she absorbed the idea that this number matters, that ninety seconds of crying unanswered is either nothing or the beginning of something, and she cannot remember which, only that she should know, only that a good mother would know without having to stand here counting.

She opens the door. She picks the baby up. They stop crying almost immediately — the way babies do, since the crying was never really about the ninety seconds. It was about being alone in a room, and that ends the moment someone walks in. She holds them and feels the specific relief of a threat that's passed, and underneath the relief, already, the next thought arriving: was that fast enough. Did that count as responsive. Is this the kind of moment that gets remembered by their body for the next forty years — filed away as evidence about whether the world is safe — and did she just fail it, standing in a hallway, counting.

This particular mother has never read Bowlby. She couldn’t tell you what a strange situation procedure is. But she is carrying his theory around anyway — or a version of it, thinned out and sharpened into something closer to a verdict: your child’s whole future rests on how correctly you respond right now, today, to this cry, and there is a wrong answer, and you might be giving it.

That is not what attachment theory says. It’s what attachment theory became, on its way from a research journal to a parenting Instagram account, and the distance between those two things is where a great deal of maternal suffering has been quietly manufactured.

What the Theory Actually Claimed, and What Got Bolted On

John Bowlby’s original work was not a morality tale. It began as a study of what happens to children under conditions of real, sustained deprivation, children in postwar institutions, children separated from every caregiver for long stretches, children with nobody responding to them at all. Mary Ainsworth took that starting point somewhere different. Her Strange Situation was not a laboratory recreation of institutional deprivation. It was a standardized separation-and-reunion procedure, built to observe how ordinary infants in ordinary homes organized their behavior around a caregiver’s brief absence and return, and it sorted what she saw into patterns, secure, insecure-avoidant, insecure-resistant, based on those responses (Ainsworth et al., 1978). It was careful, specific research, built to examine variation in everyday caregiving, not only its extremes.

What neither piece of research was ever built to describe was a single hallway, a single ninety seconds, a single ordinary Tuesday inside an otherwise attentive relationship. But that is exactly where it ended up living, translated into a culture that took a theory rooted in chronic deprivation and a lab measure of ordinary variation, and flattened both into a verdict on individual moments of normal, unavoidable imperfection, until an entire generation of mothers came to believe that a badly timed nap, a two-minute delay, a hard day where she had less patience than usual, might be quietly rewiring a child forever.

The early deprivation question asked what happens when reliable care is seriously disrupted. Parenting culture eventually began grading the ordinary imperfections of presence.

When Attachment Became a Parenting Method

There is another distinction worth making, because it explains exactly how the theory ended up living in a hallway instead of a lab.

Attachment theory is not attachment parenting. Popular culture has spent decades treating them as interchangeable, and they were never the same kind of thing. Attachment theory was a description, built from research on what tends to happen inside a caregiving relationship. Attachment parenting is a philosophy of practice, coined by pediatrician William Sears and registered nurse Martha Sears, who organized it around what they called the seven Baby B’s, birth bonding, breastfeeding, babywearing, bedding close to baby, belief in the signal value of a baby’s cry, beware of baby trainers, and balance (Sears & Sears, 2001). One is a body of findings. The other is a checklist built on top of it, using its name.

A child does not become securely attached because a mother checked the correct number of boxes. Diana Divecha, a developmental psychologist and assistant clinical professor at the Yale Child Study Center, has made this distinction plainly, drawing on the University of Minnesota’s decades-long attachment study — one of the field’s longest-running bodies of evidence: the specific practices the Sears method promotes have never been shown to predict a secure attachment on their own (Divecha, 2018). The peer-reviewed literature backs her up. A 2020 review of attachment-based parenting interventions in the Australian Journal of Psychology reached the same conclusion independently, noting that these practices may well benefit children and parents, but that the research base linking them specifically to secure attachment remains insufficient (Gregory, Kannis-Dymand, & Sharman, 2020). What the same research does show is plainer and harder to package into a list of B’s — a caregiver who is attuned, who responds, and who repairs the moments she misses. Not a caregiver who never once puts the baby down.

But culturally, the line blurred fast. Attachment stopped sounding like something researchers observed in a relationship and started sounding like something a mother was responsible for manufacturing through the correct combination of practices.

And once attachment becomes something she believes she must produce, ordinary caregiving decisions start to carry a diagnosis.

Feeding is no longer just feeding. Sleep is no longer just sleep. Putting the baby down for a nap is no longer just putting the baby down. Handing him to someone else for an hour is no longer just handing him to someone else. Each one can start to feel like evidence, filed away, about the quality of the bond she’s building.

That is where a research finding quietly turns into a performance standard. The question stops being what helps a child feel safe with the people who care for her, and becomes whether she, specifically, is doing enough of the right things to earn that safety.

Once that question lands inside a culture already built on intensive mothering, there is almost no ceiling on how closely she can begin watching herself.

The theory studied a relationship. The method sold a checklist. She has been grading herself against the wrong one.

Even Bowlby Backed Away From It

Start with something most mothers standing in that hallway have never been told. The man who built the theory did not stand behind the strictest version of it for his whole career.

The idea doing the most damage in the popular imagination is monotropy, Bowlby’s claim that a child forms one primary, uniquely important attachment, usually the mother, ranked above every other bond as a matter of biological necessity. That is the idea underneath the hallway math, the sense that only her response counts, that a grandmother or a father or a sitter is a lesser substitute rather than a full source of security in his own right. And it is the idea Bowlby himself moved away from. Working from Bowlby’s private letters, the historian Robert Duschinsky found that Bowlby came to call the strict, single-figure reading of his own work nonsense, and clarified that he had never meant a child could only be secure with one person (Duschinsky, 2020). Sarah Blaffer Hrdy, who considers herself broadly within the attachment tradition, makes the same point from the other direction: she credits Bowlby with correcting himself on this over the course of his career, under Mary Ainsworth’s influence, even as she argues he never fully escaped the assumption that the mother was the attachment figure who mattered most (Hrdy, 2009).

That is not the same as a full retraction. Bowlby stood behind the broader theory, that consistent, responsive caregiving matters, for the rest of his life. But the specific claim that turned into a mother’s private verdict on herself, one figure, one relationship, one chance to get it right, is the piece its own author walked back. The myth outlasted the man’s own second thoughts about it.

What the Wider Research Actually Found

The gap between the theory and its popular use is not a new observation. Some of the most credentialed people in the field have been pointing at it for decades, from several different directions, and the culture mostly went on without them.

Margaret Mead raised one of the earliest serious challenges, in the same 1962 volume the World Health Organization published to reassess Bowlby’s maternal deprivation claims. Mead drew on decades of fieldwork in Samoa, New Guinea, and Bali, societies in which infant care and child-adult relationships were organized very differently from the exclusive mother-infant arrangement assumed by some readings of maternal-deprivation theory, and used those comparisons to challenge the universality of its assumptions (Mead, 1962). She also pointed out something sharper: the conditions Bowlby treated as biologically necessary, one mother, constantly available, were only possible at all in a fairly modern, fairly narrow set of economic circumstances.

Sarah Hrdy came at the same gap from evolutionary biology rather than anthropology, and arrived somewhere similar. Her research on human child-rearing argues that humans evolved as cooperative breeders, not solitary ones, raising infants across a network of alloparents, grandmothers, aunts, older siblings, trusted others, who shared the physical and emotional load from the beginning of the species (Hrdy, 2009). Hrdy’s argument goes further than claiming shared care was merely harmless. She proposes that cooperative breeding created evolutionary pressure favoring increasingly sophisticated capacities for intersubjective engagement, the ability to track and respond to the intentions and mental states of more than one caregiver at once, since an infant’s survival once depended on reading several minds, not just its mother’s (Hrdy, 2009). The single, sealed mother-infant unit the myth demands was never the arrangement the species evolved inside, on Hrdy’s account, and it may not be the arrangement infants are best equipped for.

Heidi Keller picked the argument back up decades later, and pressed on the measurement itself. Cross-cultural research, including the Strange Situation’s own follow-up studies, has repeatedly found that its classifications track cultural caregiving norms as much as anything resembling universal security, German infants coded avoidant at unusually high rates in a culture that prizes early independence, Japanese infants coded resistant at unusually high rates in a culture where infants are rarely separated from a caregiver at all (van IJzendoorn & Kroonenberg, 1988). Keller has argued that in societies built around distributed, multiple-caregiver rearing, of the kind common across much of Africa and Asia, the Strange Situation may not be measuring attachment security at all, since the procedure assumes a kind of exclusive, one-caregiver bond that was never the arrangement being tested (Keller, 2022). On Keller’s account, in some caregiving ecologies the procedure may partly capture unfamiliarity with the arrangement the test assumes rather than attachment security alone.

And Jerome Kagan, one of the most decorated developmental psychologists of the last century, spent much of his later career arguing that the theory had been asked to carry more certainty than the evidence could support, that temperament and the wider conditions of a child’s life shape who they become at least as much as any classification made in a strange-situation room at twelve months old. The data assembled since has largely borne out the caution. Meta-analytic reviews find that early attachment security is only modestly associated with later outcomes, greater peer social competence (r = .19), fewer externalizing behaviors (r = −.15), somewhat fewer internalizing symptoms (r = −.07), across early development (Groh, Fearon, van IJzendoorn, Bakermans-Kranenburg, & Roisman, 2017). And when researchers went back to two of the field’s own landmark longitudinal studies and re-ran the numbers with a registered, pre-specified plan, adjusting for the demographic conditions surrounding each child, the picture held up unevenly: attachment security predicted later academic outcomes only modestly, around r = .10 to .15 in both samples, while its associations with socioemotional outcomes were small and inconsistent across the two datasets (Nivison et al., 2025). Kagan wasn’t arguing that caregiving doesn’t matter. He was arguing that a lab classification at twelve months was never going to be destiny. Attachment classification is not a forty-year psychological verdict. Associations with later outcomes exist, but they are modest, and a life unfolds inside a much larger ecology than one, of temperament, relationships, socioeconomic conditions, and everything that happens afterward.

An anthropologist, an evolutionary biologist, a cross-cultural psychologist, and one of the most eminent developmental scientists alive all said some version of the same thing. Even the theory’s own author, in his own private words, said something close to it too.

What the Myth Actually Does to a Mother

None of this reaches the hallway. She is not thinking about Mead or Kagan while she stands there counting. She is thinking, correctly and instinctively, that something matters here, and incorrectly, because nobody ever told her otherwise, that the something is a single pass-fail test she is taking right now, alone, with no retakes.

This is the actual cost of the myth, and it’s worth being specific about it, because it doesn’t look like a scientific dispute from the inside. It looks like vigilance. A mother who has absorbed the thinned-out version of attachment theory starts treating every ordinary interaction as diagnostic, every delay as potential damage, every hard day as a data point in a file that’s quietly being built against her. She loses the capacity to have a bad five minutes without it meaning something. She loses the very ordinary, very human range of being a present but imperfect caregiver, because imperfect has been redefined as a threat to the whole architecture of her child’s future self.

It also does something crueler and more specific to mothers raising children outside the narrow model the theory assumed. A mother relying on a grandmother, a sister, a rotating circle of aunties and neighbors, exactly the kind of distributed caregiving Mead documented as healthy and ordinary across most of the world, gets quietly told by the culture around her that this isn’t quite attachment enough, that the real bond is supposed to be with her, specifically, constantly, or something is being lost. She ends up guilty for accepting help that would have been unremarkable, even necessary, in most of human history.

And underneath both of these is the deepest cost, which is what the myth does to trust. A mother who believes that one miscalculated response might have lasting consequences stops trusting her own read of her child and starts outsourcing that read to a chart, a percentage, a stranger’s infographic about the four attachment styles. The theory, in its original form, was never meant to replace her judgment. Its thinned, viral version does exactly that, every day, in hallways like this one.

The science studied relationships. The myth put the mother under surveillance.

What Actually Holds Up

None of this means nothing in the original research matters, or that responsiveness is irrelevant, or that the hallway is a fiction with nothing at stake in it. What survives the critique is not nothing. Caregiver sensitivity is associated with attachment security, a large 1997 meta-analysis of 66 studies found the link real but modest, important without being the only thing that matters (De Wolff & van IJzendoorn, 1997), and a 2024 meta-analysis of over twenty thousand children found essentially the same association for fathers as for mothers, r = .26 and r = .21 respectively, further evidence that this was never a story about one uniquely irreplaceable maternal bond (Madigan et al., 2024). Attachment security itself shows real but modest associations with some later outcomes, as the numbers above show. That much holds up.

What the evidence doesn't justify is turning an individual maternal interaction into a deterministic forecast of a child's future. It can't tell her, in the specific moment, whether ninety seconds was too long. Nobody's research was ever built finely enough to measure that, and the scholars above spent long careers demonstrating exactly why. What actually predicts how a child turns out is a far wider, messier set of conditions — the temperament they were born with, the stability of the whole environment around them, the resources and support available to the people raising them — accumulated over years, not the outcome of any single hallway.

That doesn’t mean nothing she does matters. It means the thing that matters isn’t a single response. It’s the whole shape of a relationship, built across thousands of ordinary moments, most of them unremarkable, some of them late, some of them imperfect, almost none of them decisive on their own.

Return

The baby has settled against her shoulder, and the ninety seconds are already gone, already indistinguishable from every other ninety seconds in every other day of their life so far. They aren’t holding it against her. They weren’t built to. Nothing in the actual research, read honestly, says they were.

She stands there a while longer than she needs to, not because the baby needs it now, but because she does — the way a person needs a minute after putting down something heavy that turns out to have been lighter than she thought. Somewhere behind her, unheard, is a version of this same scene multiplied by a few million households — a few million mothers standing outside a few million doors, counting, certain they’re being tested by a theory that was never built to test them this closely.

It wasn’t a referendum. No study was waiting in that hallway to grade her. She has always been somebody.

That was true before she counted to ninety. It’s still true now.

References

Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment: A psychological study of the strange situation. Erlbaum.

Bowlby, J. (1969). Attachment and loss: Vol. 1. Attachment. Basic Books.

De Wolff, M. S., & van IJzendoorn, M. H. (1997). Sensitivity and attachment: A meta-analysis on parental antecedents of infant attachment. Child Development, 68(4), 571–591.

Divecha, D. (2018, May 2). Why attachment parenting is not the same as secure attachment. Greater Good Magazine.

Duschinsky, R. (2020). Cornerstones of attachment research. Oxford University Press.

Gregory, M., Kannis-Dymand, L., & Sharman, R. (2020). A review of attachment-based parenting interventions: Recent advances and future considerations. Australian Journal of Psychology, 72(2), 109–122.

Groh, A. M., Fearon, R. M. P., van IJzendoorn, M. H., Bakermans-Kranenburg, M. J., & Roisman, G. I. (2017). Attachment in the early life course: Meta-analytic evidence for its role in socioemotional development. Child Development Perspectives, 11(1), 70–76.

Hrdy, S. B. (2009). Mothers and others: The evolutionary origins of mutual understanding. Belknap Press.

Keller, H. (2022). The myth of attachment theory: A critical understanding for multicultural societies. Routledge.

Madigan, S., Deneault, A.-A., Duschinsky, R., Bakermans-Kranenburg, M. J., Schuengel, C., van IJzendoorn, M. H., Ly, A., Fearon, R. M. P., Eirich, R., & Verhage, M. L. (2024). Maternal and paternal sensitivity: Key determinants of child attachment security examined through meta-analysis. Psychological Bulletin, 150(7), 839–872.

Mead, M. (1962). A cultural anthropologist’s approach to maternal deprivation. In Deprivation of maternal care: A reassessment of its effects (Public Health Papers No. 14). World Health Organization.

Nivison, M., Caldo, P. D., Magro, S. W., Raby, K. L., Groh, A. M., Vandell, D. L., Booth-LaForce, C., Fraley, R. C., Carlson, E. A., Simpson, J. A., & Roisman, G. I. (2025). The predictive validity of the strange situation procedure: Evidence from registered analyses of two landmark longitudinal studies. Development and Psychopathology, 37(1), 147–163.

Sears, W., & Sears, M. (2001). The attachment parenting book: A commonsense guide to understanding and nurturing your baby. Little, Brown and Company.

van IJzendoorn, M. H., & Kroonenberg, P. M. (1988). Cross-cultural patterns of attachment: A meta-analysis of the strange situation. Child Development, 59(1), 147–156.