Attachment theory attempts to describe the way humans form emotional bonds with each other, particularly within long-standing relationships. Developed primarily by psychiatrist John Bowlby beginning in the 1930s and extended by his student Mary Ainsworth, it frames bonding as a developmental process shaped above all by early caregiver relationships — arguing that how a caregiver “attaches” to an infant predicts how that child’s future relationships will form.

That premise sounds plausible. It is also, from a neurodiversity perspective, deeply problematic — and for Autistic and neurodivergent people, it has caused and continues to cause direct harm.

As Autistic advocate and trainer Kieran Rose writes, the theory’s application, “if not done critically, adaptively, undismissively, thoughtfully, and with a great deal of curiosity, has the potential to cause great harm.”

A Theory Treated as Fact

Attachment theory is not established science. It is a theory — yet, as Rose observes, it has been reified into clinical truth through professional repetition and institutional inertia.

“Hear something enough times and eventually it becomes a truth, especially when it ‘makes sense.’ That may sound casually flippant, but it is what happens — and it has a tendency to happen in critical professional fields such as mental health and psychology, social work and education, who, despite being critical fields, have a tendency to be incredibly uncritical in the application of all sorts of things that have a huge impact on people’s lives.”

— Kieran Rose, Autism and Attachment Theory

The theory was constructed in a historical moment of profound limitation — without meaningful access to genetics, neurology, gender science, developmental psychology as it now exists, or cross-cultural anthropology. It was also shaped by the political and social forces of its time: sexism, racism, ableism, colonialism, and heteronormativity are not incidental to the theory’s architecture. They are baked in. The normative family structures, the mother-infant dyad as the primary unit of analysis, the Western laboratory setting as universal — these are not neutral choices. They are ideological ones.

This matters because theories treated as facts get applied. They move from academia into clinical settings, from clinical settings into schools, from schools into safeguarding decisions. When a flawed theory travels that far without being questioned, the harm it does scales accordingly.

The Blank Slate Problem

At its foundation, attachment theory assumes that children are effectively empty vessels — that social and relational selfhood is constructed from the outside in, by the quality of early caregiver attunement. Rose draws a sharp and troubling connection between this assumption and a more familiar piece of deficit ideology:

“There is a famous line from Ivar Lovaas, the man responsible for developing [ABA]: ‘You see, you start pretty much from scratch when you work with an autistic child. You have a person in the physical sense — they have hair, a nose and a mouth — but they are not people in the psychological sense. You have the raw materials, but you have to build a person.’ I’m not equating attachment theory to ABA, or EIBI, but you can hopefully recognise how some of the founding principals may be similar. The thought of children as effectively blank slates that are waiting to be filled and moulded.”

— Kieran Rose, Autism and Attachment Theory

This comparison is not made to shock — it is made to be precise. Both frameworks assume that the child’s relational selfhood is something to be built by external agents. Both locate the source of “correct” development in the authority and behavior of adults. Both erase the child’s own intrinsic relational architecture as something that deserves to be met rather than molded.

This stands directly against what neurodiversity-affirming frameworks understand: that neurodivergent people have intrinsic ways of being, processing, connecting, and relating — difference that is not deficit, not damage, not emptiness waiting to be filled.

Maternal Blame and the Refrigerator Mother

Bowlby’s theory placed the mother at the center of everything. The caregiver’s emotional attunement was framed as the primary variable determining a child’s relational future. If the attachment didn’t look typical, the mother failed.

Rose connects this directly to Bruno Bettelheim’s “refrigerator mother” hypothesis — the catastrophic and now-discredited claim that Autistic children were Autistic because their mothers didn’t love them enough.

“Parental blame, particularly against mothers, permeates social care and education narratives, especially so when it comes to Autistic children. So again, hopefully, you can see how easy it is for an uncritical adoption of a theory, or ideology, to a narrative that a professional may deem problematic because of uninformed and ignorant views.”

— Kieran Rose, Autism and Attachment Theory

The architecture of blame is structurally identical. Mother’s love and attunement — measured against a neurotypical standard she may never have been told existed — becomes the variable that explains her child’s difference. For Autistic families, this is a machine for producing guilt. Autistic parents, who are more likely to be Autistic themselves, are already navigating systems designed around neurotypical norms. Layering a theory that attributes their child’s distress to their relational failure adds weight to a burden that was never theirs to carry.

This is deficit ideology directed at caregivers. It names a person as the source of a problem that is structural, biological, environmental, and systemic. It is broken systems thinking inverted — locating the breakage inside the family rather than in the world the family is trying to navigate.

Two Years, and Then It Is Over?

Bowlby placed the make-or-break window for attachment in the first two years of life. Miss it, and the damage is done. Michael Rutter’s Maternal Deprivation Reassessed (1972) challenged this determinism directly, demonstrating that relational growth and repair are possible across the whole lifespan — that the door Bowlby closed at two years old does not stay closed.

This matters particularly for Autistic people. Many Autistic people — especially those who receive late diagnoses — find that the deepest moments of recognition, belonging, and genuine connection come not in infancy but in adulthood, often through Autistic community. A framework that locates the essential window of attachment in the first two years, and reads everything after as remediation of earlier failure, cannot account for this. It mistakes the timing of the theory’s own instrument for the timing of human connection itself.

The Strange Situation as Neurotypicality Test

The primary empirical basis for attachment styles is Mary Ainsworth’s “Strange Situation” experiment — an artificial laboratory procedure that observed infants’ reactions to brief separations from their caregiver, then classified them as securely attached, insecure-avoidant, insecure-resistant, or disorganized.

Rose identifies multiple layers of failure in this instrument.

“That experiment was used to classify infants into attachment styles (secure, insecure-avoidant, insecure-resistant, disorganised), and has been criticized for lacking ecological validity — it is an artificial laboratory setup and therefore is unlikely to accurately reflect natural attachment behaviours.”

— Kieran Rose, Autism and Attachment Theory

The experiment was conducted on WEIRD populations — Western, Educated, Industrialized, Rich, Democratic — predominantly white, middle-class, Western families. In Japan, infants consistently showed higher distress in the procedure — not because of insecure attachment, but because separation from caregivers was culturally unusual for children that age. The instrument was reading cultural norms, not attachment quality. It was ethnocentric in the 1960s; critics said so then.

For Autistic and neurodivergent children, the problem runs deeper. The Strange Situation was calibrated to neurotypical proximity-seeking behavior: looking to the caregiver for reassurance, using the caregiver as a “safe haven,” expressing distress in expected ways upon separation and relief upon reunion. An Autistic child who self-regulates through parallel play rather than proximity, who has atypical eye contact, who processes separation differently — that child gets classified as insecurely attached. The instrument cannot distinguish neurotype from relational failure. It is a neurotypicality detector masquerading as an attachment assessment.

An Autistic child showing “secure” attachment through intense interests, sensory seeking, or side-by-side presence rather than face-to-face proximity is not insecurely attached. They are attached differently, in ways the framework has no language for.

Monotropism, Sensory Safety, and a Different Way of Attaching

Here the neurodiversity-affirming critique becomes generative rather than purely critical. Rose, drawing on Monotropism theory and the work of Australian advocate Kristy Forbes, offers a framework for understanding how Autistic people may seek and construct felt safety through channels that classical attachment theory cannot see.

“For Autistic people seeking Psychological safety, who are looking to feel safe, we might look to plug into a familiar, safe-feeling electrical field given off by a safe person, like a caregiver. And if we can’t find one? How much safe energy is there to plug into in, say, a school environment, where there are multitudes of environmental and systemic factors that might make you feel unsafe, or harm you.” — Kieran Rose, Autism and Attachment Theory

Monotropism — the attentional framework developed by Murray, Lesser, and Lawson (2005) — describes the Autistic tendency toward deep, focused attentional investment, attending intensely to fewer things rather than distributing attention broadly. Rose suggests this shapes how Autistic people orient toward safety: seeking to “plug into” the sensory-energetic field of a known, trusted person. When negative experiences have accumulated — sensory needs invalidated, environments hostile, predictability absent — the predictive models built from those experiences carry that negativity forward. Safety becomes harder to locate. The baseline expectation shifts toward unsafety.

What attachment theory reads as “disorganized” or “insecure” attachment may in many cases be a coherent, adaptive response to environments that have consistently failed to provide safety. The attachment-seeking is intact. The environment has failed to meet it.

This reframe is foundational. The question is not: why can’t this child attach correctly? The question is: what would this child need the environment to provide in order to feel safe enough to connect?

Masking: The Variable Attachment Theory Cannot See

Attachment theory attempts to describe unconscious relational bonding. Masking — the developmental, unconscious, safety-seeking suppression of authentic self-expression in response to stigma — is equally unconscious, equally relational, and fundamentally absent from attachment theory’s framework.

“Attachment attempts to describe the unconscious bonds that form within interpersonal relationships, so you’d think that there may be a relationship between the two dichotomies. But no, still no. And this again, is why Attachment Theory falls short. Masking is a key component of Autistic experience, yet… nothing… So what we see when we describe ‘Attachments’ might not be the reality of what’s happening. What we think, what we see, and what is actually occurring, might not be the same thing.” — Kieran Rose, Autism and Attachment Theory

If an Autistic child is masking in the presence of an assessor — performing expected relational behavior to manage the threat posed by the professional context — then any “attachment assessment” is not measuring attachment. It is measuring the child’s capacity to perform neurotypical relatedness under social pressure. Classifying that performance as “secure” or “insecure” tells us nothing about the child’s actual relational life. It tells us whether the child has learned to hide it.

This is a particularly serious problem given that assessments of attachment frequently occur in precisely the kinds of high-pressure, unfamiliar, professionally mediated settings most likely to activate masking. The instrument and the context collude to produce data that is systematically misleading about Autistic children specifically.

The Double Empathy Problem

The most important theoretical reframe for understanding attachment and autism is Damian Milton’s Double Empathy Problem (2012). Communication and connection difficulties between Autistic and non-Autistic people are not located in the Autistic person alone — they arise from the meeting of two different neurological styles, with both parties misreading each other.

When professionals use neurotypical attachment frameworks to assess Autistic connection, they are not measuring attachment. They are measuring how closely Autistic connection resembles neurotypical connection — and finding it lacking. Not because connection is absent, but because it is different. The connection is there. The theory does not have the vocabulary to recognise it.

This is not a minor methodological caveat. It reframes the entire project of assessing Autistic attachment through standard instruments as categorically unreliable. Every classification of an Autistic person as “insecurely attached” using tools designed for and normed on neurotypical populations should be held with this question: is this measuring the quality of the person’s connection, or the distance between their connection and a neurotypical template?

What Gets Erased: Structural Factors and Interpersonal Victimisation

Classical attachment theory has almost no vocabulary for the structural conditions that shape family life. Social class, economic instability, housing insecurity, community isolation — these are not central to the framework. Their absence is not a minor gap.

Rose points to the specific implication for Autistic families:

“To be Autistic is hereditary and that it runs in families, means that the statistics around low employment rates mean that there is an increased likelihood that an Autistic child is going to be from a poorer economic background or lower social class.” — Kieran Rose, Autism and Attachment Theory

When a theory that erases structural factors is applied to families who are disproportionately navigating those structural pressures — poverty, unemployment, inaccessible systems, lack of support — the result is predictable. The structural conditions that shape the family’s capacity to provide consistent, low-stress caregiving are rendered invisible. What remains visible is the caregiver’s emotional attunement, measured against a neurotypical standard, in a moment of professional scrutiny. This is not analysis. It is blame without context.

Equally invisible through an attachment lens is a further structural reality: Autistic people face extraordinarily high rates of interpersonal victimisation from very early childhood — in environments that punish difference and reward compliance, that consistently fail to meet sensory and safety needs. The relational difficulties that professionals read as signs of insecure attachment are often rooted not in early caregiver failure but in the cumulative impact of navigating hostile environments. Locating the source of relational difficulty inside the dyad — inside the child and their primary caregiver — erases this. It makes harm look like pathology.

The Neurodiversity-Affirming Reframe

The goal is not to abandon the insight that felt safety matters. Safety matters enormously. Predictability matters. The felt sense of being known and held by someone who sees you clearly — that is real, and neurodivergent people need it at least as much as anyone.

What the neurodiversity-affirming reframe requires is understanding that Autistic people seek and experience safety through different channels — sensory, predictive, monotropic — than classical attachment theory assumes. The attachment system is not broken in Autistic people. The framework for assessing it is broken.

What looks like “insecure” or “disorganized” attachment is frequently a coherent adaptive response to environments that have failed to provide safety, or to assessments that cannot see the child clearly. The child is not failing to attach. The system is failing to meet them.

Before any professional applies an attachment framework to an Autistic person, the more honest questions are: Is this person safe? Are their sensory needs met? Do they have spaces where they can be themselves without having to perform? Have the adults around them genuinely tried to understand their communication, their sensory world, their way of connecting?

Supporting Autistic children’s connection and felt safety means following their lead: understanding their sensory profile, building predictability, recognizing parallel presence and shared interest as valid forms of closeness, and — above all — reducing the environmental and systemic sources of threat that make safe connection impossible.

The question has never been whether the child can attach correctly. The question is what the world needs to do to become a place where connection feels possible.

The Stimpunks Frame

By the Stimpunks test: attachment theory in its classical form is deficit ideology dressed in relational language. It locates pathology in the dyad, blames caregivers (especially mothers), has no structural vocabulary, and codes neurological difference as relational failure. It fails “broken systems, not broken people.” The maternal determinism is particularly pernicious for neurodivergent families who are already navigating systems that pathologize them at every level.

Where it’s been extended to account for neurotype variation, sensory profiles, and double-empathy dynamics, it becomes more defensible — but that requires discarding most of what makes it distinctively “attachment theory.”

Attachment theory in its classical form fails on every axis relevant to the Stimpunks frame:

It is deficit ideology — it classifies relational difference as dysfunction. It is individualist — it locates the problem in the dyad, not the system. It is normative — it uses a narrow neurotypical population to define what “healthy” attachment looks like. It is deterministic in a way that erases the Autistic child’s intrinsic relational and sensory architecture. It generates maternal blame with direct historical lineage to the refrigerator mother narrative. And it is epistemically fragile — a theory treated as fact, applied uncritically by professionals who wield enormous power over Autistic children and their families.

The neurodiversity-affirming reframe isn’t to abandon the insight that felt safety matters — it’s to recognize that Autistic people seek and construct felt safety through different channels (sensory, predictive, monotropic), that environments routinely fail to provide those conditions, and that what looks like “insecure attachment” is often a coherent adaptive response to systemic failure. The system isn’t broken because the child didn’t attach correctly. The system is broken because it was never designed for this child.

Resources

Kieran Rose, Autism and Attachment Theory, The Autistic Advocate
Kieran Rose and Amy Pearson, Autistic Masking: Understanding Identity Management and the Role of Stigma, Jessica Kingsley Publishers, 2023
Damian Milton, On the Ontological Status of Autism: The ‘Double Empathy Problem’, Disability & Society, 27(6), 2012
Murray, D., Lesser, M., & Lawson, W., Attention, Monotropism and the Diagnostic Criteria for Autism, Autism, 9(2), 2005
Michael Rutter, Maternal Deprivation Reassessed, Penguin, 1972

Stimpunks Foundation

Attachment Theory

▶ Table of Contents