ABA Was Never Neutral

What the research actually shows about compliance, masking, harm, and the institutional drive to normalize autistic people.

Editor’s note: This essay draws on peer-reviewed research, meta-analytic literature, and published community surveys to examine the evidence for harm associated with Applied Behavior Analysis (ABA). It is written from a systems perspective, not a clinical one. The author is an autistic adult and institutional design researcher.

Applied Behavior Analysis is the most widely recommended, most heavily funded autism intervention in the United States. It is covered by nearly all state Medicaid programs. Pediatricians recommend it. School districts fund it. Insurance pays for it. And for decades, its status as the gold-standard treatment for autism has been treated as settled.

It is not settled.

This essay is not arguing that every ABA program is abusive, or that no autistic person has ever benefited from behavioral intervention. Some have. Modern, child-led implementations are meaningfully different from the compliance-driven models that generate the most criticism, and practitioners working in good faith within an evolved framework deserve that distinction. What this essay is arguing is narrower and, in some ways, harder to dismiss: a field that has consistently failed to look for harm cannot credibly claim it has found none. A growing body of research, including some of the most rigorous large-scale evidence syntheses in the field, is raising serious questions not just about whether ABA works, but about whether the research was ever designed to catch what it costs. Those are two different problems. This essay is concerned with both.

A Framework Built to Erase

To understand the current debate, you have to understand what ABA was designed to do.

Applied Behavior Analysis descends from the work of psychologist Ivar Lovaas, whose methods in the 1960s and 1970s included electric shock, food deprivation, and physical restraint as aversive tools to suppress autistic behavior. The 1965 Life Magazine photographs of Lovaas’s methods, depicting children being subjected to painful stimuli in the name of treatment, are not ancient history. They are the documented origin of the framework that underlies modern ABA.

Lovaas himself articulated the premise clearly: autistic children needed to be built from scratch. The goal was not to support an autistic child’s development. It was to produce a child who appeared neurotypical. That goal, critics argue, never fully left the field, even as the methods modernized and the explicit use of aversives declined.

Modern ABA programs typically emphasize positive reinforcement and describe their aims in terms of skill building. Many practitioners are genuinely working in good faith within an evolved version of the framework. But the foundational premise — that autism is a collection of behaviors to be extinguished rather than a neurotype to be supported — is still embedded in how ABA treatment goals are set, who sets them, and what counts as success.

The Research Was Not Designed to Find Harm

When a treatment is described as “evidence-based,” that phrase carries weight. It implies a body of research with consistent methodology, adequate controls, and systematic attention to both benefits and side effects. By those standards, the ABA literature has a problem.

Project AIM (Autism Intervention Meta-Analysis) is a multi-institution research collaboration that has conducted the most comprehensive systematic review of early childhood autism intervention research to date. Their updated 2023 analysis, published in the BMJ and covering studies from 1975 through 2021, did not find that ABA was ineffective. What they found was more structurally damaging: the quality of the evidence is too inconsistent to justify the confidence placed in it. Kristen Bottema-Beutel, a Boston College professor and one of the study’s principal authors, has stated plainly that “it doesn’t seem to be on anyone’s radar that these interventions could cause harm,” describing a field with a striking unwillingness to examine that question.

That unwillingness shows up in the data. Project AIM coded every included study for whether researchers monitored and reported adverse events. Among studies that described no monitoring procedures at all, 76% reported no adverse events. Among studies that did implement at least some monitoring procedures, only 50% reported no adverse events. When researchers looked for harm, they found it more often. When they did not look, they did not find it. This is not a minor methodological footnote. It is a structural feature of how ABA research has been produced.

The U.S. Department of Defense’s Autism Care Demonstration adds another layer of concern. This is not a small academic study with self-selected participants. It is a federal program’s own internal outcome data, covering thousands of military dependents receiving ABA services over multiple years. Its annual reports found weak evidence for ABA overall: in one analysis, approximately 76% of recipients showed no improvement after a year, roughly 9% showed measurable worsening, and only 16% showed improvement per therapist measures. Spending on ABA through the program increased significantly in the same period. Those findings do not prove that ABA caused harm in every case. But they make it very difficult to sustain the claim that the evidence base is robust.

The PTSD Association

In 2018, researcher Henny Kupferstein published a study in the journal Advances in Autism examining posttraumatic stress symptoms in autistic people who had been exposed to ABA. Surveying 460 respondents, she found that exposure to ABA predicted higher rates and more severe posttraumatic stress symptoms. Adult participants in particular showed significantly greater symptom severity with increasing exposure to the intervention.

This study has been contested, and the contestation is legitimate. The research relied on an online survey with self-selected participants, which introduces significant methodological limitations. Those limitations are real. But the response from much of the ABA research establishment was not to design better studies. It was to dismiss the finding. A 2025 letter to the editor in the same journal summarized the situation clearly: “Unless and until there is clear scientific evidence against Kupferstein’s preliminary finding of a link between ABA and PTSD, we think this finding should be taken seriously.” Seven years later, the rigorous follow-up studies that would resolve the question either way remain largely absent from the literature.

A contested finding that no one has tried to disprove is not proof of innocence. It is proof of a field that does not want to ask the question.

Compliance Is Not the Same as Wellbeing

ABA’s core mechanism is behavior modification: reinforcing desired behaviors and extinguishing undesired ones. What counts as “undesired” has historically included stimming, atypical eye contact, unconventional social responses, and other behaviors that are, from a neurodivergence perspective, natural autistic traits rather than deficits.

Critics argue that ABA often teaches masking — the suppression of visible autistic traits in order to appear neurotypical — rather than genuine skill development. The research on masking makes this concern concrete. A 2024 study published in Autism in Adulthood found that increased autistic masking is associated with increased depression, anxiety, burnout, and exhaustion. Learning to pass as neurotypical is not the same as being supported. In many cases, the research suggests, it works in the opposite direction.

There is also the question of what ABA studies actually measure. The majority of outcome research focuses on behavior change, compliance, language scores, and adaptive skill assessments — not on mental health, self-determination, or subjective wellbeing. A 2024 scoping review documented that many autistic individuals with lived experience of ABA describe their treatment as traumatic precisely because the intervention was oriented toward how they appeared rather than how they felt. Gains in eye contact and reduced stimming look like success in a study that only measures those variables. They do not tell you what the child experienced, or what they carry forward into adulthood.

Long-term outcome research compounds this problem. There is limited evidence that early intensive ABA improves adult quality of life, mental health, employment, or relationships over the long term. The field has invested heavily in measuring short-term behavioral compliance and comparatively little in following up on whether the people who received intensive intervention in childhood are doing well as adults. That gap is not accidental. It reflects what the system was designed to care about, and it tells you exactly who the research was built for.

Who Gets to Define the Problem

Running beneath the empirical debates is an ethical question that the research has largely avoided: who decides what autistic behavior needs to change, and who consents to the process?

ABA is most commonly delivered to children who have little or no meaningful assent in the treatment. Goals are typically set by clinicians and caregivers, and the autistic person’s own preferences and experience are rarely part of the formal design process. A 2025 paper in the Journal of Social Issues examined autistic adults’ lived experiences of ABA and found that the field’s focus on compliance to norms set by others — rather than self-determined goals — was central to participants’ negative experiences. The experience of being trained to perform behaviors one did not choose, in service of an appearance of normalcy one did not want, recurred across accounts.

This is a separate line of critique from whether ABA feels distressing in the moment. The ethical argument is about the structure of the intervention itself: a framework built to shape a child toward an external standard, without that child’s input, is not neutral by design. When the external standard is neurotypicality, the structural message is that the child’s natural way of being is the problem.

But there is a deeper institutional question beneath even that. When clinicians, insurers, school systems, and policymakers all define autism as a behavioral problem requiring correction, and autistic people are structurally excluded from that definitional process, the treatment goals that emerge will reliably reflect the priorities of everyone except the person receiving the treatment. This is not a failure of individual practitioners. It is how institutions that embed a narrow prototype of normal cognition produce treatment frameworks: by deciding, in advance, who the intervention is for and what success looks like, without asking the people whose lives are at stake.

Precision Over Polemic

It is worth being precise. The existing research does not prove that ABA causes PTSD in every recipient, or that no autistic person has ever benefited from behavioral intervention, or that all ABA practitioners intend harm. Some autistic adults report that ABA helped them develop skills they wanted. Modern, naturalistic, child-led ABA approaches are meaningfully different from the compliance-driven, high-intensity models that generate the most criticism. Harm is not uniformly distributed across all implementations of a framework with dozens of variations.

But what the research does establish is this: the evidence base for ABA is built on studies that largely did not look for harm. When researchers looked, they found it more often. The most cited foundational study in the field’s history has basic methodological problems that went unchallenged for decades. The largest real-world dataset shows weak outcomes and some measurable worsening. The association between ABA exposure and posttraumatic stress symptoms has not been resolved by better research. And the autistic community — the people with direct experience of the intervention — has consistently and clearly communicated concern that has been structurally excluded from the evidentiary record.

The strongest version of the harm argument is not that ABA is always abusive. It is that a field designed to measure behavioral compliance has systematically failed to measure the things that matter most to autistic people: autonomy, dignity, mental health, and the experience of being treated as a person rather than a set of behaviors to correct.

When the same experiences are repeatedly dismissed across different people and contexts, that is not randomness. That is structure.

---

Works Cited

Sandbank, M., Bottema-Beutel, K., et al. (2023). Autism intervention meta-analysis of early childhood studies (Project AIM): updated systematic review and secondary analysis. BMJ, 383, e076733. https://doi.org/10.1136/bmj-2023-076733

Kupferstein, H. (2018). Evidence of increased PTSD symptoms in autistics exposed to applied behavior analysis. Advances in Autism, 4(1), 19–29. https://doi.org/10.1108/AIA-08-2017-0016 [Expression of Concern issued by the journal; investigation ongoing as of 2025. Methodological limitations discussed in essay body.]

Evans, J.A., Krumrei-Mancuso, E.J., & Rouse, S.V. (2024). What you are hiding could be hurting you: Autistic masking in relation to mental health, interpersonal trauma, authenticity, and self-esteem. Autism in Adulthood, 6(2), 229–240. https://doi.org/10.1089/aut.2022.0115

Marshall, S. (2025). Autistic experiences of applied behavior analysis (ABA): Toward improved autistic-centered supports. Journal of Social Issues. Advance online publication. https://doi.org/10.1111/josi.70037

Najeeb, P. & Quadt, L. (2024). Autistic well-being: A scoping review of scientific studies from a neurodiversity-affirmative perspective. Neurodiversity, 2. https://doi.org/10.1177/27546330241233088

Wilkenfeld, D.A. & McCarthy, A.M. (2020). Ethical concerns with applied behavior analysis for autism spectrum disorder. Kennedy Institute of Ethics Journal, 30(1), 31–69. https://doi.org/10.1353/ken.2020.0000