ABA and behavioral therapy

Applied behavior analysis (ABA) builds a skill by breaking it into small steps and teaching each step with prompts, repetition and rewards. The psychologist Ole Ivar Lovaas developed the method with autistic children at the University of California, Los Angeles, in the 1960s.^10,30

Lovaas's best-known paper, published in 1987, compared 19 children who received intensive behavioral treatment with a separately assembled control group of 40. It reported that 47% of the treated children, 9 of the 19, reached what it called "normal educational and intellectual functioning", against 2% of the control group. The groups were not assigned at random, and reviewers have argued that the original claims could not be substantiated; later gains come mostly from studies that were not randomized, and fewer effects can be estimated once the weakest designs are excluded.^10,27,7 In May 1965 Life magazine had profiled the program in a photo essay titled "Screams, Slaps and Love", which described slaps and an electric shock given to one child.²⁶

Lovaas also co-authored a 1974 paper with the psychologist George Rekers that set out to reduce a boy's "deviant sex-role behaviors", part of the research program known as the Feminine Boy Project.¹¹ The child in that work was later identified as Kirk Andrew Murphy, who died by suicide in 2003. Later reviews of the research document the ethics concerns it raised, and the journal that published it issued an expression of concern in 2020.^28,31 No source establishes that the treatment caused his death.

Aversive punishment outlasted the early years. The Judge Rotenberg Center in Massachusetts used contingent electric skin shock for decades. In March 2020 the US Food and Drug Administration banned the devices (85 FR 13312);¹² in July 2021 a federal appeals court vacated the ban (3 F.4th 390);¹³ in March 2024 the FDA proposed the ban again (89 FR 20882); as of October 2026 that proposal has not been finalized, and the devices remain legally marketed.¹⁴ Autism trailed the wider disability-rights movement: Ari Ne'eman, writing in Disability Studies Quarterly in 2010, argued that autism advocacy had long lagged the broader disability-rights tradition.²⁹

ABA-based early intervention is the most widely used support for autistic children in the United States, and the one health insurance most often pays for.^c3,4 Almost all of the research is in young children, mostly under 8, and covers short periods.^1,2,7 In the studies that report it, the samples are mostly boys, so less is known about how these interventions work for girls.^8,9 The benefit evidence rests mostly on studies that did not randomly assign children to a group, and both the analyses that report gains and those that question them say those studies carry serious or high risk of bias.^1,2,7 No study found as of October 2026 reports adult outcomes from a randomized trial of ABA.

Two 2026 syntheses by researchers in the behavior-analytic and early-intervention field report large gains. An umbrella review by Brian Reichow and colleagues pooled six meta-analyses covering 36 unique studies and 2,241 children, reported IQ effect sizes above 0.50 and adaptive-behavior effect sizes above 0.35, and called the evidence "consistent, replicated"; the six meta-analyses share many of the same primary studies.¹ An individual-participant-data meta-analysis by Sigmund Eldevik and colleagues combined 15 studies, with data on 341 children who received early intensive behavioral intervention (EIBI) and 280 comparison children, and reported effect sizes of 0.66 for adaptive behavior, 0.87 for intellectual functioning and 1.36 for reduced autism severity. That team also reported that a higher share of the EIBI group met criteria for reliable change and scored in the non-clinical range, with a number needed to treat of 4.1 to 6.9. It writes that "all studies had a serious risk of bias due to the lack of random assignment", and concludes that EIBI "should currently be considered as the preferred treatment" for children.² The 1987 cohort itself was followed to a mean age of 11 and a half years, and that report said the treated group kept its gains, but the follow-up was by the same laboratory and was not independent.³²

Longer follow-up of independent cohorts is thinner. A 10-year follow-up of a community-based EIBI cohort found no lasting advantage: 34 of the original group of 71 preschool children were reassessed, alongside 28 age-matched comparison children who had received community treatment as usual, and although raw differences favored the comparison group on cognitive, adaptive and language functioning and on overall clinical severity, they were no longer significant once baseline IQ was controlled.³ Restricting the evidence to stronger studies changes what survives. An updated meta-analysis of 252 studies and 13,304 children still found significant effects when results were limited to randomized trials, including for behavioral interventions on social-emotional or challenging behavior; only the narrower step that also excluded outcomes at high risk of detection bias left a single significant effect, although the size of the effects was often similar across the steps. The authors report that adverse events were "poorly monitored, but possibly common".⁷ The syntheses also disagree on how much more therapy buys: the IPD team reports that higher intensity went with larger change,² while an analysis of 144 studies found no relationship between any measure of intervention amount and outcome.⁸

Some autistic adults who had behavior-analytic intervention as children report lasting distress. The study most often cited is a 2018 online survey by Henny Kupferstein, which pooled 243 autistic adults with 217 caregivers of autistic children, 460 respondents in all, recruited through social media, support groups and adult gatherings, and found that 46% of those exposed to ABA met a threshold for post-traumatic stress symptoms. It is a self-selected sample, and the adults and the caregivers were reporting about different people; its finding is correlational.⁹

A larger and more recent study used a different kind of data. Researchers at the University of Wisconsin and Ohio State examined a national database of privately insured autistic youth under 18, matching 17,120 who had an ABA record with 17,120 who did not. Youth who received ABA had 30% higher odds of a mental health hospitalization (odds ratio 1.30). The study is correlational and comes from insurance claims, which record what was billed; it found no link between how much ABA a child received and PTSD, suicidality or length of stay.⁴ The systematic reviews add that the trials themselves did not track harms well.⁷

In 2021, Kristen Bottema-Beutel and Shannon Crowley examined autism intervention articles published over one year in eight ABA journals. They found that 84% of studies had at least one author who worked as an ABA clinical provider or a paid training consultant to such providers, but that this was disclosed in only 2% of studies; among studies whose statements claimed the authors had no conflicts, 87% were written by researchers who had them.⁶ A five-year update by Bottema-Beutel and colleagues examined 119 studies and 450 authors in the same eight journals: 78% of authors had a clinical or consultancy conflict, 93% of studies had at least one such author, 8% disclosed any conflict, 2% disclosed these, and 93% of statements claiming no conflicts were false.⁵ The authors write that disclosure has not improved.⁵

ABA's defenders say the evidence is good and that the therapy is a benefit autistic people should not be denied. Writing in the 2026 umbrella review, Brian Reichow and colleagues call the pooled evidence "consistent, replicated";¹ the IPD team, led by Sigmund Eldevik, calls EIBI the "preferred treatment" for children.² The American Academy of Pediatrics, in its 2020 clinical report on autism, states that there is "an increasing evidence base to support behavioral and other interventions to address specific skills and symptoms".¹⁵ The Council of Autism Service Providers, a trade association of autism service providers most of whose members deliver ABA, presents ABA as effective and well researched and, in May 2026, disputed how a New York Times investigation portrayed its evidence base;^21,33 Travers and Tincani, writing in 2025, call the criticisms "unfounded criticism and longstanding myths".²²

The behavior-analytic professional bodies have moved on the harshest practices. The Association for Behavior Analysis International voted in November 2022 to "strongly oppose the use of contingent electric skin shock under any condition".¹⁹ The Behavior Analyst Certification Board's ethics code, effective in 2022, requires restrictive or punishment-based procedures only after less intrusive means have failed.²⁰

Its critics, many of them autistic, say the intervention is aimed at suppressing autistic behavior and that its evidence base is weaker than it is presented as being. The Autistic Self Advocacy Network, in a report dated December 2021, argues that "'new ABA' is no more acceptable or ethical than old ABA" and that "the evidence base for ABA is quite weak".¹⁶ The researchers who publish the conflict-of-interest audits are part of the same critical position,^5,6 and the autism researchers Damian Milton, in 2018, and Michelle Dawson and Sue Fletcher-Watson, in 2022, add the ethics and harms critiques.^24,23

There is a middle view. The researcher Giacomo Vivanti, in 2024, writes to reconcile the "conflicting narratives" of early intervention and neurodiversity.²⁵ Amy Gravino, an autistic woman who trained in ABA and counsels autistic students, told Garcia that she considers herself "a moderate" on the subject; Garcia draws the line at harm, writing that autistic people "must not be forced to suppress what makes them who they are".^c3

For the benefit question, what is missing is a randomized trial of comprehensive ABA-based intervention carried through to adulthood: randomized trials of early intensive behavioral intervention exist, but they are small and none follows its participants into adulthood.^34,35 That is also the design that would answer the question autistic adults care about. Such a trial would assign children to groups at random, pre-register its outcomes before it begins, follow participants into adulthood, and measure what adults say matters, such as quality of life, mental health and autonomy. A second route is an independent re-analysis of the pooled primary data: the two 2026 syntheses reuse the same non-randomized studies, and a re-analysis by a group with no ties to providers would test whether the reported gains hold.^1,2 For the harm question, what would settle it is a prospective controlled study: one that follows children who do and do not receive ABA, measures mental health with pre-registered outcomes and independent assessment, and separates the effect of the intervention from the factors that led a family to choose it in the first place.^4,9

Many autistic writers describe ABA as something done to them rather than for them. Bianca Toeps writes that the therapy "up to a few decades ago, was also commonly used to 'cure' gay people", and that "many autistics who underwent ABA describe the therapy as traumatic"; she describes sessions of up to forty hours a week in which children are told "quiet hands!" and made to look the therapist in the eye.^c1 Devon Price argues that ABA "is focused on training Autistic kids to fake a neurotypical personality" and that it teaches children "to hate themselves and obey all adults".^c2

Some autistic writers compare ABA to conversion therapy. Julia Bascom's essay "Quiet Hands" opens with being made to hold her hands still as a child and records the phrase as an instruction, not a description;^c5 Melanie Yergeau, an autistic rhetoric scholar, quotes the autistic writer Sparrow R. Jones: applied behavior analysis "equals forced compliance, total control, and constant surveillance".^c4 The comparison is contested, and not every autistic writer agrees: Eric Garcia records the view of autistic people who defend ABA and describes Amy Gravino's "moderate" position.^c3