Does ABA Therapy Work? What the Research Says for Autistic Children

What “ABA” does and does not mean

Applied behavior analysis is a field that uses learning principles to understand and change behavior. In autism care, the label is used for several things:

Swipe or scroll to see all columns

This table is here to help you read the research; it cannot classify a particular program. AAP notes that behavioral and developmental models have important distinctions and significant overlap. If a provider says they “do ABA,” you can ask what that looks like in practice. [2]

What do the large reviews show?

Project AIM: wide search, cautious conclusions

The 2023 BMJ update by Sandbank and colleagues pooled controlled group studies of non-drug interventions for autistic children younger than 8. It included 252 studies (reported in 289 papers) with 13,304 participants: 173 randomized controlled trials and 79 quasi-experimental studies. The authors categorized interventions as behavioral, developmental, NDBI, technology-based, and others, and grouped outcomes by domain. [3]

The main findings, in plain terms:

• When results from randomized trials were counted, behavioral interventions showed a summary effect for social-emotional or challenging behavior outcomes (Hedges' g 0.58; 95% confidence interval 0.11 to 1.06). The authors say this estimate was mainly driven by unmasked caregiver or teacher reports. [3]
• When caregiver and teacher reports were removed, no summary effect remained significant for behavioral interventions. Significant effects remained for developmental interventions on social communication and NDBIs on social communication and measures of autism diagnostic characteristics. [3]
• When outcomes at high risk of detection bias were also removed, only one significant summary effect remained: NDBIs on measures of diagnostic characteristics of autism (g 0.30; 95% CI 0.03 to 0.57). [3]
• Quality gaps were common. About 65% of outcomes were at risk of detection bias and about 46% came from caregiver or teacher reports. In most studies, participants and families could not be masked to which intervention they received. [3]

“Detection bias” means an outcome may be measured differently because the person measuring it knows who received the intervention. Parents who have invested months of effort in a program may sincerely perceive improvement that an independent assessor would not record. That does not mean parents are wrong about what they see; it means the strongest evidence needs outcomes that cannot be influenced by that expectation. [3]

Effect sizes in this review are statistical summaries. They are not a prediction of how any one child will respond, and a non-significant result does not prove an approach has no benefit. [3]

Comprehensive ABA: moderate effects on some outcomes in a smaller meta-analysis

A 2023 BMC Psychiatry meta-analysis by Eckes and colleagues examined comprehensive ABA-based interventions in 11 studies with 632 participants. Compared with treatment as usual, minimal treatment, or no treatment, it found medium effects on intellectual functioning (standardized mean difference 0.51; 95% CI 0.09 to 0.92) and adaptive behavior (0.37; 0.03 to 0.70). Language ability, symptom severity, and parental stress did not improve beyond improvements in the control groups. [4]

The authors note that children's average age at intake ranged from 2.2 to 5.5 years, most were from Western countries and without co-occurring conditions, and group assignment in most studies was based on parental preference. They caution that the findings may not apply to every child or setting. [4]

What U.S. clinical guidance says

CDC states that behavioral approaches have the most evidence for treating symptoms of autism, are widely accepted among educators and health care professionals, and are used in many schools and clinics. CDC also notes that treatments often combine behavioral, developmental, educational, and other approaches. [1]

AAP's 2020 clinical report describes early intensive behavioral intervention as supported by a few randomized trials and a substantial single-subject literature, and states that when only randomized trials are considered, few interventions have sufficient evidence to be endorsed for children younger than 12 years or for adolescents. It also lists characteristics of empirically supported interventions, including systematic assessment, family input and shared decision-making, individualized measurable goals, trained providers, and ongoing measurement of progress. [2]

Both points matter when looking at the evidence. Behavioral methods are common and have research behind them, while the strength and generalizability of that research remain debated, especially when bias is considered. [1] [2] [3]

One head-to-head trial with a sensory-based therapy

A 2025 randomized trial by Schaaf and colleagues compared occupational therapy using Ayres Sensory Integration, ABA, and no treatment in autistic children with sensory differences. Both treatment groups improved on individualized goals compared with no treatment, and the two treatment arms improved at comparable levels. The groups did not differ significantly from no treatment on a measure of daily living skills. The intervention was 30 one-hour sessions. Only the abstract was checked for this article; sample size and full methods were not reviewed. [8]

This is one trial of a short, specific program and should not be treated as a ranking of therapies. For the sensory-based approach it tested, see Sensory Differences in Autistic Children: What Research Says and When Occupational Therapy May Help.

The harm question

A 2021 review in Autism by Bottema-Beutel and colleagues examined 150 reports of non-drug interventions for young autistic children. Only 11 reports mentioned adverse events at all. Fifty-four reports described why participants withdrew; in 10 of those, the reason could be classed as an adverse event and in 8 as an adverse effect. Another 12 gave reasons too vague to classify. Only the abstract was checked. [5]

The 2023 Project AIM update repeated the finding that adverse events were poorly monitored, adding that some studies provided information suggesting they occurred. It concluded that benefits could not be weighed against potential adverse effects because monitoring and reporting were inadequate. This applies across early intervention research, not just ABA. [3]

Here's what those gaps in reporting mean:

• It does not show that ABA, or any named program, causes harm.
• It does not show that ABA is safe.
• It does mean families and clinicians have less information than they should about possible downsides, and that watching a child's distress, withdrawal, and willingness is part of responsible practice. [3] [5]

What autistic adults say

A 2026 mixed-methods systematic review by Nosova and colleagues synthesized 19 studies of autistic adults' views on psychosocial and educational interventions for young autistic children. Themes included hearing the child's voice, keeping childhood playful, celebrating autistic lives, and understanding autistic challenges. A cross-cutting theme, “Damage done,” described negative experiences among autistic people who had gone through early interventions. In the quantitative studies, participants endorsed well-being, adult support and accessibility, and reducing harmful behaviors, and did not endorse outcomes aimed at reducing autism traits. [6]

There are a few things to keep in mind. This review covered early interventions broadly, not only ABA. The included studies were varied, and the authors call for more research with diverse autistic people, including autistic parents, children, and people with higher support needs. The same review also found support for reducing harmful behaviors and for adult help. [6]

A 2026 BMJ clinical review (Penner and colleagues) similarly concludes that interventions studied for developmental outcomes have shown benefits, but that adverse events and outcomes prioritized by autistic people need to be incorporated. For this article, the abstract and the review's table defining behavioral interventions were checked; the full review was not audited line by line. [7]

What the evidence cannot tell us

• Which program is best for your child. Reviews combine results across groups, so they cannot predict how an individual child will respond. [3] [4]
• How long benefits last. The reviews discussed here mostly report outcomes near the end of intervention or short follow-ups; this article does not draw conclusions about adult outcomes. [3] [4]
• Whether results from research programs apply to community services. Intensity, provider training, and fidelity vary. AAP lists trained providers and fidelity as characteristics of empirically supported interventions. [2]
• Whether ABA is better or worse than developmental approaches. Project AIM found stronger and more consistent signals for developmental and NDBI approaches on social communication measures, but that is not a head-to-head superiority trial, and it estimated a small number of outcomes. [3]
• Whether a goal is worth pursuing. Research can estimate whether a goal changes. It cannot say whether the goal respects the child's autonomy and well-being. Autistic adults' priorities are relevant here. [6]

What you can do now

1. Start with the goals, not the label. Ask what skill, behavior, or situation you want to improve and why. Communication, safety, self-care, play, and reducing distress are different from teaching a child to hide autistic traits. [2] [6]
2. Do not delay other supports. Speech-language therapy, occupational therapy, family coaching, and school services may be considered alongside any behavioral support. See How to Get Early Intervention Services for the U.S. process and Does AAC Delay Speech? if communication is a priority. [1] [2]
3. Ask how progress and comfort will be measured. Request both skill data and information about how your child responds emotionally. [2] [3]
4. Check that you can say no or pause. A provider should welcome questions, explain methods, and adjust the plan if your child is distressed. This is a practical expectation for ethical care, not a research-derived rule. [5] [6]
5. Seek a second view when stakes are high. If a recommended schedule is very intensive or you are unsure, ask your pediatrician or another qualified team to discuss alternatives. [2]
6. Know when to act quickly. If a child loses skills they previously had, or something seems to change suddenly, contact the pediatrician promptly rather than waiting for a scheduled review. This is general safety guidance, not a conclusion from the reviews above.

If you are still working out whether a developmental assessment is needed, see Early Signs of Autism in Toddlers and Speech Delay vs. Autism. Neither is a diagnostic test, and neither article decides what therapy is right.

Frequently asked questions

The bottom line

The research does not support a simple slogan in either direction. Behavioral approaches are widely used and have a research base, but large quality-focused reviews show that much of the evidence is vulnerable to bias, harms are not well tracked, and effects differ by approach and outcome. Autistic adults ask us to weigh well-being, consent, and acceptance alongside skills. A good plan names specific goals, measures both progress and distress, involves the child and family, and can change. [1] [2] [3] [4] [5] [6]