Published by Unseen Progress, an independent publisher of caregiver research. Last reviewed 2026-05-10. Part of the autism research overview.
Short answer. ABA is a category, not a single therapy. Contemporary ABA delivered inside an NDBI framework looks substantially different from the older compliance-focused ABA models that drew the most criticism — but the label alone does not tell you which version a family is being offered. The research-backed differences are about who sets the goals, how progress is measured, whether aversives are used, and how the child's preferences shape the plan (Schreibman et al., 2015; Sandbank et al., 2020). Naturalistic Developmental Behavioural Interventions (NDBIs) are the current evidence-based default for young autistic children; older discrete-trial-only ABA is not.
Schreibman and colleagues' 2015 consensus paper formalised the NDBI category — Early Start Denver Model (ESDM), JASPER, Pivotal Response Treatment (PRT), Enhanced Milieu Teaching, and related protocols — by identifying the active ingredients shared across them. Those ingredients include child-led activities, naturalistic teaching opportunities, embedded reinforcement using the child's own motivation, and developmentally calibrated targets. The randomised-trial evidence base sits primarily on Dawson and colleagues' 2010 RCT of ESDM (showing gains in IQ, language, and adaptive behaviour at age 4 after two years of intervention), Kasari's JASPER trials on joint attention and play, and Koegel and Koegel's PRT work on motivational pivots.
NDBIs are not a rebrand of ABA. They draw on behavioural principles (reinforcement, prompting, fading) but embed them inside developmental science — meaning the targets are joint attention, symbolic play, communication, and self-regulation as developmental capacities, not isolated discrete skills.
The Sandbank et al. (2020) meta-analysis of autism early intervention is the most comprehensive recent synthesis. It found that NDBIs produced the strongest and most consistent effects on social-communication and language outcomes for young autistic children. Behavioural interventions (the broad ABA category) produced effects too, but the evidence base was weaker than commonly assumed — much of the older ABA literature has methodological limitations (small samples, lack of blinding, control-group issues) that the meta-analysis flagged.
Contemporary ABA practice has moved substantially toward NDBI-aligned principles. Many board-certified behaviour analysts (BCBAs) now deliver intervention that looks naturalistic, child-led, play-based, and neurodiversity-informed. The label "ABA" covers both this and older discrete-trial-only models that ran 30–40 hours a week of adult-led drills aiming at "indistinguishability from peers." These are not the same intervention, even if they share a regulatory category.
The criticism of ABA — much of it from autistic adults who experienced it in the 1990s and 2000s — typically targets a specific configuration: high-hour discrete-trial training, heavy adult control, suppression of stimming, use of aversives or "extinction" of behaviours that were actually communicating distress, and the explicit goal of making the child appear neurotypical (Dawson, M., 2004, The Misbehaviour of Behaviourists, separate from Geraldine Dawson; autistic-adult-led literature). This configuration is associated with worse mental-health outcomes in autistic adults, which is part of why neurodiversity-informed practitioners have moved away from it.
The point of the research distinction is not that ABA is uniformly bad or uniformly good. It is that what specifically the provider does matters more than the label, and parents can ask explicit questions to find out.
Across Schreibman et al. (2015), the AAP clinical report (Hyman, Levy, Myers, 2020), and the Sandbank meta-analysis, four questions reliably distinguish contemporary NDBI-aligned practice from older compliance-focused ABA:
A good answer is specific (joint attention frequency, spontaneous communicative initiations, tolerance of a specific transition, play-skill emergence) and measurable. A bad answer is general reassurance ("we'll work on behaviours as they come up") or "indistinguishability from peers" as a primary goal. The Schreibman consensus is clear that NDBI targets are developmental capacities measured in naturalistic settings, not isolated drills counted in trials per hour.
NDBI sessions are built around the child's preferred activities, materials, and pace. Reinforcement is embedded in the activity itself rather than tokens or treats unrelated to what the child cares about. If the answer involves a fixed daily curriculum the child is brought through regardless of interest, that is a closer match to older discrete-trial ABA than to NDBI.
Contemporary practice should not use aversives. Stimming should not be treated as a behaviour to eliminate — it typically serves a regulatory or sensory function (Prizant, 2015). A provider that frames stimming as inappropriate or targets it for extinction is not aligned with the current evidence-and-ethics consensus.
Good practice involves parents (and where age-appropriate, the child) in goal-setting. The plan should reflect what the family cares about — communication, regulation, participation, autonomy — not a pre-set curriculum the provider imports unchanged. Parent-mediated NDBI components have substantial evidence in their own right (Wetherby and colleagues' Early Social Interaction work; Kasari's parent-mediated JASPER trials).
The research does not establish that any single intervention is right for every autistic child. Lord, Bishop, and Anderson (2015) document the wide trajectory variation across the autism population; Pickles and colleagues' UK longitudinal work (PACT trial and follow-ups) reaches a similar conclusion about the heterogeneity of response. A child's profile, family circumstances, and the specific provider's skill all matter alongside the protocol label.
The research also does not establish a single "right number of hours." The original ESDM trial used roughly 20 hours per week of therapist-plus-parent-delivered intervention. Higher intensity is not always better; fit and fidelity matter more than dose past a threshold (Sandbank et al., 2020).
The label — ABA, ESDM, JASPER, PRT, "behavioural therapy" — is the start of the conversation, not the end. Parents who walk in with the four questions above, and who push for specific 90-day measurable outcomes, get a clearer picture of what the intervention will actually look like than parents who rely on the category name.
The decision to take or leave a provider is not a verdict on a whole field. It is a verdict on what this provider, this week, will be doing with this child — measured against the research-backed criteria above.
---
Unseen Progress publishes long-form caregiver research and builds research-backed daily trackers for the families covered. See the full autism research overview for the complete framework.