Published by Unseen Progress, an independent publisher of caregiver research. Last reviewed 2026-05-10. Part of the autism research overview.
Short answer. Yes — regression in the second year of life is a real and documented pattern in a meaningful subset of autistic children. Prospective studies of infant siblings followed from birth (Ozonoff and colleagues; Pickles and colleagues' UK work) show that a substantial proportion of autistic toddlers lose previously acquired words, eye contact, social responsiveness, or pointing between roughly 14 and 24 months. The regression does not mean a parent caused it, does not mean the child will not recover skills, and is not the same thing as the older "fading away" narrative. Early identification and NDBI-based intervention are the research-backed responses (Hyman, Levy, Myers, AAP 2020; Schreibman et al., 2015; Dawson et al., 2010).
The early literature on autistic regression relied on parent recall, which had obvious limitations. The more recent prospective infant-sibling studies — following younger siblings of autistic children from birth, with structured developmental assessments every few months — have produced a clearer picture. Ozonoff and colleagues' work, alongside the broader Baby Siblings Research Consortium, finds that a substantial minority of children later diagnosed with autism show measurable loss of social or language skills between roughly 14 and 24 months, with a peak window around 18–24 months. The exact proportion varies by definition and measurement, but the pattern is robust enough that contemporary diagnostic frameworks (DSM-5; AAP 2020 clinical report) explicitly acknowledge regression as one of two common onset patterns alongside early plateau without prior typical development.
Pickles and colleagues' UK longitudinal work, including the PACT trial and follow-ups, similarly documents heterogeneity in early developmental trajectories. The key finding across the prospective literature is that "regression" is not a single phenomenon — it ranges from a quiet plateau where development slows below typical pace, to a clear loss of words and social behaviours that were previously present.
Parents who describe regression most often describe one or more of the following changes between roughly 15 and 24 months:
1. Loss of words. A child who had 5–20 words stops using them, sometimes within weeks. The words may return later, or be replaced by different ones, or stay lost. 2. Loss of social responsiveness. Reduced response to name, less spontaneous eye contact, reduced shared smiling, less pointing or showing. 3. Loss of joint attention. A child who used to follow a parent's gaze or point things out stops doing so. 4. New behavioural patterns. Increased repetitive behaviours, increased sensory seeking or avoiding, increased difficulty with transitions.
The change is sometimes abrupt (over weeks) and sometimes gradual (over months). Parents often describe an inflection point that, in retrospect, marked the transition.
The research does not support any single environmental cause of autistic regression. The vaccine hypothesis has been investigated extensively across multiple large epidemiological studies (Madsen et al., 2002; Jain et al., 2015; Hviid et al., 2019) and the evidence is clear that vaccines do not cause autism or autistic regression. The timing coincidence — vaccines and the regression window both falling in the second year of life — does not survive controlled study.
The current research-supported view is that autism is highly heritable (twin studies converge on heritability around 80%) and that the second-year onset window reflects when typical development demands the social-communication and language capacities that autistic neurodevelopment makes harder. The "regression" may reflect a developmental trajectory that was always going to diverge, becoming visible only when the demands of the second year surface the underlying difference.
This is where the longitudinal research is most useful for parents. Lord, Bishop, and Anderson (2015) and the broader trajectory literature show that children who regressed in the second year reach later outcomes across the full autism range — some integrate substantial language and social skills, some develop more limited spoken language and rely on AAC, some sit between. The regression itself is not a strong predictor of long-term outcome once intervention begins. Dawson and colleagues' 2010 ESDM RCT included children with regression-pattern onset and found the intervention produced gains across the sample.
The clinical and parent-community consensus is that the period immediately after regression is the most disorienting for the family but is also when intervention has the strongest evidence base. Acting on the regression — rather than waiting to see if skills return on their own — is the research-backed move.
Across the AAP 2020 clinical report, the Schreibman NDBI consensus, and the Dawson and Rogers ESDM work, the steps with the strongest evidence are:
If a child between 14 and 24 months has lost words, eye contact, or social responsiveness, that is itself an indication for evaluation regardless of other factors. The AAP recommends not waiting through "wait and see" cycles when regression is reported. Documentation — video clips of skills before and after, dated notes — accelerates the evaluation process.
Naturalistic interaction, child-led play, embedded communication opportunities, and joint-attention scaffolding can be implemented at home from the moment regression is observed. Wetherby and colleagues' Early Social Interaction work provides a manualised parent-coaching approach with evidence in this window.
If words have been lost, AAC supports — picture exchange, simple speech-generating devices, gesture-pairing — can begin immediately. The research is clear that AAC does not impede the return of spoken language; it supports communication while the language system reorganises (Schlosser & Wendt, 2008; broader AAC literature).
Post-regression development is highly variable week-to-week. Marking specific skills now — number of words used, frequency of eye contact at meals, joint-attention bids per hour of play — and re-checking at 90 days produces a clearer signal than daily impressions can.
The research does not establish a single mechanism for regression, does not provide a reliable advance test for which infants will regress, and does not give clinicians a way to predict which post-regression children will recover the most skills. The trajectory variation is wide and the predictors are imperfect (Lord et al., 2015; Pickles and colleagues). What is settled is that early intensive NDBI-based intervention beginning as soon as concerns are documented is the research-backed response.
---
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.