Autism research is entering a new and pivotal phase, as growing recognition of its biological and clinical heterogeneity presents challenges to previous paradigms and the adequacy of a single diagnostic label. Moving forward, incorporating a neurodiversity lens and embracing individual variability in autism will be a necessary step to improve research, outcomes and lives.
Autism spectrum disorder comprises a group of neurodevelopmental conditions that shape how individuals experience and interact with the world, including differences in social and emotional interactions and verbal and nonverbal communication. Often referred to as simply ‘autism’, it is a label that applies to more than 1% of the global population, which places it alongside common neurological conditions in prevalence. Importantly, autism is not a neurodegenerative disorder but instead is a lifelong, multifaceted condition that encompasses wide variation in behavior, sensory processing, intellectual ability and adaptive functioning. This heterogeneity in presentation, capacity and developmental trajectories makes identifying its etiology, refining its epidemiology and delivering effective support a complex and rapidly evolving challenge.
Credit: Marina Spence
The study of autism has evolved considerably over the past century. In 1911, Eugen Bleuler used the term ‘autismus’ to describe a symptom of schizophrenia characterized by withdrawal from reality. In the 1940s, Leo Kanner identified a distinct condition in children marked by social and behavioral differences, repetitive behaviors and echolalia. Although they were foundational to child psychiatry, these early accounts framed autism narrowly, emphasizing deficits and promoting the notion of a singular ‘infantile autism’ phenotype. They also perpetuated now-discredited psychogenic theories, including the idea that ‘cold’ parenting caused the condition.
By the 1970s and 1980s, advances in psychiatric research and methodology shifted the field toward developmental and epidemiological models, and autism was formally classified as a neurodevelopmental disorder. Twin and family studies established strong heritability, decisively moving the field away from psychogenic explanations. Subsequent research has demonstrated substantial genetic diversity, with multiple genetic and molecular pathways contributing to the phenotypic complexity observed in clinical presentation. These biological differences may underpin distinct behavioral, sensory, motor and cognitive profiles. Investigators have also examined environmental factors associated with an increased risk of developing autism, including advanced parental age, prenatal exposure to pollutants and maternal nutritional factors, as well as potential gene–environment interactions.
Over the past two decades, reported autism prevalence has risen substantially. Much of this increase reflects more recent expanded diagnostic criteria, greater awareness and improved access to evaluation — particularly among adults. Yet heterogeneity and high rates of co-occurring conditions can complicate recognition, leading to delayed or missed diagnoses, especially in children whose communication styles or sensory sensitivities do not align with conventional expectations of autism.
In addition to the potential for hampering diagnoses in individuals, restricted or outmoded diagnostic criteria can adversely influence research on autism. Studies have often relied on narrow populations — for example, by excluding individuals who are nonverbal or from lower socioeconomic backgrounds. Inconsistent definitions and non-standardized measures further complicate efforts to track prevalence and outcomes. Addressing these shortcomings requires more-inclusive sampling and improved measurement tools, as well as moving away from more-medicalized, deficit-centric models toward difference-based conceptualizations of autism.
Central to the next evolution in the research landscape is tackling some of the underlying questions about what drives the heterogeneity observed in autism. In the March issue of Nature Mental Health, two Perspectives confront these questions from distinct angles, each challenging the adequacy of the singular term ‘autism’ and calling for conceptual change to better identify mechanisms, guide interventions and predict outcomes.
In a Perspective by Lombardo and colleagues, the authors propose an early-life, behavior-focused stratification framework called AUTISMS-3D (A3D), in which ‘D’ refers to disability as conceptually distinct from differences in development. Their approach distinguishes between two broad subtypes on the basis of non-core features such as language, intellectual, motor and adaptive functioning. Type I (‘disability’) and Type II (‘difference’) are proposed to show distinct structural and functional neuroimaging profiles that lead to divergent developmental trajectories and life outcomes. By identifying subtypes early, the A3D framework aims to accelerate the discovery of differential biological mechanisms, improve prognostic precision and tailor support more effectively. Importantly, the authors argue that clearer subtype distinctions may also help reconcile medical and social models of autism, supporting both scientific rigor and destigmatization.
In contrast, another Perspective, by Lin and coauthors, questions whether stable subtypes of autism can or should be identified at all. The authors argue that many attempts to define subgroups have failed because of ‘artifactual heterogeneity’ produced by broad diagnostic criteria that aggregate people with widely varying profiles. Rather than seeking subtypes, they propose variability itself as a defining feature of autism. Their person-centered framework focuses on idiosyncrasy or highly individualized neural and cognitive patterns that persist even after common sources of variance such as age, sex or comorbidities are accounted for. Detecting these patterns requires moving beyond group averages to examine outliers often lost in statistical noise. The authors describe a dynamic developmental landscape characterized by individualized neural ‘attractors’, suggesting that precision medicine approaches, combined with longitudinal tracking of people at a greater likelihood of developing autism, may better identify critical developmental windows and optimize supports.
Any framework for understanding autism should align with a neurodiversity perspective by recognizing autism as part of natural human variation rather than merely reducing it to a disorder or disability. Perhaps the most persistent barrier in this endeavor is the continued presence of stigma. Integrating biological insight with social understanding will be essential to developing policies and support that reflect the diversity of autistic people and a crucial point at which researchers, clinicians, educators and caregivers work collaboratively to enable individuals to live their best lives, wherever their place on the spectrum.