Summary: Subtle differences in hand and finger movements during everyday grasping tasks can reliably distinguish autistic from non-autistic individuals. Using machine learning to analyze natural grasping kinematics, researchers achieved roughly 85% accuracy in classifying autism spectrum disorder (ASD) from these motion patterns.
These motor signatures often appear early in development and could complement current diagnostic approaches that rely on behavioral markers emerging later in childhood. The study suggests a pathway toward simpler, scalable screening tools based on natural hand movements, which could enable earlier diagnosis and faster access to interventions and support for autistic individuals.
Key Facts:
- ~85% Classification Accuracy: Machine learning models distinguished autistic from non-autistic participants using grasping motion data with high precision.
- Early Motor Signals: Differences in fine motor control during grasping may be detectable earlier than many behavioral diagnostic signs.
- Scalable Assessment Potential: Analyzing natural hand movements could become a low-cost, accessible approach to support screening and diagnosis of ASD.
Source: York University
Timely diagnosis of autism spectrum disorder remains a global challenge. New research from York University finds that the way people grasp objects—specifically the kinematics of thumb and index-finger movements—may provide a clear and measurable indicator of ASD.
An international research team applied machine learning to precise measures of naturalistic hand movements. Participants performed a simple precision grasp: they used their thumb and index finger, fitted with tracking markers, to pick up blocks of varying sizes, lift them, replace them in the same location, and return their hand to the starting position. The recorded finger trajectories formed the basis for kinematic analysis and classification.
Lead author Associate Professor Erez Freud (York University, Department of Psychology and the Centre for Vision Research) reports that the classifier consistently identified ASD with approximately 85% accuracy. The models captured subtle differences in timing, trajectory, and coordination of digit movements, revealing distinct kinematic signatures associated with autism.
Autism affects approximately one in 50 Canadian children, and delayed diagnosis can limit access to early support services. The research team emphasizes that motor markers—such as the fine-grained features of grasping—often emerge in infancy or early childhood, meaning they could reduce the age at which reliable screening is possible.
Professor Batsheva Hadad of the University of Haifa, a senior collaborator on the study, explains that traditional diagnostic markers often focus on social and communicative behaviors that develop later. By contrast, motor markers provide an earlier window into neurodevelopmental differences and may help clinicians identify children in need of assessment sooner.
Participants in the study were young adults, matched on age and IQ between autistic and non-autistic groups. The use of adults—rather than children—allowed the researchers to control for potential developmental delays and focus on persistent, trait-like differences in motor control. Across subject-wise cross-validation, classifiers exceeded 84% accuracy, while trial-wise analyses and receiver operating characteristic (ROC) metrics showed strong performance (area under the curve values above 0.95 for subject-wise and above 0.85 for trial-wise analyses).
The study highlights that naturalistic precision-grasp tasks, combined with modern machine learning techniques, offer a powerful method for extracting diagnostic information from movement data. Unlike complex or clinical motion-capture setups, similar analyses could potentially be adapted for accessible devices or apps that record simple hand tasks, making early screening more scalable and affordable.
While the findings are promising, the authors note that further validation across larger and more diverse populations—especially in younger children—is needed before such methods can be integrated into clinical practice. Still, the research opens a promising avenue: objective, movement-based biomarkers that could augment behavioral assessments, speed up diagnosis, and guide earlier interventions to improve long-term outcomes for autistic individuals.
About this Autism research news
Author: Sandra McLean
Source: York University
Contact: Sandra McLean – York University
Image: Image credit to Neuroscience News
Original Research: Open access.
“Effective autism classification through grasping kinematics” by Erez Freud et al., published in Autism Research. DOI: http://dx.doi.org/10.1002/AUR.70049
Abstract
Effective autism classification through grasping kinematics
Autism is a complex neurodevelopmental condition characterized by differences in social communication and behavior, alongside frequent motor abnormalities. Motor features commonly appear in early childhood, making them important targets for earlier diagnosis and intervention.
This study evaluated whether kinematic features extracted from a naturalistic precision-grasp task could reliably distinguish autistic from non-autistic participants. We recorded movements from two markers placed on the thumb and index finger and used subject-wise cross-validated classifiers to evaluate diagnostic performance. The models achieved accuracy scores above 84%.
Receiver operating characteristic analyses indicated strong classification power, with area under the curve values above 0.95 in subject-wise analyses and above 0.85 in trial-wise analyses. These results demonstrate that subtle differences in motor control during a simple grasping task can be captured effectively and used to discriminate between autistic and non-autistic individuals.
The findings support further development of accessible, movement-based diagnostic tools that could augment existing assessment methods and help deliver earlier, more reliable identification and support for people on the autism spectrum.