Empirically derived subclassification of the autistic syndrome.

Siegel et al. (1986) built a new way to sort kids with autism. They used only what you can see: actions, not blood tests or brain scans.

The team wrote clear rules for each subtype. Then they checked if the groups looked different in real clinics.

The tool split the children into smaller, more alike groups. Each group had its own pattern of skills and needs.

Early markers, like how they used toys or looked at faces, lined up with the new subtypes.

McQuaid et al. (2024) did the same idea with adults. Their two subtypes stayed stable for years and predicted later mood and IQ scores.

Beglinger et al. (2005) used a social subtype plan before EIBI started. Kids tagged as aloof gained fewer IQ points than active-but-odd or passive kids.

Courchesne et al. (2019) faced the same heterogeneity headache. They fixed IQ testing for minimally verbal preschoolers by leaning into strengths instead of one-size-fits-all tasks.

You can borrow this logic today. Pick a few top-priority traits—social approach, toy play, or response to name—and write plain definitions. Sort your new learners into these homemade subgroups, then match teaching speed, prompt level, and reinforcement type to each mini-profile instead of using the same plan for every child with an autism label.