Unsupervised Categorization in a sample of children with autism spectrum disorders.

Rojahn et al. (2012) watched kids sort pictures without help.

Some kids had autism. Some did not.

The team made the sorting game easier or harder to see when kids picked only one feature.

On easy sets the autism group used one feature more than peers.

When the sets got hard, both groups slid into the same simple plan.

Task load, not diagnosis, drove the drop.

Preston (1994) once taught four kids with autism a sorting trick. All reached perfect multi-feature sorts and kept the skill. That study says one-feature sorting can be fixed with training.

Cicchetti et al. (2014) later showed that listener training—just asking kids to point to “the animal”—sparked full categorization and naming for most. Together the three papers form a line: bias seen, bias trained, bias fixed.

McGonigle-Chalmers et al. (2010) looked almost the same but found no group gap after matching IQ. The clash fades when you see Margaret kept age and IQ tight while J let those vary, letting task difficulty shine through.

Do not label a child as “over-selective” forever. First check how hard the task is. Ease the load or add quick listener drills like V et al. and you may see flexible sorting right away.