Brief report: generalisation of word-picture relations in children with autism and typically developing children.

Hartley et al. (2014) taught eight preschoolers to match words to pictures. Four kids had autism and no speech; four were typically developing.

Training used a tabletop game. Kids learned that one spoken word went with one picture of a toy. Next they saw new toys that matched the trained toy only in color or only in shape. The team asked, "Which of these also gets the same name?"

Typical kids picked the shape-match almost every time. They acted like shape is what counts for a name.

Autistic kids picked shape or color about equally. Color pulled their choices even though it was irrelevant. They treated any shared feature as a good reason to extend the word.

Kunda et al. (2011) review shows many autistic learners lean on visual cues. Calum’s color bias is a live example of that visual-first style.

Polo-López et al. (2014) ran a similar matching game the same year. They also saw shaky generalization after training. Together the two papers warn that simply teaching a match is not enough; you have to program the right cue.

Floyd et al. (2021) found verbal autistic students struggle with words that carry more than one meaning. Calum’s toddlers show the front end of the same problem: they grab any surface feature and call it the name. Both studies say, "Don’t assume typical word-learning rules."

When you teach a new word to a low-verbal child with autism, explicitly point to the shape and say, "This shape means the name." Block the color cue by showing the same toy in many colors. Keep trials until the child picks shape over color on three different sets. This one tweak can stop overgeneralization before it starts.