Training two severely retarded adolescents to ask questions.

Two teens with severe intellectual disability got adult-led sessions. The goal was to teach them to ask 'What is it?' when shown an unknown object.

First the adults tried modelling alone. They said the question while showing the item. That failed. Next they added a full verbal prompt plus candy or soda for each correct question.

Prompt plus reinforcement produced quick, stable question-asking. Modelling alone did almost nothing.

The skill appeared in the first session with the new package and stayed high.

Lahey (1971) seems to disagree. That study doubled preschoolers' adjective use with modelling only. The key difference: the 1971 kids were neurotypical Head Start pupils; these teens had severe ID. Modelling alone may be enough for typical kids, but learners with ID need the extra prompt and reward.

Kay et al. (2020) and Tavassoli et al. (2012) extend the idea. They show that the prompt you used last is the one that works fastest next. Rotate echoic and tact prompts to keep efficiency high.

Yuan et al. (2019) looks like a clash. They found reinforcement during error correction did not speed learning in preschoolers with autism. Note the task: error correction, not new question teaching. The teens here were learning a brand-new response, not fixing an old one.

Skip modelling-only trials for learners with severe ID. Start with an immediate full verbal prompt and deliver a small edible or social reinforcer right after each correct question. Once the response is steady, fade the prompt word by word. Track which prompt form you used last; it will likely work fastest next session.