Naturalistic observations of elicited expressive communication of children with autism: an analysis of teacher instructions.

The author watched 12 minimally verbal kids with autism during preschool routines.

She coded every time a teacher used a model, a verbal prompt, or just waited.

She also coded each child’s request, comment, or protest that followed.

Modeling plus a verbal prompt gave the best lift—kids talked or gestured more right after.

Still, the total number of child messages stayed very low, even with these helps.

Most turns were requests; comments and protests were rare.

Rosenthal et al. (1980) already showed that prompts alone don’t create new mands—you need direct teaching.

Chiang (2009) now echoes that view: teacher modeling helps a bit, but frequency stays flat without formal training.

Carnett et al. (2016) took the next step and paired systematic instruction with an iPad voice output device.

Their kids learned to ask questions, kept the skill, and used it in new places—proof that structured teaching beats simple prompting.

If you run classroom routines, don’t expect modeling and prompts to do the heavy lifting. Use them as warm-ups, then add explicit mand training or SGD instruction. Target each request or comment under the real cue you want the child to use later. That combo—prompt now, teach next—turns rare, fragile bids into steady, spontaneous communication.