An investigation of extinction-induced vocalizations.

Three preschoolers with autism used sign language to ask for toys. None of them spoke.

The team first gave toys only when the child signed. Then they stopped. They wanted to see if the kids would try talking instead.

They also tested giving extra praise for any sound, but only after they saw what plain extinction did.

When signing no longer worked, all three kids made new sounds. Words, word tries, and babble jumped up right away.

Just giving praise for sounds, without stopping the signs, did nothing. The jump happened only after the signs stopped working.

Cengher et al. (2020) later showed the same trick with teens. After they stopped one-word signs, the kids started full sentences with words like “I want the red car, please.”

Takashima et al. (1994) saw the same pattern in play. When a trained toy move no longer earned candy, kids invented new ways to play with the same toy.

McReynolds (1969) did the opposite move: brief timeout from all reinforcement made one preschooler stop noise and start clear words. Together these studies show withholding reinforcement can both cut bad sounds and spark good ones, depending on what you set up next.

If a client uses signs or devices but never tries speech, you can run a quick probe. Hold the signed mand for a minute and wait. New vocal tries may pop out on the first try. Capture them with praise or the item, then build from there. It is a low-cost way to check if the voice box is ready to join the conversation.