Teaching generative use of sentence answers to three forms of questions.

Seven children with intellectual disabilities practiced answering three kinds of questions.

An adult first modeled a full-sentence answer. The child then copied it and got praise or candy.

Sessions kept going until each child could answer new, untrained questions the same way.

Every child learned to answer with complete sentences.

They also used full sentences on brand-new questions that had never been trained.

No extra teaching was needed for this transfer to happen.

Dodd et al. (1977) used the same model-plus-praise package on staff. Counselors watched a short demo and then praised clients more often. Both studies show the combo works for different age groups and behaviors.

Zhi et al. (2024) seems to disagree. They found that praise added nothing extra while teaching listener responses; correction alone did the job. The gap is about the behavior type. The 1975 paper teaches expressive answering, a skill that benefits from social reward. The 2024 paper teaches listener discriminations, where quick error correction matters more than praise.

Kahng et al. (1999) pushed fluency instead of rules. Adults learned English articles through rapid practice and generalized to new sentences. Like the 1975 children, they produced untrained verbal forms without grammar lectures.

You can add this two-step routine to any language goal. Model the full sentence once, then reinforce each correct copy. After a few short rounds, test with new questions. If the learner still answers in phrases, give another model and keep reinforcing. The study says you will likely see full sentences spread to questions you never practiced, saving you from teaching each one separately.