The effects of single instance, multiple instance, and general case training on generalized vending machine use by moderately and severely handicapped students.

The team worked with high-school students who had moderate or severe intellectual disability. Each teen needed to learn to buy snacks from vending machines they had never seen.

The researchers compared three teaching sets. Single-instance showed one machine. Multiple-instance showed three machines that looked alike. General-case also used three machines, but the machines differed in buttons, coin slots, and item slots. The goal was to see who could use any new machine later.

Students who got the general-case set could buy from brand-new machines. Students who saw only look-alike machines often got stuck when a button or slot was different.

General-case training won. Three varied examples beat three similar examples or one lonely example.

Milata et al. (2020) repeated the idea with teens with autism. They added video clips and prompts and still saw wide transfer to new debit-card readers. The 1984 vending study set the stage; the 2020 study shows the same logic still works decades later.

Collier et al. (1986) built on the idea by adding "don't do this" examples. Their students with ID learned to clear cafeteria tables in untrained spots. The 1984 paper proved varied exemplars matter; the 1986 paper says you can sharpen the lesson with clear non-examples.

Phillips et al. (2019) looks like a mismatch at first. They taught only one five-step sequence and saw weak transfer. Their result quietly agrees with Sprague et al. (1984): one narrow lesson is not enough. You need multiple, varied exemplars for true generalization.

If you want a learner to use any microwave, any bus reader, or any vending machine, pick three machines that span the differences. Train with that varied set and probe on brand-new ones. One example is a gamble; three strategic examples give you generalized use.