On the role of covarying functions in stimulus class formation and transfer of function.

The team paired shapes with mild shock for adults without disabilities. Later they taught the same adults to pick one shape for money and avoid another. The question: would the shock pairings alone build a stimulus class that later controls choice?

A second group got the same shapes but no shock. Everyone then did the money game with new shapes that had shared the earlier pairings.

Only the shock group showed transfer. They quickly treated the new shapes like the old ones—approaching or avoiding without extra training. The no-shock group acted as if the shapes were unrelated.

This tells us respondent conditioning (simple pairings) can create a class that later guides operant behavior.

Pérez-González et al. (2003) extends this idea. They showed that contextual cues alone can transfer control to brand-new conditional tasks. Together the two studies say both painful pairings and neutral context cues can build usable stimulus classes.

Earlier work used reinforcers instead of shock. Dube et al. (1989) and Dube et al. (1987) formed classes by linking shapes to food. The 2002 paper swaps food for shock but lands on the same outcome—indirect relations create classes that guide later behavior.

Roche et al. (1997) looks like a contradiction at first. They found early pairings can block later conditional learning. The difference is timing: B et al. tested whether pairings interfere with new conditional tasks, while G et al. tested whether pairings help transfer after the class is already built.

You now have a low-effort way to build stimulus classes. A few contingent pairings—tone with token, picture with snack—can set up a class that later speeds teaching of conditional discriminations or safety rules. Try inserting a brief pairing phase before your standard matching-to-sample program and watch for faster emergence of untaught relations.