Operants.

Siegel (1971) wrote a think-piece, not an experiment. The paper asks: what counts as an operant?

Old rules said an operant must produce a clear consequence each time. The author says that rule breaks down during extinction, superstition, and transfer tests.

The paper finds the classic definition too tight. It offers a wider view so the same unit can cover responses that no longer pay off, responses that pay off randomly, and responses that move to new places.

Rosenthal et al. (1980) extends the idea. They apply the broader operant view to naming and show that a single naming unit can join speaker and listener roles.

Koegel et al. (1992) uses the same lens on extinction. They add an evolutionary twist: variation and selection explain why bursts happen when payoff stops.

Lewon et al. (2026) moves the ball again. They fold Pavlovian cues into the widened operant frame so you can plan for both respondent and operant sides of a procedure.

If you treat only responses with clear payoffs as operants, you will miss key behavior in extinction, superstition, and generalization probes. Use the wider unit to see bursts, accidental reinforcement, and emergent naming as part of the same class. Your data sheets, task analyses, and parent explanations all get simpler because one story covers more cases.