Choice and transformed interreinforcement intervals.

Pigeons pecked two keys in a lab chamber. Each key led to a different food schedule. The schedules looked the same on paper. The twist was how each schedule ended. Sometimes it stopped after the first food pellet. Other times it stopped after a set time. Parrott (1984) wanted to see if this small rule change would flip the birds' favorite key.

The birds' favorite key reversed when the ending rule changed. Same food rates. Same delays. Only the exit rule differed. The result shows that local rules beat overall statistics. Birds did not just count food per minute. They reacted to how each bout ended.

Alba et al. (1972) first showed that chaining cuts preference. Parrott (1984) goes further: even after chaining, you can flip choice by tweaking the exit rule. Duncan et al. (1972) also found that extra links hurt preference. Parrott (1984) agrees, but shows the hurt can be undone with a simple rule change.

Landon et al. (2002) later split short-term from long-term choice effects. Their data fit J's story: moment rules matter as much as long-run rates. Green et al. (1987) added probabilistic reinforcers and still saw orderly preference. Again, the local structure, not just the average payoff, drove the birds.

When you write a token board or set a timer, think exit rule, not just rate. A child may prefer a 5-min art session that ends after one sticker over a 5-min session that ends at the bell, even if stickers arrive at the same pace. Test both setups. Let the learner vote with their feet.