Choice between reliable and unreliable outcomes: mixed percentage-reinforcement in concurrent chains.

Green et al. (1987) let pigeons pick between two keys in a two-part chain.

One key always paid off. The other key paid off only half the time.

The team then changed how long each part of the chain lasted to see if delay altered the birds’ preference.

The birds strongly favored the sure-thing key.

When the wait times in the chain grew, the preference for the reliable key grew with them.

The results fit an extended delay-reduction model: shorter wait to sure food equals stronger choice.

Mazur et al. (1992) ran a near-copy of the task but mixed up the number of links instead of the payoff odds. They still saw a pull toward the risky side, yet adding a green light cue wiped the pull out. Together the studies show both probability and signals steer choice.

Martin et al. (1997) kept the 50 % vs 100 % comparison and simply inserted a five-second gap before the food signal. That small gap cut the birds’ risky preference in half, proving the signal itself acts as a conditioned reinforcer.

Anonymous (1995) also used the 50 % vs 100 % setup and found the risky key won only when the delays were short. Once delays lengthened or a gap was added, preference flipped to the sure key. The pattern lines up with Green et al. (1987) and underlines the same rule: delay controls the strength of preference.

For BCBAs the lesson is plain: clients will gravitate toward the option that cuts wait time and delivers sure reinforcement. If you want a student to pick the new task, keep the delay to reinforcement short and the payoff odds high. Check your schedules—adding extra steps or uncertain rewards can quietly weaken the power of your program.