Choice: Effects of changeover schedules on concurrent performance.

Reberg et al. (1979) watched pigeons choose between two keys. Each key paid off on its own variable-interval schedule.

A changeover delay punished quick switches. The team tracked how often birds hopped between keys.

They wanted to know if switching itself followed the same math as staying.

Birds switched keys in the same ratio as the food they earned. If left key paid 70%, about 70% of all hops landed there.

The animals sampled each side just enough to keep the payoff odds straight.

Neuringer et al. (1968) already showed that longer delays cut switching. Reberg et al. (1979) added the idea that each hop is a data point, not an error.

Krägeloh et al. (2003) later doubled sensitivity by telling the birds which side was richer and by keeping the delay. Their signaled ratios sharpened the same sampling rule.

Fahmie et al. (2013) looked backward, not forward. They found that birds still liked a key that once had long delays, even when payoffs later matched. This seems to clash with pure moment-by-moment sampling, but the tests were probe-only with no delay present, so history could speak without current punishment.

When you run concurrent schedules in the lab, treat every changeover as information, not noise. Keep the delay short enough to let sampling occur, but long enough to block superstitious flips. If you shift reinforcer ratios, give at least six sessions so the new pattern can overwrite old hop habits.