The effects of terminal-link fixed-interval and variable-interval schedules on responding under concurrent chained schedules.

Byrd (1972) let pigeons pick between two keys. Each key started its own chain of schedules.

The first link was always a short variable-interval. The last link was either fixed-interval or variable-interval.

The team changed the absolute length of the terminal FIs while keeping the ratio between them the same.

Birds only matched relative time to relative reinforcement when the FIs were 5 s versus 10 s.

At 20 s versus 40 s the same 2:1 ratio fell apart. Choice flattened even though the rate ratio stayed 2:1.

When both options ended with VI schedules that shared the same harmonic mean, choice looked the same no matter which VI was on which side.

FARMEMOORHEARSKELLEHER et al. (1964) saw response rate rise as the terminal reinforcer got closer. Byrd (1972) adds that the absolute wait time, not just the relative wait, decides which chain is chosen.

Steege et al. (1989) later swapped pigeons for humans and found maximization instead of matching. The bird data from Byrd (1972) help explain why: once terminal delays grow past a few seconds, pigeons also stop matching.

Wilkinson et al. (1998) ran concurrent VI-EXT and saw no exclusive choice, seeming to break the matching law. Byrd (1972) shows the law holds only in a narrow FI window, reconciling the two studies.

When you build token boards or chained schedules, keep the last wait short. A 5-s delay keeps choice aligned with reinforcement rate; a 20-s delay does not. If you must use longer delays, expect clients to spread time evenly no matter how tokens accrue. Test both options in your next session and watch the split change.