An analysis of response and time matching to reinforcement in concurrent ratio-interval schedules.

Researchers placed pigeons in a chamber with two keys. Each key paid off on a different schedule. One key used a ratio schedule. The other used an interval schedule.

They recorded two things: how many pecks the birds made on each key, and how much time they spent near each key. They wanted to see which measure better matched the actual payoff rates.

The birds' peck counts lined up with the payoff rates. But time spent near each key did not. When the two measures disagreed, response counts were the reliable guide.

This means you should trust response data over time data when schedules clash.

Goldman et al. (1979) later tested adult humans with the same setup. They added a twist: a light signaled when payoff was ready. The signal pushed choice away from the lit key. This shows the matching law holds across species, but signals can bend it.

Blue et al. (1971) ran a similar pigeon study two years earlier. They found that slow, 0.5-second steps in changeover delay kept payoff rates closer to what was programmed. Garcia et al. (1973) used a fixed delay, so their payoff rates stayed clean.

Fahmie et al. (2013) came 40 years later. They showed that a history of longer changeover delays can make a key more preferred even when payoff rates are equal. This builds on the 1973 finding by adding learning history as another layer.

When you run concurrent schedules in practice, track responses, not just time. If a child splits time between two tasks but only earns tokens on one, the token count—not the clock—tells you which schedule is actually working.