Reconceptualized Associative Learning

Gallistel (2025) rewrote the rules of learning. Instead of tracking how close two events are in time, the paper says animals and people track the informativeness ratio. That ratio compares how often an event happens with a signal versus without it.

The paper pulls data from both Pavlovian and operant studies. It shows the same math fits both types of learning. The key is not time, but how much the signal changes the odds of the event.

Learning works like a statistician, not a stopwatch. The brain asks: does this bell make food more likely? If yes, strong learning. If the bell gives no new info, learning stays weak.

This single rule explains why long delays sometimes work and why short delays sometimes fail. The delay only matters if it changes the informativeness ratio.

Calamari et al. (1987) first said rules are not simple cues but messages about contingencies. Gallistel (2025) extends that idea to all of associative learning. Both swap old labels for new, math-based ones.

Dews (1978) showed that Pavlovian relations can drive operant contrast. Gallistel’s model unites those findings under one ratio. The 1978 data now fit neatly inside the new theory.

White (1995) found stimulus control fades with time since the last schedule change. That decay looks like a drop in informativeness, not mere forgetting. The new view keeps the data but gives a cleaner reason.

Sherwell et al. (2014) added brief cues to sharpen time-based discrimination. Gallistel predicts those cues work only when they raise the informativeness ratio. Same method, deeper explanation.

Stop asking “how close in time?” Ask “how much does this cue change the odds?” When you design a teaching trial, check if the SD truly raises the chance of reinforcement compared to baseline. If it does not, the learner may not notice the link, no matter how fast you deliver the reinforcer.