New Theory of Learning Upends the Lessons of Pavlov’s Dog

Explore

TThere are many different types of learning, but one of the simplest is associative learning. This is when we learn to associate one thing with another, like the ringing of a bell with food, in the famous example of Pavlov’s dog.

The more we experience the stimulus and reward, the faster we learn. At least that’s what we thought. New research published in Natural neuroscience throws a wrench into this model.

Neuroscientists Vijay Mohan K. Namboodiri and Dennis Burke of the University of California, San Francisco, trained mice to associate a sound with a sip of sugar water while varying the delays between trials: Some mice received a lesson every 30 to 60 seconds, while others waited 5 to 10 minutes. Even though the mice in the second group received fewer rewards than the mice in the first, they learned at the same rate.

“It turns out that the time between these signal-reward pairs helps the brain determine how much to learn from this experience,” Namboodiri explained in a statement.

Read more: “A glimpse into the minds of bees”

In other words, increasing the frequency of stimulus and reward combinations leads to diminishing returns over a short period of time, in the same way that cramming before an exam is not as effective as learning the material throughout the semester.

“What this tells us is that associative learning is less about ‘practice makes perfect’ and more about ‘timing is everything,'” Burke said.

Associative learning is mediated by dopamine, and during the learning process our brain begins to release the feel-good neurotransmitter in anticipation of reward. An examination of the mice’s brains revealed that those in the group that waited longer between lessons released dopamine after fewer trials than mice getting more frequent rewards. Basically, mice that underwent spaced trials got a better neurochemical bang for their buck.

The researchers then changed things by spacing the trials at 60-second intervals, but rewarding one group of mice only 10% of the time. The 10 percent also only needed fewer trials before releasing dopamine, whether there was a reward or not.

While these findings may change the way we view learning and addiction – another dopamine-mediated condition – Namboodiri plans to next apply his research to artificial intelligence.

“A model that borrows from what we discovered could potentially learn more quickly from fewer experiments,” Namboodiri said. “But right now, our brains can learn much faster than our machines, and this study helps explain why.”

Enjoy Nautilus? Subscribe for free to our newsletter.

Main image: ArtemisDiana / Shutterstock

• Jake Currie

  Published on February 20, 2026

  Jake Currie is a writer based in Brooklyn, New York.