Scientists at the University of California, San Francisco (UCSF) have challenged a long-standing belief about how associative learning works. Traditionally, it was thought that repetition—the number of times an event is paired with a reward—was the main driver of learning. However, new research suggests that the timing between rewards plays a more significant role.
Vijay Mohan K. Namboodiri, PhD, associate professor of Neurology and senior author of the study published in Nature Neuroscience on February 12, explained: "It turns out that the time between these cue-reward pairings helps the brain determine how much to learn from that experience." He added that when experiences occur closer together, "the brain learns less from each instance," which may help explain why students who cram for exams often do not perform as well as those who study over longer periods.
The research team, including postdoctoral scholar Dennis Burke, PhD, conducted experiments with mice to test their theory. The mice were trained to associate a brief sound with receiving sugar-sweetened water. Some trials were spaced 30 to 60 seconds apart while others had intervals of five to ten minutes or more. Despite receiving many more rewards in closely spaced trials within the same timeframe, mice did not learn faster than those with fewer but more widely spaced rewards.
Further investigation showed that when rewards were spaced further apart, mice required fewer repetitions before their brains began releasing dopamine—a chemical linked to learning and reward—in response to the sound cue alone.
In another experiment, researchers played the sound every 60 seconds but only gave sugar water 10% of the time. Even so, these mice needed far fewer actual rewards before their brains started responding with dopamine after hearing the cue.
These findings could have implications for understanding both learning and addiction. For example, cues associated with smoking can trigger cravings even if nicotine is not always present. Continuous delivery methods like nicotine patches might disrupt this association by providing steady rather than intermittent exposure.
Looking ahead, Namboodiri plans to explore whether this revised model of learning could improve artificial intelligence systems. Current AI models typically require many repeated interactions to refine predictions—a process similar to traditional views on human learning.
"A model that borrows from what we've discovered could potentially learn more quickly from fewer experiences," Namboodiri said. "For the moment, though, our brains can learn a lot faster than our machines and this study helps explain why."
