Per-trial learning scales with time between rewards in mice
Feb 22nd 2026
Across multiple conditioning protocols, mice and their mesolimbic dopamine signals learned more per trial when rewards were spaced further apart, making total conditioning time nearly constant despite far fewer trials.
- Per-trial behavioral learning increased proportionally with the inter-reward interval (IRI) across 30 to 600 second spacings.
- Cue-evoked dopamine responses in nucleus accumbens developed per trial in the same proportional way and preceded the behavioral change.
- Because learning per trial rose with IRI, total time to reach conditioned behavior was effectively constant across groups despite a 10 to 20 fold difference in trial counts.
- At an extreme IRI of 3600 seconds the 1/IRI proportional rule broke down: animals still learned faster but not at the predicted rate and cue-evoked dopamine reached higher asymptotes.
- A retrospective dopamine-based model called ANCCR accounts for the proportional scaling by updating cue given reward at each outcome and normalizing over the same temporal window, which predicts learning rate proportional to IRI.
- The same proportional scaling of per-trial learning was observed in aversive cue shock conditioning, showing the rule applies to both appetitive and aversive outcomes.