2025 commentary · Neuron

learning on a break

Xiao and colleagues found that brief bursts of hippocampal activity during a short break were associated with how people later combined what they’d learned. in our commentary, we discussed which experiences the brain might replay during rest, and how that could make later decisions easier.

say you learn that restaurant b has better food than a, and that c has better food than b. you can work out that c beats a by following the chain (this is called ‘inference’: working out a relationship you never learned directly). but could your brain link a and c ahead of time, so the answer is ready when you need it?

Xiao and colleagues recorded from electrodes inside participants’ brains while they learned relationships like these, took a short break, and then made new comparisons. brief bursts of activity in the hippocampus, a memory hub, during the break were associated with how participants organized and used what they’d learned afterward. in our commentary, we discussed what kinds of experiences the brain might replay during a break like this, and how that could help with future decisions.

imagine learning about four restaurants one comparison at a time. you liked the food at b better than a, and then decided that c is better than b. to compare a and c, you’d chain those together: a → b → c.

you can do the same when comparing the service at these restaurants. now you have two separate rankings of the same four places, shown as the two rows on the right. participants in the study learned relationships like these between unfamiliar items, along two different dimensions.

after learning these relationships, participants in the study rested for five minutes without learning anything new. from the outside nothing is happening, but the brain can be going back over what it learned and linking it together. this is called ‘replay’: the brain reactivating recent experiences, while you rest.

on the right, click the button and watch the replay run a → b → c. going over that sequence can also strengthen a direct a–c link, so next time you won’t need to go through b. thicker lines are links that are easier to ‘bring to mind’ and make choices between

the animation pauses when the a–c link forms.

later, someone asks you: a or c? there are two ways to answer. if you already have a direct a–c link, you just use it. if you don’t, you can still get there by walking through what you do remember, one link at a time: c beats b, and b beats a, so c beats a.

it’s like getting to a friend’s apartment. if you’ve been there before, you just go. if you haven’t, you take the way you know: drive to the grocery store you always go to, then walk over from there. it takes longer, but it gets you there. replay during a break could build the direct route before you need it through ‘simulation’ (i.e., replay), and Xiao and colleagues found that participants got better at these new comparisons after resting.

switch between “recall the learned links” and “use the map” to see both routes.

Xiao and colleagues could measure what happened during the break. their participants already had electrodes in their brains for medical reasons, which let the team record brief bursts of fast electrical activity in the hippocampus called ‘ripples’. ripples are thought to be one of the moments when replay happens.

activity around these ripples, along with coordination between the hippocampus and medial prefrontal cortex, a region that helps both organize knowledge into maps and track how much options are worth, was associated with the map participants formed and the comparisons they made afterward. on the right is our example map after a full break of replay: some links are now ready to use, while others would still need the long way around.

this is where our commentary comes in. the brain has far more experiences to keep track of than it can rehearse, so it has to prioritize. which experiences it replays would decide which comparisons are ready when the next choice arrives.

we proposed a few possibilities: replaying the most rewarding options more often, replaying the links that connect the two kinds of knowledge (here, food and service), or simply replaying at random. the four maps on the right show the same break under each strategy, plus one more that favors the weakest links. each strategy prepares a different set of comparisons.

each map is the average of many breaks with that strategy, so the differences come from the strategy itself rather than chance.

in our commentary, we discussed what the brain might choose to replay, and what that would mean for later decisions. our final figure compares a few possibilities: replaying the most rewarding options more often, replaying links that connect the two kinds of knowledge (here, food and service), or replaying at random. these are our proposals for how rest could prepare you for future choices; the original study did not compare them.

the demo lets you try those three, plus a fourth: replaying whichever links are currently weakest, as if the brain prioritized what it remembers least. every version starts with the same a → b → c example. each replay strengthens the two links it visits plus the direct link between the ends, and links that don’t get replayed fade a little. different replay strategies end up preparing different comparisons.

see the original figure

main result figure for neuron commentary showing the effect of rest on cognitive maps — figure from our commentary. view full size →

read the full commentary

from separate experiences to a shared map

each row ranks the same four places by a different quality.

thicker lines are links that are easier to bring to mind; shrinking halos show a place’s position becoming clearer.
based on the discussion in Orloff & Boorman, 2025.

two rankings learned separately

replay during a break

a comparison that’s ready to use

listening in on the break

our idea: choosing what to replay

from separate experiences to a shared map