bioRxiv · 2026

preferences shape how the brain maps options

we found evidence that brain regions known for navigation organize options by what they’re worth to each person. a map like this could help people evaluate options they’ve never encountered.

preprint · not yet peer reviewed

how do you choose something you’ve never tried? say it’s your usual, dependable coffee shop versus a new café that might be amazing but could also be a letdown. even before you taste anything, its menu and reviews give you clues about how good the coffee could be and how likely it is to deliver, compared with the coffee shops you already know.

those two clues, how good something could be and how likely you are to get it, are the basic ingredients of most uncertain choices. in the lab, we turned them into amounts of money and chances of winning, which we can control precisely. while 35 adults chose between these gambles, we used fMRI (which tracks blood flow as a stand-in for brain activity) to ask whether the brain lays them out as a ‘cognitive map’, an internal map of how options relate, like its maps of places. we found map-like patterns in entorhinal cortex, a navigation hub, and in medial prefrontal cortex, which helps both represent maps and track value. these patterns followed each participant’s own preferences (their ‘subjective value’), not just the objective numbers.

which gamble would you rather play? make these hypothetical choices—then i’ll explain what your brain is likely doing to help make these choices.

participants in our study made tradeoffs like these. we’ll come back to your answers in a moment.

let’s put one of those offers on a map. option a offered $8 with a 30% chance of winning. go across to its chance, then up to its amount. that’s where it sits.

on the map to the right, each dot is a possible offer. farther right is a better chance and farther up is more money, so the top-right corner is a big reward you can count on (i.e., $10 with 100% chance). each offer gets a spot on the map, even one you’ve never seen.

for now, the map is spaced by the plain numbers, the same for everyone.

option b offered $3 with a 80% chance. same rule, different spot. one option has more money at stake, and the other gives you a better shot at winning something.

the line between them is the tradeoff. a map keeps track of how all the options relate, so you can compare any two without memorizing an answer for every pair.

the two colored dots are the offers on the cards.

here’s an example pair. chance sets how far right each offer sits, and amount sets how high.

the map isn’t fitted to you yet. it’s the same layout for everyone, with your comparisons drawn on it.

each line connects one pair you saw, and a ring marks the one you picked.

here are two offers you haven’t seen. we can place them right away because we know their chances and amounts. you don’t need to have played either one to compare them.

same with a new café. its menu and reviews hint at how good the coffee could be and how likely it is to deliver. that’s enough to place it among the cafés you already know. this is what makes a cognitive map useful: new things fit into an existing structure that represents café features.

in our study, we tested whether brain regions known for navigation help organize these options.

but the same offer can mean different things to different people. say you just need a decent cup of coffee. the familiar Starbucks is the safe bet, and a bad cup from somewhere new is not worth the risk.

with money, that’s like preferring a good chance at a smaller reward. this map is stretched to show those preferences, which we call a ‘subjective value’ map: higher means the reward feels more valuable, and farther right means the chance feels better. now b’s better odds count for more than a’s bigger prize, even though the offers themselves haven’t changed. to describe preferences like these, we use a computational model: a few equations, fit to someone’s choices, that capture how much they weigh chances against amounts.

with these example preferences, the model picks b. this isn’t your map yet.

now say you’re feeling adventurous. you’re curious about the new café, and a shot at amazing coffee is worth the risk of a bad cup.

watch the two offers. a keeps most of its height, while b slides down and to the left. the bigger prize now matters more than the better odds, so the model picks a. the riskier option got more appealing without its odds changing at all.

these are two example ways to weigh the same offers. next, a map based on your answers.

this is an example map. the interactive activity can sketch one from your answers.

five choices can’t pin down exactly how someone weighs odds and rewards, so a map like this is only a rough sketch. in the lab, we ask participants many more questions. (it’s also not a personality test, or a prediction of your coffee order.)

based on this map, the model predicts which of the two new offers would be preferred.

so does the brain use a map like this? while 35 participants compared offers in the scanner, we looked for a ‘grid-like code’, a signal that comes from cells the brain uses to create ‘cognitive maps’ of different spaces (the details tab explains how it works).

we found this signal in the two brain regions we expected: entorhinal cortex and medial prefrontal cortex, where grid-like signals have been found before. importantly, and what we found most interesting, the signal matched each participant’s own subjective value map, not the objective amounts and chances. what’s more, the brain showed this signal even when participants were making comparisons they had never made before. that last part is why i think the brain organizes options as a map: if it only stored a value for each pair it had already seen, a new comparison would have nothing to draw on, but a map lets you place a new option relative to the ones you already know.

next: what “grid-like” means.

we found two kinds of evidence for a subjective value map. the first is about direction (the ‘grid-like code’). grid cells, neurons first found in the entorhinal cortex of navigating animals (a discovery that won a Nobel Prize), fire at locations arranged in a hexagonal grid, like a honeycomb. (this is not an accident! hexagons are the most efficient way to tile a flat surface, and they show up all over nature: in honeycombs, snowflakes, a fly’s eye, and even the storms at the poles of Saturn and Jupiter. but i digress…) grid-like signals have since been found in medial prefrontal cortex too. with fMRI, we looked for a signal that varied with this hexagonal pattern as participants compared gambles, treating each comparison as a step in some direction across the map.

the question was which map the signal followed: the objective one, or each participant’s subjective one. in entorhinal and medial prefrontal cortex, it followed the subjective map, including for pairs participants hadn’t practiced in training.

the second is about location (the ‘2D positional code’). if two options sit close together on someone’s subjective map, does the brain respond to them more similarly? we found that it did. this held even across two different sets of shapes we used to stand for the options, so the brain was tracking where an option sat on the map, not what it looked like.

about the demo: in the real task, participants learned amounts and chances from abstract shapes, then compared combinations in the scanner. here we just show the numbers. when the demo says “the model picks” an option, that’s our computational model’s prediction: it weighs each option’s chance and amount by the estimated preferences, combines them into one value, and favors the higher one. it doesn’t simply pick whichever option is closer to the top-right corner.

see the original figure

figure 2 shows objective and subjective value spaces, decision vectors, and predicted sixfold grid-like modulation — figure 2 · objective and subjective decision vectors, and the grid-like pattern tested in the preprint. view full size →

read the full preprint

choice 1 of 5

the brain evidence

a navigation system, applied to value

green marks where we found the grid-like signal following each person’s own map.

complete figure 3 with brain maps, individual estimates, symmetry controls, and behavioral correlations — fMRI · 35 adults choosing between gambles

best known for keeping track of where things are as we navigate.
it helps represent maps of how things relate and track how much options are worth.

which would you rather play?

putting an option on the map

a choice becomes a line on the map

your choices on the map

room for a new option

preferences reshape the map

the same offers on an adventurous day

a rough sketch of your map

the personal map showed up in the brain

entorhinal cortex

medial prefrontal cortex