New Study Explains Why No Two People See the Same Room the Same Way

Walk into a crowded airport, a public swimming pool, or a car repair shop, and the eyes immediately get to work: scanning, darting, landing on some things and skipping right past others. A common assumption is that whatever is brightest or most dramatic in a scene grabs the gaze. A new study says something more personal is going on. The way each person explores a new place is as distinctive and steady as a fingerprint, shaped by the meanings and concepts they carry around in their head.

Researchers at Dartmouth College and the University of California, San Diego, writing in the Proceedings of the National Academy of Sciences (PNAS), had 61 people put on virtual reality headsets and freely explore 100 immersive, 360-degree scenes, places like airports, auto shops, and public pools, while eye-tracking recorded every glance. Two people looking at the exact same scene often looked at very different parts of it, and those differences were not random. Each person’s pattern held steady, showing up again across completely different environments and across testing sessions days apart. Each viewer, it turned out, had a personal way of exploring the visual world, a stable “attentional fingerprint.”

More surprising was what drove those fingerprints. It came down not to how things look, but to what things mean.

How Meaning Shapes the Attentional Fingerprint

To find out what guided people’s eyes, the team built computer models that predicted where each person would look using three kinds of information: the basic layout of a scene, the visual appearance of its objects, and the abstract meaning of what was in it. For that last piece, they used a large language model trained on text. Models like this learn how words and concepts relate through enormous amounts of writing, so they grasp that a flag and a military jet are connected by the idea of patriotism, even though the two look nothing alike.

Here is where things got interesting. The language model, which gets no visual input and never touches a single pixel, was not the best at predicting gaze across the board; the vision model held that edge. Where the language model pulled ahead was in explaining the uniquely individual part of gaze, the part that set one person’s viewing apart from everyone else’s. Spatial layout mattered. Visual features mattered. But when it came to the deeply personal side of how someone looks at the world, meaning came out on top.

That result challenges a long-standing view of how attention works. For decades, researchers mostly explained where people look by what stands out visually: bright colors, sharp edges, movement, faces. This work doesn’t set those factors aside, but it shows they miss the personal layer of how a given person sees.

How the Study Worked

Each of the 61 participants wore a VR headset with built-in eye tracking that logged exactly where they looked, moment by moment, for 16 seconds per scene. Because the scenes were full 360-degree spaces, participants could turn their heads and bodies to look anywhere, giving each person plenty of room to show their own habits.

To build the meaning-based model, online raters wrote detailed descriptions of small patches within each scene, sorted into three levels of depth: a plain label (“a hat”), a description with context (“a hat that is on her head”), and a deeper read on purpose or function (“a hat that is on her head and could be keeping the sun from her eyes”). Feeding these richer descriptions into the language model made it noticeably better at predicting where individuals looked, which confirmed that deeper meaning, not just object names, was doing the work.

As a check, the researchers scrambled the descriptions so the words stayed just as rich but no longer matched the correct part of the scene. Accuracy fell apart, dropping even below the plain-label version. That ruled out the idea that the model was simply riding on longer or fancier text.

About half of the participants came back a week later to explore 40 brand-new scenes. Their individual patterns held. A model trained on a person’s first-session data could still identify that person’s gaze in the second session. Spatial layout alone could not manage that; the visual and, above all, the meaning-based models kept each fingerprint intact over time.

Timing told its own story. In the first couple of seconds, spatial habits led, with people glancing toward the center of a scene, a well-documented tendency. Visual features kicked in soon after, as viewers picked out specific objects. Meaning came on more slowly, building across the full 16 seconds and eventually overtaking both spatial and visual factors. That order makes sense, since it takes a beat to grasp what a scene is about before deeper associations can steer the next glance.

What an Attentional Fingerprint Could Reveal

Researchers say these results open up questions well beyond the lab. If gaze reflects stable priorities about meaning, future studies might use it to explore how different groups organize what matters to them, though the paper frames this as a direction to pursue rather than a proven tool. One example the authors point to is autism, in which differences in attention are well documented. Future work could test whether those differences trace back to visual-level or meaning-level factors. That question is wide open: no such test has been run, and the paper takes no side on the answer.

Privacy is worth a mention too. Gaze patterns here were distinctive enough to tell participants apart within this specific dataset, but the authors stress that this is not the same as eye movements working as a reliable ID the way a fingerprint or a face scan does. Showing a personal signature inside a controlled study is a long way from picking someone out of a crowd. That gap matters as VR and augmented-reality systems with eye tracking spread into daily life.

One broader idea, which researchers say has been studied little, is that each person moves through the world with a private map of meaning, a set of priorities about what things are and how they connect, quietly steering attention from the moment they step into a new place. Where the eyes land depends less on what stands out in a scene and more on what already matters to the person looking.

---