Made by AI, Directed by David Stein

Search
XR / Spatial Computing

Adding Smell to VR Made People Navigate Worse Before It Made Them Better

Two studies with fan-based scent delivery found that conflicting smell and vision produced systematic directional bias — and distance-linked scent improved accuracy at the cost of time.

4 min read

Smell is the sense that immersive technology keeps promising and never ships. It has been the punchline since Smell-O-Vision in 1960, and the reason is partly engineering — scents are slow to clear, hard to mix, and need consumables — and partly that nobody has demonstrated it does anything a display could not.

“Smelling the Way: Olfactory Modulation of Spatial Estimation and Path Integration in Virtual Reality”, posted 26 September 2026 by Siyeon Bak, Junho Kim, Dongyun Han, Sun-Jeong Kim and Isaac Cho, is an attempt at the second question. It asks whether scent can carry spatial information — not atmosphere, but where things are.

The two experiments

Study one, 24 participants: find the source. Participants traversed a linear route and tried to locate a scent source, with fan-based olfactory devices delivering the smell.

Study two, 19 participants: triangle completion. Participants walked two legs of a triangle and then tried to return to the start — the standard path integration task, which measures whether you can keep track of where you are by dead reckoning rather than by landmarks. Varied by rotation angle and scent condition.

The results, which are better than a clean win

When vision and olfaction conflicted, participants showed systematic directional biases without improved accuracy.

Read that carefully, because it is the sharper finding. A conflicting scent cue did not produce noise — it produced a consistent bias. People did not become randomly worse; they became reliably wrong in a particular direction. That means the olfactory signal is genuinely being incorporated into the spatial estimate rather than ignored, and when it disagrees with vision the brain does not simply discard it.

For anyone building multisensory work, that is a warning with teeth. An olfactory cue that is decorative — a forest smell in a forest scene, not tied to any particular location — is not neutral. If the scent’s implied source disagrees with what the visuals say, you are introducing a directional error into your user’s spatial model, and they will not notice it happening.

Distance-modulated scent feedback decreased navigation error but increased travel distance and time.

When the scent got stronger as you neared the target, people got closer to the right answer — and took a more cautious route to get there, wandering further and pausing longer before committing. The authors read this as a verification strategy: the scent gave a signal worth checking against, so participants checked.

That trade is familiar from every other navigation aid ever studied. More information, more accuracy, more deliberation. Whether it is a good trade depends entirely on whether your application values getting there or getting there right.

What to take from it

Scent can carry spatial structure, but only if you make it spatial. The gradient condition — intensity as a function of distance — is what produced the accuracy improvement. Scent as ambience produced bias. The variable has to mean something.

Fan-based delivery is the practical approach, and its limits shape the design. Fans direct airflow rather than releasing into a room, which gives you some directionality and much faster onset and clearance than a diffuser. But the response is still slow relative to a display refresh, and the resolution is coarse. Do not design an olfactory cue that needs to change quickly or specify a precise bearing.

Test for conflict, not just for presence. The interesting failure here was not “scent did nothing.” It was “scent did something wrong, consistently, without anyone noticing.” An evaluation that only asks whether users liked it would have missed the entire finding.

This is a path-integration result, which generalises further than navigation. Path integration is how you know where you are when you cannot see a landmark — it underlies the sense of a room’s extent, of how far you have come, of where the door was. Anything that perturbs it perturbs presence. Which is to say the relevance here extends well past wayfinding, into any immersive piece where the audience moves through a space and needs to retain a sense of it.

Related Reading