When you stop smelling anything: olfactory fatigue, and how many materials to evaluate at once

Anyone who has run a long evaluation session knows the feeling. The first five are clear, the eighth is vague, and by the twelfth you are not sure the strip in your hand has anything on it at all.

The culprit is olfactory adaptation. Tiredness has nothing to do with it, and neither does material strength. Adaptation follows rules specific enough to design around, so a session's second half doesn't have to be wasted.

Adaptation is stimulus-specific

Dalton's 2000 review in Chemical Senses puts it plainly: repeated or prolonged exposure to an odorant produces a stimulus-specific decrease in sensitivity to that odorant, and sensitivity recovers over time in the absence of further exposure.

Three things follow.

1. It is specific. Adaptation mainly affects the odour you have been smelling rather than your sense of smell as a whole. "My nose is broken" is usually the wrong diagnosis: you have lost that one material, and something genuinely different will still register.

2. Magnitude and duration depend on concentration and exposure length. Stronger and longer means deeper adaptation and slower recovery. A working rule falls straight out of that: evaluate dilutions. The earlier piece on dilution argued for them so you can smell what a material really is; here is a second reason. Neat material costs you the rest of the session.

3. Switching nostrils does not work. Dalton notes that stimulating one nostril produces adaptation in both: more profound, with slower recovery, on the stimulated side, but the other side drops too. Adaptation therefore has both peripheral and central components.

So "this side is dull, I'll use the other one" gets you nowhere. Stop and let the nose recover.

Cross-adaptation follows similarity, not structure

In 1992 Cain and Polak ran a neat experiment. They chose two molecules that are structurally very different but smell almost identical, both bitter-chocolate: trimethyl pyrazine (TMP) and 2-propionyl-3-methyl furan (PMF).

Subjects breathed an adapting concentration continuously for two minutes and rated test concentrations during brief respites. The results:

• With the adapting stimuli matched in perceived intensity, self-adaptation came out equivalent and cross-adaptation between the two was symmetric.
• The second experiment is the important one. TMP and PMF cross-adapted strongly with one another, yet their effects on three other odorants (anethole, ethyl butyrate and 2,3-pentanedione) were substantially weaker.

Cross-adaptation reflects perceptual similarity, not chemical similarity. Two molecules from different structural worlds that smell the same will cancel each other out.

For session design this means the thing that wears you down is a run of similar materials in a row. The total count matters much less.

Similarity in our data

If cross-adaptation follows similarity, it is worth knowing how similar materials within one odour family actually are.

We took 3,790 materials with odour data and supply information (each with at least three descriptors) and computed pairwise overlap of their descriptor sets (Jaccard).

One methodological note: the calculation deliberately excludes the primary odour family term itself. Materials in the same family obviously share that word; including it would inflate within-family similarity by definition and make the result meaningless.

Also, the median material carries only five descriptors, so absolute overlap is low by construction. The ratio is what matters, not the raw figure.

	Mean overlap	Pairs sharing no descriptor at all
Within the same family	0.090	45%
Across families	0.036	70%

Within-family similarity is 2.5× the across-family figure, and seven in ten cross-family pairs share not a single descriptor.

That is why evaluating ten citrus materials in a row is the worst possible schedule. They overlap heavily, they cross-adapt, and by the fifth you are guessing.

How to arrange a session

1. Interleave families; never run one type consecutively. To compare five citrus materials, do not smell them back to back. Citrus → woody → citrus → balsamic → citrus. The materials in between overlap little, so they cross-adapt little.

2. Keep a session to roughly 8–12 materials. No single study gives this number directly; it is an operating figure derived from "adaptation depends on exposure". What you should actually do is calibrate your own. Re-smell the first material at the end of every session, and if it has gone noticeably weak, distrust the back half and do fewer next time. I fade fast myself; past ten strips I catch myself guessing, so I tend to stop early.

3. Use the first material as a reference and repeat it last. That is how you run rule 2, and it doubles as honest quality control: you know whether the session's data is usable or needs redoing.

4. When fatigued, leave the room. Dalton's point is that recovery happens in the absence of further exposure. A break taken in a room full of fragrance does nothing for you. Go outside, or at least somewhere odourless.

5. Dilute, small, brief. Both concentration and duration deepen adaptation. Touch the strip to the dilution rather than soaking it, and keep your sniffs short.

Two common misconceptions

"Coffee beans reset your nose." Popular, but inconsistent with stimulus specificity. Coffee is simply another strong odour: it adapts you to coffee and leaves your adaptation to the previous material in place. To recover, stop smelling.

"My nose isn't working today." Usually you are adapted to one class of odour. Try something from a completely different family. If that registers, your nose is fine and the running order was the problem.

References

P. Dalton, Psychophysical and behavioral characteristics of olfactory adaptation, Chemical Senses, 25(4), 487–492 (2000). doi:10.1093/chemse/25.4.487

W. S. Cain & E. H. Polak, Olfactory adaptation as an aspect of odor similarity, Chemical Senses, 17(5), 481–491 (1992). doi:10.1093/chemse/17.5.481

---

Next: record keeping — what an evaluation note should actually contain, why "smells nice" is not data, and what a note looks like that still makes sense six months later.