Is Recalling a Scene from Its Smell Related to How Easily We Can Describe It?

When we encounter a nostalgic smell, sometimes we simply know we have smelled it before; at other times, we recall the scenery of a particular place. It may be important not to treat these as the same process in odor memory.

A study analyzing experimental data from 106 people found that remembering an odor was associated with the strength of its pleasantness or unpleasantness and the number of words used to describe it. Recalling its associated scene, meanwhile, was related to personal familiarity and the ability to describe it with words that did not overlap with descriptions of other odors.

Separating “I know this” from remembering a scene

The study aimed to divide odor-evoked retrieval into recognition of the odor itself and recall of an associated scene, and to investigate the features related to each process.

The researchers pooled data from four studies using the same basic procedure. The sample comprised 106 young French adults: 65 women and 41 men, with a mean age of 22. They were not all newly recruited into a single experiment.

Over three days, participants explored a different landscape image each day for seven minutes. Clicking designated locations on the image caused a device to deliver an odor. Each landscape was paired with three odors, giving nine odors across the three days. Participants were encouraged to explore thoroughly but were not explicitly instructed to memorize anything.

The combinations of odors, landscapes, and locations within images were not designed to be semantically congruent. Assignments varied between participants so that pre-existing associations alone would not provide the answers.

Testing memory for odors and scenes on day four

On day four, participants encountered the nine previously presented odors and nine new ones. They first answered whether they had smelled each odor before. If they answered yes, they selected the corresponding image from the three landscapes and then a location within it. The interval between initial exposure and testing was 24 to 72 hours.

The recognition analysis was restricted to 949 trials involving previously experienced odors. It was not an overall measure of old-versus-new classification accuracy that also included correct rejections of novel odors.

Scene retrieval was analyzed in the 736 trials where participants correctly recognized an odor. A response counted as correct if they selected at least the correct landscape image. They did not have to identify both the image and its location correctly.

Averaged across participants, about 77% of previously experienced odors were recognized, and the corresponding landscape was correctly recalled for about 55% of those. The numbers illustrate that remembering an odor does not necessarily mean remembering its associated scene.

Rating and describing the odors afterward

After the memory test, participants rated the odors’ pleasantness, intensity, and familiarity, and described them freely in words if they could. Descriptions could take any form, including a possible source, a metaphor, or a feeling.

Even equally short descriptions differ in content: identifying something as “eucalyptus” is not the same as saying “I don’t know.” A long description might recount rich memories or reflect uncertainty about an odor’s identity. The researchers analyzed such language rather than limiting themselves to whether people supplied the correct name.

Verbal richness was measured by the number of words after removing articles, conjunctions, and generic terms such as “odor,” and standardizing word forms. Another measure captured how much the words overlapped with the same person’s descriptions of other odors. This was calculated from overlap between sets of words, not from an AI measure interpreting deeper meaning.

The study’s “emotional strength” was not a score from an independent emotion questionnaire either. It represented how far a pleasantness rating lay from neutral, regardless of direction.

Remembering an odor: pleasantness, unpleasantness, and richer descriptions

The researchers used machine-learning models to classify memory outcomes from these features. Data were split so that the same participant did not appear in both training and evaluation sets, and the researchers examined which cues the models used.

For odor recognition, the main cues were the number of descriptive words, the distance of pleasantness ratings from neutral, and odor intensity. Richly described and strongly affecting odors were more likely to push the model’s prediction toward successful recognition.

In the machine-learning analysis, unpleasant odors in particular were associated with recognition. A separate statistical analysis, however, also found greater recognition at both ends of the pleasantness scale. The findings therefore cannot be reduced to “only unpleasant odors are remembered.”

Recalling scenery: distinctive words and familiarity

The relevant cues changed when the researchers examined whether participants could recall the corresponding landscape after recognizing the odor. Here, familiarity and the absence of overlapping words with other odor descriptions were important.

Compared with using the same terms, such as “sweet” or “floral,” for many odors, describing each with different words made the model more likely to predict correct landscape recall. The authors suggest that conceptual cues distinguishing odors may reduce confusion between scenes.

The relationship with familiarity was not simply an upward slope. Both highly familiar and almost entirely unfamiliar odors were associated with better retrieval, while intermediate familiarity was less favorable, producing a U-shaped relationship.

A nearly novel odor may be memorable because of its novelty, while a well-known one may provide access to existing knowledge and memories. These are interpretations of the results, however, not distinct brain mechanisms directly measured and confirmed in this experiment.

More words do not necessarily mean better memory

Crucially, odor ratings and descriptions were collected after the memory test. This was not an experiment showing that people trained to describe odors in detail beforehand remembered them better. Possibilities such as memorable odors also being easier to describe require separate investigation.

The models’ predictive performance was also limited, especially for recognition failures and scene retrieval. These findings concern laboratory memories in young French adults, not a model that predicts every instance of important life memories coming back to mind.

The paper reports financial support from L’Oréal Research & Innovation for one researcher, and the authors declared no conflicts of interest.

Summary

At the stage of knowing that an odor had been smelled before, emotional strength and richness of language were relevant. At the stage of recalling an associated landscape, distinctive descriptions and familiarity mattered.

In particular, the association between better scene recall and both extremes of familiarity, rather than its middle range, suggests that odor memory is not determined simply by how accustomed we are to a smell. Distinguishing remembering an odor from recalling the scene beyond it allows a more detailed consideration of how language and emotion relate to memory.

References

What makes a scent trigger a memory? A cognitive decomposition of odor-evoked retrieval. iScience. 2025;114467. doi:10.1016/j.isci.2025.114467