Citrus, floral, woody. Arranging scents on a wheel or map can make it easier to find similar fragrances or explain your preferences. These arrangements reflect not only perfumers’ experience: shared patterns can also be found in rating data.
A 2009 study analyzing two existing sets of fragrance-material evaluations identified dimensions including a contrast between fresh and warm–powdery qualities. Yet the same word can occupy a different position on a map when a different fragrance material serves as its reference. The central question was what needs to be standardized to create a shared odor map.
Exploring relationships between odor descriptors in two datasets
The first dataset involved six perfumers comparing 309 compounds with 30 reference materials. Bergamot oil, for example, represented fresh, and eugenol represented spicy; each characteristic was rated from 0 to 9. The researchers reanalyzed existing values compiled through averaging and consensus.
The second dataset comprised written odor descriptions of 44 natural and 75 synthetic materials, 119 in total. After removing infrequent terms and words describing intensity alone, among others, the researchers encoded which of 66 descriptors appeared for each material. They tried both equal weighting of primary and secondary characteristics and different weights for the two.
They used principal component analysis, a method that summarizes patterns of co-occurrence among many descriptors in a small number of dimensions. Rather than recruiting new members of the public to study perfume preferences, this study explored the structure of odor descriptions through experts’ ratings and word use.
Fresh on one side, warm and powdery on the other
A prominent pattern in the first dataset placed fresh, citrus, watery, and green close together, opposite terms such as sweet, powdery, and balsamic. Powdery had the strongest negative correlation with fresh.
The authors interpreted this dimension as a contrast between freshness and warmth. They also considered its relationship with familiar fragrance descriptions, such as rich versus light and watery versus dry. Rotating the plot from the other dataset for comparison produced a similar arrangement, with fresh and warm opposing one another.
Another major dimension contrasted floral and soft descriptors with terms relating to woods and herbs. Comparison with product classifications in contemporary perfume guides showed a close correspondence with differences in the terms commonly used for women’s and men’s fragrances.
Here, the gender-related interpretation rests on product categories and the frequency of descriptor use at the time. It does not identify a biological boundary between men’s and women’s sense of smell or preferences, nor does it restrict who can choose a fragrance.
Even balsamic changes position when the reference changes
An important finding was that differences between maps could sometimes be explained by examining their reference materials, rather than dismissing them as mere variability in ratings.
For example, the traditional Jellinek map used for comparison represented balsamic with Peru balsam, which evokes sweet vanilla. The numerical dataset reanalyzed here instead used olibanum resinoid. Because the latter is not equally sweet, the authors interpreted this as explaining why balsamic occupied a position farther from the sweet side.
Similarly, patchouli oil served as the reference for dusty in this dataset, but was frequently selected to represent earthy in another survey of perfumers. Changing the reference convention can change the label attached to the same material’s position. Sharing a map therefore requires agreement not just on words but on which material, smelled at which concentration, each word represents.
Two proposed changes to the 2008 Fragrance Wheel
The researchers also compared their results with Michael Edwards’ 2008 Fragrance Wheel, from a guide classifying more than 5,700 perfumes at the time. Its overall arrangement overlapped with the analytical results, but they proposed two changes to improve consistency.
One was to move fougère from the center of that wheel to a position between dry woods and citrus. This would place fougère, with its fresh herbal and lavender impressions combined with a mossy foundation, between those categories.
The other was to swap fruity and water. Fruity and floral lay close together in the rating data, and the swap also improved correspondence with the product classifications of the time. These were research proposals for the 2008 version, not statements about the current wheel’s arrangement.
Shared patterns captured by a map, and differences it leaves out
The authors considered the similar arrangements obtained from two different datasets to support the possibility of a common odor map. They concluded that agreement among perfumers about reference materials would be a prerequisite.
However, the first two dimensions explained only 31.7% of the variability in the numerical-rating dataset and 19.5% in the verbal dataset. Other dimensions captured remaining differences, including animalic and fatty–aldehydic characteristics. A two-dimensional map is a useful summary, not a way of fitting every aspect of odor onto a plane.
Some of the older maps used for comparison also carried labels suggesting effects, such as calming, stimulating, or sensual. What this study examined, however, was the relationship between odor evaluations and descriptors. The authors themselves noted that such physiological or psychological effects of individual fragrance materials require separate demonstration.
Summary
Analysis of existing perfumer evaluations revealed structures overlapping with traditional odor maps, including a contrast between fresh and warm–powdery scents. The researchers also proposed moving fougère and swapping fruity and water on the 2008 Fragrance Wheel.
At the same time, descriptor positions depended on the reference materials. When reading an odor map, adjacent categories can provide useful clues, but considering the materials and words behind those categories can make the map more concrete and useful.
One author was affiliated with Procter & Gamble at publication, and the acknowledgments list research support from the company. The other author received postdoctoral support from the Fulbright Program and Spain’s Ministry of Education and Science. No separate conflict-of-interest statement was identified.
Reference
Zarzo, M., & Stanton, D. T. (2009). Understanding the underlying dimensions in perfumers’ odor perception space as a basis for developing meaningful odor maps. Attention, Perception, & Psychophysics, 71(2), 225–247. https://doi.org/10.3758/APP.71.2.225 (Public original-paper PDF)
