Yellow for bergamot, brown for caramel lactone. When people smell a material and choose a color that suits it, their answers can have things in common. A 1996 study examined this correspondence using both color names and actual color samples.
The questions were whether associations between smells and colors are shared across people, and whether they persist over time. Two experiments and a retest about 2 years later revealed some group-level tendencies within people’s freely chosen associations.
Describing 20 smells with 11 color names
The first experiment involved 94 people aged 18–40, recruited through newspaper advertisements. They were screened by self-report for their sense of smell and nasal condition, and also underwent a color-vision test. This was not a study of fragrance experts.
The 20 fragrance materials included bergamot oil, lavender oil, jasmine absolute, and patchouli oil. The researchers used a mixture of natural extracts and synthetic materials and, in preliminary testing, adjusted perceived intensities to approach that of a reference material. The samples were presented in white containers that concealed the color of the contents and were labeled with three-digit codes.
Participants smelled each material and distributed a total of 5 points among 11 color names: red, orange, yellow, green, blue, purple, brown, white, pink, gray, and black. They could give all 5 points to one color or divide them among several. Odor order was randomized for each participant, with at least 20 seconds between presentations.
For comparison, they first completed tasks assigning colors to mood words and describing line shapes through moods and colors. The aim was to check whether an existing method for investigating cross-sensory correspondences could be replicated and extended to smell.
Some smells converge on brown; others spread across several colors
Analysis of the point distributions showed that color choices were uneven for many materials. In a test corrected for ties arising from the many zero scores, all 20 showed distributions different from random point allocation. Before correction, lavender, jasmine, and star anise had not reached the significance threshold, so the results also differed according to the analytical method.
The patterns varied between materials. Caramel lactone attracted notably high scores for brown, whereas points for an artificial civet material were spread across brown, white, gray, black, and other colors. The findings fit the idea that different smells elicit different patterns of color choice, rather than that each smell has one correct color.
High-scoring colors were not simply the result of a few people assigning many points. Colors with mean scores in the top quarter received points from about 36% of participants on average. There were shared tendencies within the group, even without unanimity.
Similarities in color allocations remained about 2 years later
Fifteen of the original participants completed the same task again about 2 years and 2 months later. For each smell, the researchers compared the group-mean profiles of points assigned to the 11 colors at the first test and retest for these same 15 people.
The mean correlation for odors was r = 0.53. This was close to the value of 0.57 for describing lines with colors and lower than the value of 0.84 for describing mood words with colors. Nine of the 20 smells had individually significant correlations, and the highest-scoring color name was unchanged for 7.
This indicates some persistence in color associations at the group-average level. It does not measure the probability that each individual will choose the same color every time, and the retest involved only 15 people. The authors also distinguished the finding from the nearly unchanging individual experiences associated with strong synesthesia.
Choosing from 1,565 color samples rather than words
The word red may bring different shades to mind for different people. In the second experiment, therefore, 50 people aged 18–40 who had not taken part in the first experiment smelled the same 20 materials. This time, they selected the single chip that best represented each smell from 1,565 Munsell color samples.
These samples specify hue, chroma or saturation, and value or lightness numerically. A dedicated light box was used, with the room lights turned off, to standardize viewing conditions. The odor container was placed near the participant’s nose, so participants smelled it without holding it.
For 13 of the 20 materials, the selected hues showed statistically significant clustering. Selected lightness and saturation also differed between materials. The results suggest that correspondence with smell concerns not only broad color names but also impressions involving different shades.
A further 30 people then named color chips close to the group mean for each of those 13 odors. For 10 of the 13 chips, the most frequently used color name had also ranked highly in the first experiment. Examples included yellow for bergamot, brown for caramel lactone, red for cinnamaldehyde, and pink for methyl anthranilate.
There were also mismatches. For artificial civet and star anise, the associations with brown or black that stood out in the first experiment were not reproduced. The authors suggested possible explanations including insufficient representation of browns in the chip set and the absence of pure gray and black samples. The way the available choices are constructed also matters to the color impressions obtained.
Where do shared associations come from?
The authors emphasized that the color-name task and the task using actual color samples provided converging evidence. Odor associations are not determined solely by separate personal memories: there can also be something shared across people.
These experiments alone do not establish the mechanism. For example, the authors suggested that a cinnamon-like smell might evoke red cinnamon candy. Yet color preferences also occurred for materials that are generally hard to name, and lavender was not simply associated with purple or blue. The relative contributions of object names, life experience, and sensory qualities remain open questions.
This was not a study of colors appearing before someone’s eyes when they smelled an odor. It examined correspondences between senses by asking participants to choose colors that suited smells. Moreover, the results concern adults recruited in one area and 20 fragrance materials; treating them as a universal color chart would require further tests across cultures and materials.
Summary
Choosing colors to suit smells revealed shared tendencies among participants, such as brown for caramel lactone and yellow for bergamot. All 20 materials showed nonrandom color-name allocations after the statistical correction, and the hues chosen from actual color chips clustered for 13 materials. Lightness and saturation also differed between materials, and the two methods broadly corresponded.
About 2 years later, a moderate association remained between the same 15 people’s group-average point allocations. Representing smells through color need not be merely a passing whim: it can draw on associations that are shared and persistent to some extent. The study examined that possibility through both words and color samples.
Research context and disclosures: At the time of the study, all three authors were affiliated with Givaudan-Roure Fragrances. The published paper’s correspondence note lists Synesthetics, Inc. as the first author’s new affiliation. The acknowledgments name contributors to the experiments, fragrance materials, statistics, and other aspects of the work, but no separate funding or conflict-of-interest statement could be identified.
Reference
Gilbert, A. N., Martin, R., & Kemp, S. E. (1996). Cross-Modal Correspondence between Vision and Olfaction: The Color of Smells. The American Journal of Psychology, 109(3), 335–351. Original paper / Author-shared full text
