829 cross-sensory translation actions reveal four findings:
Olfaction's priority pathway through touch, the competitive resonance of scene and character,
absence as negative-space synesthesia, and 88.9% structural sharing vs. 3.7% lexical overlap among translators
Humans possess a fundamental but rarely consciously recognized cognitive ability: describing one sensory experience using the vocabulary of another. We say sounds are "bright," colors are "warm," and personalities are "sharp." This is not rhetoric, not literary imagination, but a foundational mechanism through which the human perceptual system processes complex information.
In cognitive science, this phenomenon is called synesthesia. In its narrow sense, synesthesia refers to a neurological phenomenon in which a minority of individuals automatically experience a secondary sensory perception when receiving a primary sensory stimulus (for example, seeing colors when hearing specific musical notes). But in broader linguistic and cognitive science research, synesthesia refers to the cross-sensory mappings pervasive in human language: describing taste through touch, sound through vision, emotion through spatial relations.
These cross-sensory mappings are not random. Research shows they follow systematic directional patterns. Spence (2011), in a large-scale crossmodal correspondence study, found that mappings between certain sensory channels are more natural, more frequent, and more easily understood than others. Williams (1976) observed even earlier that synesthetic metaphors in language tend to flow from "lower" senses (touch, taste) to "higher" senses (vision, hearing) — a principle known as the directionality of synesthesia.
But olfaction holds a special position within this system.
Unlike vision (red, blue, bright, dark) and hearing (loud, quiet, sharp, deep), human language contains virtually no native vocabulary exclusive to olfaction. We lack an independent system of olfactory adjectives. This means describing a scent must borrow the language of other sensory channels: calling an aroma "warm" (borrowing temperature), "silky" (borrowing touch), "transparent" (borrowing vision), "deep" (borrowing space). Olfactory expression is inherently dependent on synesthesia — there is no option to "not borrow."
References: Spence, C. (2011). Crossmodal correspondences: A tutorial review. Attention, Perception, & Psychophysics, 73(4), 971–995. | Williams, J.M. (1976). Synesthetic adjectives: A possible law of semantic change. Language, 52(2), 461–478. | Majid, A. & Burenhult, N. (2014). Odors are expressible in language, as long as you speak the right language. Cognition, 130(2), 266–270.
This is why olfaction — particularly the fragrance domain — is an ideal entry point for studying synesthetic language. Every sentence describing a scent is a forced cross-channel translation. The question is not "whether synesthesia occurs" but "which channel is borrowed, why that channel, and what happens during the translation."
If synesthesia is the underlying mechanism of olfactory expression, why choose tea fragrance as the research entry point?
Because tea fragrance has the largest "translation loss" of any perfume category. This loss is not information disappearing — it is a grammar mismatch. Perfumery grammar (top-middle-base notes, fragrance pyramids, concentration tiers) and Chinese tea culture grammar (cultivar and origin, processing craft and vintage, brewing rituals, literati aesthetics) are two entirely different meaning systems. When a tea-infused perfume needs to be translated for consumers, the narrator faces not "how to describe a scent" but "how to make two incompatible grammars converse."
Our data validates this judgment. Among the 4,096 possible narrative signatures defined by the six-layer density model, tea fragrance actually covers only 98 — a coverage rate of just 2.4%. No single signature holds a dominant position; no industry-recognized "standard narrative template" has formed. By contrast, most mature categories (floral, woody, oriental) have developed recognizable narrative conventions. The tea fragrance narrative space is highly dispersed — translators are each inventing their own translation strategies without a shared grammar to borrow from.
It is precisely this "no standard answer" state that makes tea fragrance an ideal window for observing synesthetic translation mechanisms. In categories with established narratives, you see convention; in tea fragrance, you see each translator's improvised cross-channel mapping strategies — and the underlying structure of those strategies is synesthesia.
Our analysis of 139 tea fragrance products and 829 cross-sensory translation actions aims to answer exactly these questions. Here are four main findings.
Among 829 bridge actions, we statistically classified all synesthetic mappings originating from olfaction by their target channel:
What this number means: when a person faces the scent of tea and needs to translate it into language, the first channel activated is not visual imagination ("what color does this scent look like" — only 4%) or spatial association ("what place does this scent remind me of" — only 3%), but somatic simulation — "what does this scent feel like to touch," "what temperature is it."
Touch and temperature together account for nearly half of all synesthetic translations. Vision accounts for only 4%. Here is what this 49% looks like in bloggers' actual language:
"Raw astringency" — a word simultaneously evoking interpersonal awkwardness and physical texture roughness
"Slowly enveloping" — a gentle, sustained wrapping sensation
"Pulling that sweetness back, reining it in"
The common feature of all three examples: bloggers are not describing abstract scent attributes but simulating tactile events the body is undergoing — roughness, being wrapped, pulling force. The olfactory signal is translated into physical processes the body can "participate in."
This finding needs to be understood in a larger context. Over the past decade, "pairing scents with visuals" has become a dominant communication paradigm in the fragrance industry — from brand advertising to packaging design to social media seeding, visual imagery is consumers' first interface with fragrance. Recent AI-driven synesthetic design experiments (such as MIT's Anemoia Device) also start from vision: using image generation models to create pictures corresponding to scents.
But our data shows that consumers' spontaneous synesthetic translation runs in the opposite direction. When confronted with tea fragrance, the perceptual system preferentially activates the tactile channel, not the visual channel. Two levels of explanation may apply:
Evolutionary level: Olfaction and touch both belong neuroanatomically to the proximal senses — requiring physical contact or very close proximity for activation. Vision and hearing are distal senses — capable of receiving signals from a distance. Cross-channel mapping between proximal senses may be more natural than proximal-to-distal mapping, because they share an "intimate contact" perceptual framework.
Tea's embodied memory: Tea as a beverage — not merely a scent — is first and foremost a bodily experience in human memory: the warmth of the cup in the hand, the texture upon entering the mouth, the sensation in the throat after swallowing. When the nose encounters tea fragrance, what is summoned first may not be the image of a tea garden, but the tactile feel and temperature of the cup in one's hand. The synesthesia of tea is not "smell → see" but "smell → touch."
If synesthetic design starts from vision (image → concept → scent) while consumers' perceptual synesthetic direction runs through touch (scent → touch → emotion), a directional mismatch exists between the design pathway and the perception pathway. One possible correction: embed tactile cues within visual communication — not "what color does this tea fragrance look like" but "what texture does this tea fragrance feel like." Vision as the communication medium, touch as the perceptual anchor.
It is worth noting that 49% is the category-wide average. Correspondence analysis (CA) results show that different sensory expression experts (bloggers) have significantly different channel preferences: some systematically favor touch, others favor visual-spatial, still others favor gustatory synesthesia. This leads to Finding 4 of this article — individual differences among synesthetic translators. But before that, two other structural findings.
When analyzing the six-layer density structure of narratives, we discovered a counterintuitive phenomenon: the co-occurrence rate of the Scene layer (C) and the Character layer (D) is extremely low.
Intuitively, a rich perfume narrative should simultaneously contain scene ("in a dimly lit tea room") and character ("like a person wearing a black coat"). But the data shows that when narrators decide to add depth beyond sensory translation (B-layer), scene and character are not complementary — they are competitive.
The third principal component (PC3, explaining 13.6% of variance) in the six-layer density PCA precisely captures this substitution relationship: C and D load in opposing directions on PC3 (C loading +0.778 vs. D loading −0.591). The C↔D correlation coefficient in the correlation matrix is only 0.17 — the lowest among all layer pairs.
Here is what this substitution relationship looks like in three real narratives:
Scene-type (C-only) — imiss香氛实验室 on Bvlgari White Tea
"Steaming hot" · "mist swirling" · "drifting leisurely"…
Character-type (D-only) — EROS闻香识人 on Thé Noir 29
"INFJ: defensive boundaries ('leave me alone') + hidden warmth (revealed inadvertently)"
Rare dual-layer type (C+D) — 闪闪好香 on Beast Youth · Ink
Scene: "Misty-rain tea mountains of Jiangnan" × Character: "Chinese old money, composed and unhurried"
One possible explanation comes from cognitive load. Scene construction and character projection are two distinct cognitive operations: scene requires the narrator to look outward — constructing a physical environment, arranging space, light, weather, and activity; character requires the narrator to look inward — projecting a psychological image, assigning identity, posture, gender coding, and social attitude. In a few-hundred-word perfume description, simultaneously completing two cognitively opposing constructions is costly.
A deeper possibility: scene and character may serve the same functional role in meaning construction — they are both "preparation for emotional landing." Data shows that regardless of whether you take the scene path or the character path, the E-layer (Emotional Shape) increases significantly (from the default 0.52 to 1.79–2.00). They are two different roads to the same destination, not two complementary building blocks.
This substitution relationship may reflect a more general constraint in human narrative cognition: when converting sensory experience into communicable meaning, people tend to choose one meaning-construction pathway and follow it deep, rather than simultaneously maintaining two parallel pathways. Whether this holds in other sensory categories — for example, whether wine flavor descriptions also choose between scene-building and character-building — is a cross-category question awaiting verification.
During the structured extraction of bridge actions, we discovered a special class of linguistic phenomena: some synesthetic mappings translate olfaction not into another "present" perception, but into a state of "absence" — not sinking, not rising, unfinished, suspended, blank space.
"This jasmine blossom suddenly hangs suspended in the rising steam of hot black tea"
We call this phenomenon negative-space synesthesia — conveying perceptual information more precise than positive description through describing the absence, interruption, or incompleteness of perception.
This is not an isolated case. Negative-space synesthesia recurs throughout tea fragrance narrative:
"No floral to stir things up, no citrus violently overpowering the tea, and certainly no powder barging in"
"Green-fresh, but not as green-fresh as green tea; smoky, but not as rich and smoky as black tea"
"Never over-expresses" · "Perfectly avoids the heavy, oppressive sillage typical of common white floral perfumes"
Among the 21 bridge actions flagged as "meaning excess," negative-space synesthesia appears 3 times — the most frequent type after "cultural semantic layer." They share a feature: positive classification labels (such as "spatial perception" or "dynamic-temporal") can capture the functional category of these expressions, but cannot capture their source of power — which comes from the naming of absence.
At a more macro level, the negation structure analysis of the tea fragrance category also confirms the prevalence of negative-space logic. Among 164 negation structures, 45% negate concentration and intensity — the category's identity is established not by declaring "what we are" but by declaring "what we are not." Negation is not a failure of description; it is the core grammar of the tea fragrance category.
This finding has a structural resonance with the concept of "留白" (negative space / blank-leaving) in Chinese aesthetic tradition. During extraction, we repeatedly encountered meaning excess from the cultural semantic layer — certain Chinese words (such as "骨感" [bone-quality], "留白" [blank-leaving], "克制" [restraint]) continue to carry cultural aesthetic weight that the classification framework cannot fully contain even after being coded as synesthetic bridges. "骨感" is not merely a structural metaphor; in the Chinese aesthetic context it points to the value judgment that "the interior matters more than the exterior." This suggests that synesthesia is not culturally neutral — the same bridge mechanism may carry different weights of additional meaning across different cultural language systems.
On the role of "negative space" in perceptual research, see: Arnheim, R. (1974). Art and Visual Perception: A Psychology of the Creative Eye. On the cognitive function of "虚" (emptiness) and "留白" (blank-leaving) in Chinese aesthetics, see Li Zehou's The Path of Beauty and its analysis of "意境" (artistic conception) — "意境" is essentially a form of structured negative-space perception.
When performing correspondence analysis (CA) on 829 bridge actions, a key finding emerged: the organizing variable of synesthetic translation is the translator, not the product.
Blogger-level CA's first two dimensions explain 88.9% of total inertia — nearly all structural information is captured in just two dimensions. By comparison, product-level CA explains only 59.4%. This means: the same perfume is translated into different sensory channels by different bloggers, while the same blogger tends to activate similar channels across different perfumes.
Each blogger has their own stable "synesthetic signature" — a systematic preference for translating olfaction into touch, visual-spatial, gustatory synesthesia, or dynamic-temporal channels.
An illustrative example: different bloggers facing the Bvlgari tea fragrance line — the same brand, the same product type — produce entirely different translations:
Oliver Walsh → Weight + Rhythm channel
"Light and not heavy" · "warm-warm, soft-soft" · "Layers by layers"
imiss香氛实验室 → Temperature + Material channel
"Cold-crisp" · "bright" · "smoky" · "immersive and restrained"
小空试香水 → Negation + Intermediate-state positioning channel
"Green-fresh, but not as green-fresh as green tea; smoky, but not as rich and smoky as black tea"
Same brand, three entirely different synesthetic signatures. This is not "inconsistent description" — it is three perceptual systems, facing the same stimulus, each activating different cross-sensory channels.
| Blogger | Nearest Dimension | Bridge Actions | Translation Style |
|---|---|---|---|
| EROS闻香识人 | Somatic channel | 66 | Touch and temperature as anchor, character projection as center of gravity |
| 香香学姐Luna | Somatic channel | 63 | High-efficiency sensory coding, compact format |
| 小空试香水 | Visual-spatial | 35 | Rich spatial imagination, cross-cultural comparative perspective |
| imiss香氛实验室 | Visual-spatial | 29 | Scene-driven narrative, environmental atmosphere priority |
| Oliver Walsh | Dynamic-temporal | 25 | Visual appearance imagery, first-person declaration |
| 微笑的二大爷 | Gustatory synesthesia | 12 | Taste-olfaction crossover, extensive tasting experience |
This finding carries important methodological implications, as it forms a precise structural parallel with a classical problem in the sensory evaluation field — panelist calibration.
In traditional sensory evaluation, systematic differences among panelists are a well-known problem. Even after standardized training, different assessors exhibit individual style differences when using the same descriptive vocabulary — some panelists are more sensitive to acidity, others to texture, some use extreme values more frequently. This is not noise but a structural characteristic of each individual's perceptual system. The standard approach is to incorporate panelist effects into statistical models, or align through multiple calibration rounds.
What we observe in blogger data is exactly the same phenomenon, except occurring in natural language rather than standardized assessment. Channel preference differences among bloggers are not "inaccurate descriptions" — they are each person's perceptual system, when facing the same olfactory stimulus, structurally tending to activate different cross-sensory channels.
But there is a crucial distinction: panelists share a vocabulary after training, with differences expressed in usage frequency and intensity. Bloggers have received no unified training, and differences are expressed in the vocabulary itself. Our lexical analysis shows: lexical overlap among the 9 bloggers' bridge vocabulary is only 3.7%. 96.3% of vocabulary is personal style.
Yet what they share is dimension — all use touch to translate olfaction, just with one saying "silky" and another saying "dense." 3.7% lexical sharing + 88.9% dimensional structure sharing — this combination means that extracting category-level synesthetic structure from natural language is viable, provided the unit of analysis is the bridge action (cross-channel translation), not vocabulary (specific word choice). This provides methodological justification for deriving category-level conclusions from blogger data.
Individual differences among synesthetic translators manifest not only in sensory channel selection but profoundly in meaning-layer construction. A vivid case is "tea fragrance = introvert" — a narrative framework that recurs throughout the tea fragrance category. Multiple bloggers link tea fragrance to introverted personality. But when we disassemble each blogger's specific construction of "introvert," we find that the same word points to entirely different people.
Six translators, the same category, the same cluster of perceptual qualities related to "quiet," "restrained," and "non-overflow" — yet the Character layer is constructed in entirely different ways: classification labels, internal spectrum, scene emergence, visual imagery, cultural identity, or outright absence.
When a brand sees a market research report stating "tea fragrance consumers are introverts," this conclusion carries nearly zero information — because "introvert" points to entirely different character constructions across different perception-expression systems. It is not a label that can be directly applied to product positioning, but a surface consensus that needs to be decomposed into specific dimensions. The same logic applies to any seemingly unified consumer language label — "premium feel," "cool-toned," "healing-type" — in different translators' hands, they may point to vastly different perceptual structures and emotional shapes.
This also explains why the D-layer (Character Projection) is completely absent in 56% of tea fragrance product narratives. It is not that consumers don't care about character — it is that character construction is the highest-cognitive-cost layer among all six. It requires the translator to make a series of precise choices: labels or scenes? Psychological type or cultural identity? Project onto the wearer or onto the tea itself? Most narratives, facing this decision cost, choose to skip.
Touch priority, scene-character substitution, negative-space synesthesia, translator individual differences — these four findings are not four independent conclusions. Together, they sketch a structural portrait of synesthetic language:
First, synesthetic translation has directional preferences, and these preferences may be category-specific. The primary synesthetic landing point of tea fragrance is touch. Coffee might be temperature. Incense might be space. This means no universal "sensory → language" mapping model exists — each category needs its own synesthetic map.
Second, meaning construction faces cognitive constraints, manifesting as path selection rather than layer stacking. The competitive relationship between scene and character suggests that when converting sensory experience into communicable narrative, people do not progressively add information onto a framework but select one of multiple possible paths and follow it deep. The source of narrative depth is not "covering everything" but "choosing one path and going all the way."
Third, absence and negation are active expressive devices in synesthetic language, not information voids. Tea fragrance category identity is built through negation; negative-space synesthesia conveys precise perceptual information through describing absence. This poses a challenge for natural language processing: standard sentiment analysis and keyword extraction cannot identify the information carried by "what is not said." A complete synesthetic analysis framework must be able to encode negation and absence.
Fourth, synesthetic translation structure is stable at the category level but unique at the individual level. 88.9% dimensional structure sharing + 3.7% lexical overlap — this combination means extracting category-level synesthetic structure from natural language is viable, provided the unit of analysis is the bridge action (cross-channel translation) rather than vocabulary (specific wording). This provides methodological justification for deriving category-level conclusions from blogger data.
This study is an initial exploration based on a single category — tea fragrance. The following questions await further investigation:
Is touch priority a feature of tea fragrance specifically, or a universal feature of olfactory synesthesia? Tea's embodied memory as a beverage may reinforce tactile channel activation. If one analyzes purely non-beverage fragrances (such as ambient scenting or fabric fragrance), does touch remain the primary landing point? If not, what replaces it?
Is the C↔D substitution a constraint of linguistic expression or of cognitive processing? If the competitive relationship between scene and character exists only in linguistic expression (it is difficult to do both things at once in a paragraph of text), then multimodal communication (text + image + video) might break this constraint. But if it is a cognitive processing constraint — people genuinely tend to choose one meaning pathway when comprehending sensory experience — then simultaneously conveying scene and character in multimodal communication may cause cognitive overload rather than enrichment.
Is the relationship between negative-space synesthesia and Chinese aesthetic tradition causal or co-occurring? We found a structural resonance between negative-space synesthesia and cultural concepts like "留白" (blank-leaving) and "克制" (restraint) in Chinese blogger texts. Is this because Chinese aesthetic tradition has strengthened the capacity to express negative-space perception, or because negative-space synesthesia is a universal human perceptual mechanism that simply has richer vocabulary to capture it in Chinese? Cross-linguistic comparative data is needed.
Are translators' synesthetic signatures stable? We observed that each blogger has preferred channels. But is this a stable individual trait (akin to a panelist profile in sensory evaluation), or a strategic choice that shifts with content type, target audience, or publishing platform? If the same blogger also favors the same channel when describing skincare texture, then "synesthetic signature" is closer to an individual perceptual characteristic rather than a writing strategy.