Making of: 103 cards for a vocabulary that has no dictionary
Deck v1.0 · First working version
This is the first public working version of the Canephora Descriptor Deck. Its structure and sources have been verified; its editorial texts are built to be tested, corrected and refined through use.
Where this started
In May 2025, a research team around neuroscientist Dr. Fabiana Carvalho published a research-based flavour wheel developed specifically for Coffea canephora: the Canephora Flavour Wheel (Carvalho, F.M., Alves, E.A., Artêncio, M.M., Cassago, A.L.L. & Pereira, L.L., Scientific Reports 15, 16643, 2025 — open access).
Forty-nine professional graders in Brazil and Switzerland cupped 67 Canephora samples from 13 countries, using a Rate-All-That-Apply list alongside the CQI/UCDA fine robusta protocol. The Swiss sessions took place at the ZHAW Coffee Excellence Center in Zurich. From 202 raw descriptors, the research team consolidated 170 terms; the 103 descriptors with an average mean score of at least 0.5 were included in the wheel.
The wheel matters because the industry's established sensory vocabulary was not built specifically around this species. The Coffee Taster's Flavor Wheel — first published in 1995 and redesigned in 2016 together with World Coffee Research — is supported by the separate WCR Sensory Lexicon.
As summarized in the Canephora study, that lexicon was developed from 105 coffee samples. Robusta appeared only among the commercial coffees, while the specialty subsets were Arabica. The positive sensory range of fine Canephora was therefore not represented in those specialty subsets.
The Canephora wheel documents what was missing: an umami category with descriptors such as shōyu, tomato, coconut water and mushroom; positive woody notes such as barrel and cedar; aromatic spices; alcoholic-fermented descriptors; and Brazilian terms that bring the tasting culture of an important producing country into the record.
The gap we decided not to fill
The study provides the terms, their hierarchy and an average mean score for each descriptor.
What it does not provide — and does not claim to provide — is a reference lexicon in the strict sense: it does not publish a sensory definition and standardized reference preparation for every Canephora descriptor.
For the Coffee Taster's Flavor Wheel, that supporting layer exists in the WCR Sensory Lexicon: 110 attributes with definitions, reference materials and intensity anchors. For Canephora, an equivalent public reference lexicon does not yet exist.
Where a Canephora descriptor has an exact or sufficiently close counterpart in the WCR Sensory Lexicon, the deck identifies it transparently as a contextual cross-reference. Where no sufficiently close counterpart was identified, the card works instead with sensory memory and association.
The distinction matters: the editorial text on the cards is there to help build sensory language, not to present itself as an official definition of the Canephora descriptor.
How the deck was built
The data layer rests on three sources: Table 3 of the Canephora study for the consolidated terms and average mean scores, the published wheel for hierarchy and spelling, and the WCR Sensory Lexicon 2.0 for contextual cross-references.
All 103 wheel descriptors were checked against the source material and compared individually with the 110 WCR attributes. The result was 40 exact matches, 12 documented near-matches and 51 descriptors with no sufficiently close WCR counterpart.
That verification has a limit. It can establish the terms, hierarchy, scores and mappings. It cannot establish whether an editorial memory anchor actually helps someone recognize or describe a sensory impression in the cup. That is what practice has to show.
What is on the cards
All 103 cards represent Tier-3 descriptors — the outer ring of the Canephora Flavour Wheel.
The front shows the family anchor, one of the study's 15 colour-coded Tier-1 categories, followed by the hierarchy and the individual descriptor.
Each card shows the descriptor's average mean score (AMS) from Table 3 of the study. AMS combines how frequently a descriptor was selected and how intensely it was rated across the evaluated coffees. It is not a percentage, a cup score or a quality judgement.
Among the 103 descriptors included in the wheel, values range from 21.4 for Caramel to the 0.5 inclusion threshold, shared by six terms including Fishy.
Below the score, the cards follow one of two paths.
Fifty-two cards are contextual WCR cross-references. They contain a short orientation written in our own editorial language and a practical “Think of” exercise using familiar objects, foods or situations.
Forty of these mappings are exact term matches. Twelve are documented near-matches. Where the terminology differs, the corresponding WCR attribute is named explicitly — for example, Caramel maps to “Caramelized”, Meaty to “Meaty/Brothy” and Garden pea to “Peapod”.
These references provide context. They are not presented as official Canephora definitions.
Fifty-one cards are sensory-memory prompts. They state plainly that the Canephora study provides no sensory definition for the descriptor and that no sufficiently close WCR counterpart was identified.
Instead, they anchor personal recall in a concrete scene — sharpening a pencil for Cedar, or standing at a harbour at low tide for Brackish water — and ask which qualities of that memory can be recognized in the cup.
Seven of these cards begin with a short factual context note, because an unfamiliar referent such as cupuaçu or cajá is difficult to recall without first knowing what it is.
A version 1, on purpose
The card texts are a first version, and they are built to be revised.
The paraphrases and memory prompts are written referent-first: the scene comes before the perception. Standardized sensory references can align people around defined attributes and intensities. This deck has a different purpose.
It is a recognition and language-training tool for people building a Canephora vocabulary. It is not a standardized sensory reference system: the memory anchors are there to help recognition and association, not to prescribe what a descriptor must taste or smell like.
There is also some freedom in this. The data gives the deck its structure; the cup does not have to behave like a database. Building associations, finding unexpected flavours and learning to name what you taste is part of the fun — and part of getting excited about Canephora.
The anchors are meant to be tested at the cupping table. Practice and feedback will show which associations help, which do not, and where the deck can improve.
The deck version appears on the colophon card, so the state of the material remains identifiable as the deck develops.
Canephora Sense aims to make sense of Canephora in Switzerland — not as something published once and left untouched, but as a process that stays open to better evidence, better language and better sensory practice.
This deck carries that idea into card form: verified where verification is possible, transparent where it is not, and deliberately unfinished where only practice can complete the work.
Sources and thanks
Descriptor terms, hierarchy and average mean scores are cited from:
Carvalho, F.M., Alves, E.A., Artêncio, M.M., Cassago, A.L.L. & Pereira, L.L. (2025). “Development of a flavour wheel for Coffea canephora using rate-all-that-apply.” Scientific Reports 15, 16643. https://doi.org/10.1038/s41598-025-99921-w
The article is published open access under CC BY-NC-ND 4.0. This page and the deck reproduce factual data with attribution; they do not reproduce the article text or its wheel graphic. The original wheel can be viewed through the publication.
The WCR Sensory Lexicon is used solely as a contextual cross-reference. Its definitions are not reproduced.
The Canephora Flavour Wheel itself resulted from a community effort involving the research team, the participating professional graders, the Canephorum events and the producers and organizations that contributed coffee samples. This deck is a downstream tool that aims to be worthy of that groundwork.
Corrections and practice feedback are welcome. If an anchor does not trigger the memory it should, or an error is found anywhere in the set, tell us. This first version is meant to improve through use.