Understanding Coffee Professionals’ Flavor Vocabulary: Early Results of a Brief Study - Issue 12 Online Exclusive
In Issue 12 of 25, Dr. ILJA CROIJMANS explored why the vocabulary we use to describe aromas matters, and shared findings from a comparative study of wine and coffee experts first published in 2016.
“An update on the status of coffee vocabulary is very much called for,” wrote Ilja. And so, when presenting this work at Sensory Summit in Zurich—complete with wine and coffee samples—Ilja also collected data. Here, he shares some of the first findings of his brief study of the attendees of “Translating Findings on Wine Language to Coffee.”
This presentation offered us a unique opportunity to update our understanding of coffee professionals’ flavor vocabulary, and more specifically, to explore how different educational backgrounds impacted attendees’ ability to describe coffee and wine.
We began the session with a 25-minute introduction exploring some of the findings on general odor language and wine expert language[1], before asking attendees to take around 15 minutes to taste and describe the aroma and flavor of three red wines. They were also asked to rank the wines for overall liking and to try and identify the wines tasted by their neighbor by discussing aroma and flavor. Crucially, each set alternated between being identified by numbers (1-2-3) or letters (A-B-C) and was ordered differently for each participant. The three wines we selected for this activity were of similar color and a moderately similar taste (i.e., we didn’t choose anything with a strong oak profile).[2]
After this, we shared information on the similarities between wine and coffee with a particular focus on coffee language and Q Graders[3] before asking attendees to take ten minutes to taste three coffees. Again, they first described each of the coffees’ aroma and flavor for themselves, ranked the coffees for liking, and matched the coffees with their neighbor by talking about smell and taste. Both the selection of the coffees for tasting and the application of the coffee labels were in line with the selection wine and application of wine labels. [4]
Throughout the session, we dimmed the lights to minimize visual cues (color, clarity, etc.) and asked attendees to complete a survey using their own device (three participants completed the survey on paper).
Hypotheses and Design
The survey tested two different things: communicative success and consistency in free description. By tracking communicative success (or the overall number of correct matches for coffee and wine), we wanted to test a hypothesis that, as the Q Grader program emphasizes consistency and communicative efficacy in the use of descriptions, Q Graders would have more correct matches than other types of coffee education. Overall, we expected coffee professionals to have more correct matches for coffees than for wines.
By looking at the consistency in the free descriptions given (looking at the consistency between attendee language in the three groups for all stimuli), we hope to be able to track whether or not recent work to find a common set of references to describe flavors in coffee was effective in creating a shared vocabulary for coffee professionals. In future analyses, we will test whether coffee experts would be more consistent when describing coffee than describing wine, with Q Graders able to describe both categories most consistently.
In future, using the data gathered in the second category (consistency in free description), we’ll also analyze the strategy attendees used when matching their experience with their neighbors’. Was there a specific strategy that led to more correct matches? We also plan to analyze the content of the free descriptions for the types of words they used (abstract, source, and evaluative terms). But for now, in this write-up, we focused on whether participants matched descriptions to drinks correctly.
Of the 98 session attendees, 77 agreed to their data being used and completed the entire survey[5], which asked them to report what level of education they had as a coffee professional. Of these, 22 self-reported to have passed the Q, either for Arabica or Robusta, 32 self-reported that they had partaken of formalized coffee education, and the remaining 22 reported not to have taken any coffee-related courses.
Results
First, we analyzed the difference between wines and coffees: Based on the amount of correct matches, are coffee experts better at communicating about coffee than about wine? Next, we looked at whether differences in the background of the participants mattered: Were participants with some coffee education, or who identified as being a Q Grader, better at matching the coffees than those who reported not to have any education in coffee?
The number of correct matches for coffees was compared with the number of correct matches for wines, by means of a rank test. This test calculates the difference between the number of correct matches for wines and matches for coffee for each participant, and then checks to see if this difference is significantly higher than 0, indicating more correct matches for coffee than wine. This was indeed the case, Z = 3.08, p = .002.
Next, we looked at whether there was a difference between the groups (no education, general coffee education, or CQI Q grader) on the amount of correctly matched wines. The chi-squared test of independence, χ2(6) = 4.95, p = .551 showed no difference between the groups.
Crucially, in contrast with our hypothesis, it did not matter what educational background the participants had when matching coffees to descriptions, either: the three groups showed comparable numbers of correct matches, as was confirmed by a chi-squared test of independence, χ2(4) = 4.95, p = .293.
Conclusion
Overall, coffee professionals were much better when matching coffee to descriptions than matching wine to descriptions, indicating that coffee professionals have domain-specific flavor expertise. This is in line with previous results from the 2016 study referenced in Issue 12 of 25[6], although there are differences in the methodology. Whereas communicative success was measured here by asking participants to match coffee to a description given by another participant, the previous study looked at consistency of the descriptions as dependent measure. Although these are related measures, they are not completely comparable.
The results further suggest educational background does not influence one’s ability to talk about coffee. Even though Q graders are consistent in rating particular attributes of coffee[7], when looking at free communication as was done here, there was no difference. There are two possible reasons for this. First, the CQI Q grader training focuses on rating many different attributes consistently, which is useful in many coffee quality situations, but this skill is not transferable to giving free descriptions, where someone has to come up with the attributes (or fitting words) themselves. A more general explanation might be that the different types of currently available education do not particularly focus on coffee description.
Descriptions of the flavor of coffee become more important with the use of internet. Sending smells or flavors digitally is not possible, and it is likely this won’t be possible in the near future either. Online coffee consumers want an indication of what they can expect to taste in the coffee they’ve purchased, and language can provide this information. As shared in Issue 12’s Spotlight feature, practicing flavor description, using flavor wheels to become acquainted with a basic vocabulary, and talking about flavor as much as possible, in as many situations as possible, and with as many coffees as possible, may help to improve flavor descriptions.
Dr. ILJA CROIJMANS is a researcher at the Utrecht University. His research has focused on the effect of flavor expertise on the way people describe and memorize smells and tastes, with a particular emphasis on wine and coffee. Read his Spotlight feature in Issue 12 of 25 here.
References
[1] Based on Majid & Burenhult, 2014; Majid & Kruspe, 2018; Croijmans & Majid, 2016; Crojimans et al., 2019.
[2] The wines used were: Clarendon Hills Cabernet Sauvignon 2012; Paul Jaoulet Aine Crozes Hermitage Domain de Thalabert 2017 (100% Syrah); and Fourcas Dupre Château Fourcas 2010 (44% Merlot, 44% Cabernet Sauvignon, 10% Cabernet Franc). The ranking of the wines per participant suggested the wines were liked equally (average ranks 1.9-2.0), with no difference in rankings across wines or education groups, χ2 (2, n = 83) = 3.2, p = .520.
[3] Based on Croijmans & Majid, 2016; Pereira et al., 2017; Hayakawa et al., 2010.
[4] The coffees used were: Guatemala Huehuetenango 2000m; Granitos de Ortiz Esperanza Costa Rica Tarrazu Red/Yellow Catuai, black honey processed 1800m; Timbuyacu San Nicolas Peru, Caturra/Pache 1850m. The ranking data suggested coffees 2 (average rank 1.9) and 3 (average rank 1.8) were liked somewhat over coffee 1 (average rank 2.3), with no difference between the groups, χ2 (2, n = 79) = 1.3, p = .862.
[5] Of these, 32 participants were female. The mean age was 36 years (SD = 8.2, [21-59]). Seven participants reported to smoke every day, and 12 reported to smoke occasionally.
[6] Croijmans & Majid, 2016
[7] Pereira et al., 2017
Special Thanks to Our Issue 12 Advertisers
This issue of 25 is supported by Pacific Foods, Bellwether Coffee, BWT Water + More, Cropster, Wilbur Curtis, iFinca, Pentair, and TONE Swiss.