As part of our study of sensory science and its practical application to specialty coffee, our sensory lab tested and demonstrated the efficacy of some common sensory errors. It’s one thing to know that these errors exist, but it’s quite another to observe and experience them. It must be emphasized again that these errors are things that we are all susceptible to as human tasters. There may be differing skill levels among cuppers. Some may even know how various of these errors occur. Even the most skilled and informed cuppers are susceptible to errors like Adaptation.
For this reason, it is (highly) preferable that cuppers be allowed to focus their talents on the assessment of coffee rather than the assessment of assessment, and God forbid, course correction while still attempting to assess coffee. At the end of the day, humans cannot help but be human. By far, the best practice is to control for sensory errors at the level of cupping design long before your cuppers start grinding coffee.
Our goal for running these trials was to better understand cupper susceptibility to sensory errors and identify where they could occur in our cupping processes. This information would be used to design practical controls in our daily cupping protocols. We’ll see that given the number of possible errors and their overlaps and interactions, perfect cupping protocols and forms are impossible. And yet, we’ll also see that the common sensory errors are addressable. We’ll see that doing so sets one’s entire cupping panel up for success, and the cost of accomplishment is no more than an open mind and a bit of creativity.
Much hay has been made in the last few years about sensory science in specialty coffee. Hopefully, people will not end up with their carts before their horses. Complexities and inefficiencies in cupping form design are less detrimental when supported by robust protocol. Everyone can benefit from giving more bandwidth to developing practical and easily applicable¹ controls for these common sensory errors.
Individuals may become better cuppers by tasting various solutions and infusions. Or they may not. A panel will always become better cuppers when their susceptibility to error is reduced or controlled.
A decrease or change in sensitivity to a given stimulus as a result of continued exposure to that stimulus or a similar one. For example, sugar can dampen the subsequent perception of sweetness (something moderately sweet may seem weakly sweet when tasted after something very sweet).
Test: A series of 10 dextrose solutions were presented to 9 trained coffee tasters (T1 – T9 in the table below), who were asked to rate the sweetness of each solution from 1 (Least Sweet) to 3 (Most Sweet).
Hypothesis: We expected that the final B solution, labeled BH2 (strength B, Hypothesis test solution 2), would be rated less sweet than the rest of the B solutions. We also thought the first BH1 solution might be rated as less sweet than the other B solutions. These two hypotheses assumed that the Adaptation Error would be localized such that a familiar level of sweetness following a novel high level of sweetness would be underrated. We were uncertain if an Adaptation effect would occur as early as the BH1 placement.
Another possibility was that an Adaptation Error would be more cumulative than localized, and the solutions in the latter half of the series would tend to be rated less sweet than their counterparts in the first half.
We included a C solution early in the series to help avoid a Habituation Error² and to give a sense of the cumulative impact of any Adaptation effect. If the first C sample were rated sweeter than the last two, this could suggest a more cumulative impact. If the C samples were rated with relative parity and the BH solutions were rated relatively less sweet than the other B solutions, this would suggest a more localized Adaptation effect.
Formally stated, our hypotheses were:
H0: B = BH
HA: B > BH
Results: The two BH solutions were equally rated as the least sweet on the table, with mean scores of 1.11. The three C solutions were rated at relative parity with mean scores of 2.78, 2.89, and 2.78. The mean value for B (excluding the BH solutions) was 1.56, while the mean value for the BH solutions was 1.11.
In addition to the likelihood of having observed an Adaptation Error, we can point out that the tasters were either biased or tentative against answering 2, though not enough to rule out an Adaptation effect. Tasters were much more confident answering 3 for the C solutions, where the mean rating was only 0.22 below the actual strength vs nearly half a point for the B solutions. C Solutions received only 5 ratings of 2 out of 27 total ratings. B solutions (excluding BH solutions) received only one rating of 3 and 21 ratings of 1 out of 45. The first B solution was rated the strongest B solution overall. The decline in B solution ratings could come from a cumulative Adaptation effect or a sort of Expectation Error³, whereby tasters were looking for A solutions despite none being presented in the test and having a reference provided.
Given the mean and deviation of the scores for the “non-hypothesis” B solutions (1, 2, 5, 6, 7) as compared to those for our BH solutions (4, 9), it is unlikely that our BH values could have occurred by chance. Using what is known as a “One Sample T Test,” we can estimate the likelihood of a sample with the BH values being pulled from a population with the B values. In this case, the likelihood is about 0.01%.
In this case, we reject the null hypothesis in favor of the alternative:
H0: B = BH
HA: B > BH
Adaptation is very difficult to control for in a high-volume sample cupping situation. Time and deadlines frequently push assessors toward Adaptation. Breaks must be built into the assessment protocol, and “pushing through” should be avoided. We want to reduce the amount of brewed coffee that a person uses to perform their assessment. Assessors must learn to do their assessments as efficiently as possible, requiring focus on the task at hand and an understanding of the protocol in use. This can be accomplished with training and test design.
Tables should be arranged with coffees from different sources or differing expected grades as much as possible. Adaptation can combine potently with Habituation and we want to interrupt those tendencies. Even in a low sample volume setting, this can be accomplished by keeping a green retention sample library. Rather than roasting through or tossing out old samples, these can be kept to help build out tables and blind the coffees that you’re actively interested in assessing. Current lots can, of course, be used in this way, too.
Adaptation counter-indicates test designs in which cuppers are presented with multiple cups for each sample set, particularly when additional demands are made to scour those cups for inconsistencies, poorly defined concepts of cleanliness, or other vague metrics. Multi-cup designs, for example, the 5-cup presentation⁴ used by the SCA, give the impression of thoroughness while undermining an assessor’s capacity for thorough analysis by dragging them into Adaptation, Habituation, and a surprising number of other errors. We are wise to be wary of protocols that are new just to be new, but we should be equally careful of the lure of “the way it’s always been done.”
While taste Adaptation will occur across origins and processing types (as the basic adaptive compounds are the same), we have observed that decoupling it from aromatic Adaptation seems to help. We encourage cuppers to vary the order in which they assess a table of coffees such that the 2nd and 3rd passes are not performed in the same order as the 1st pass. I strongly encourage cupping administrators to seek ways to present samples more than once and in different orders or groupings. Our system utilizes blind replication of samples in random placements within and across flights. The work we present in the Coffee Rose is underpinned and preceded by years of study, demonstrative trials like that presented here, and subsequent process iterations. This larger methodology is the root and stem upon which the Rose is very much a flower.
¹There is some evidence that the speed and degree of taste adaptation are reduced by mouth and tongue movement during tasting. In addition to slurping, it is also valid to “chew” and “swish” your coffee while cupping.
²Habituation Error: Tendency to continue to give the same response when a series of slowly increasing or decreasing stimuli are presented. Example: The same response of slowly increasing or decreasing stimuli is presented when cupping a series of 84-85 tables and missing 83s or 86s.
³Expectation Error: Information given with the sample may trigger preconceived ideas. Example: Naturally processed coffees are fruity in flavor.
⁴Five-Cup Presentation:
Want to be notified of the next blog release? Sign up for our newsletter below.