“Which one is better?” is often the first question at a tasting. It may not be the question your setup can answer. Before pouring, decide whether you want to describe a difference, detect a difference, or discover a preference. Each needs a different kind of response.
Choose The Question First
For description, ask what people notice and how strongly, using agreed terms. For discrimination, ask whether they can distinguish the samples under the stated conditions. For preference, ask which they like for a stated use. A sample can be distinguishable without being preferred. A preference vote does not tell you which aroma caused the choice.
For a personal serving decision, a coded pair and a short written comparison may be enough. For a production or certification decision, plan the sample preparation, assessors, repetitions and analysis before collecting answers. Keep the sample key separate, use unrelated codes, balance serving order, and collect independent responses before discussion.
Understand The Triangle
In a triangle test, a taster receives three coded samples: two from one product and one from another. The task is to select the odd sample, not the favorite. There are six arrangements: AAB, ABA, BAA, BBA, BAB and ABB. Allocate presentation orders across assessors so a position does not become the answer. This structure and its one-in-three chance probability are described by the Society of Sensory Professionals, Triangle Test, “Test Principle” and “Data Analysis”.
The diagram is the preparer's view. The taster sees only three unrelated codes; even the two matching samples receive different codes. Identical glassware, equal pours and consistent preparation help prevent accidental clues. If the question is about aroma rather than appearance, account for visible color differences as part of the design.
Whiskey's lingering sensations deserve attention. ISO's triangle-test scope specifically notes limitations for products with strong carryover or lingering flavors. A triangle also does not identify the size, direction or sensory cause of a difference. ISO 4120:2021, official abstract and scope.
A Correct Answer Is A Beginning
One correct selection can happen by chance. Repeated answers from one person are not the same as independent answers from many people, and learning or fatigue can change performance. Report who participated, how often, under what conditions, and what was counted. A formal study needs a planned analysis appropriate to those dependencies.
Consider an invented preference result: seven of twelve participants choose A, five choose B. “Seven of our twelve participants preferred A in this session” reports the result. “A is superior” adds a conclusion the count has not established. Likewise, failure to detect a difference is not automatically evidence of equivalence; a similarity study must be planned for that question.
Separate The Observation From The Cause
Consider this note: “The new recipe is harsher because the bourbon changed.” Which part records an observation, and which part proposes an explanation? Start with “I perceived more harshness in this comparison,” then record what that word meant to you and how the samples were prepared. List another plausible explanation before deciding what to check.
Perhaps the bourbon changed, but so did the syrup formulation or serving temperature. Comparing the two finished recipes can still help you choose between them; it cannot isolate the bourbon's effect. Design a check that addresses the proposed cause and say what result would make you reconsider it. You can complete this exercise using written records.
Kister's industrial troubleshooting discussion offers the guiding habit: even an obvious fault remains a candidate explanation until the evidence connects it to the problem. This exercise adapts that reasoning to a tasting record.
Kister, printed pp. 44–49 / PDF 57–62 — Literature Note
Choose A Check That Matches The Decision
Use repeat to investigate consistency, change to explore the effect of a planned preparation, and transfer to see whether the learner can use the method on unfamiliar material. They are parts of the Academy learning sequence, not interchangeable statistical tests.
A descriptive flight containing a concealed duplicate is not automatically a triangle test. A triangle requires the specified odd-sample task and an appropriate design and analysis. Likewise, twenty tasters examining one bottle do not provide twenty independent bottles or production batches. State which people, products and conditions the result can represent.
Repair A Small Experiment
A host serves A first to everyone, announces that it is expensive, then asks the group to discuss which is smoother before voting. What would you change?
Keep exploring: Method and printable record · Course overview · Previous lesson · Next lesson: Enjoyment in Context