What is the triangular cupping test?
The triangular cupping test (or triangle test) is a sensory discrimination protocol in which three cups are presented to the evaluator: two from coffee A and one from coffee B (or the reverse), unlabelled. The goal is to identify the 'odd one out' among the three. This test eliminates knowledge bias — the evaluator cannot anticipate the result from origin knowledge and must rely exclusively on sensory perception.
The triangle test is widely used in specialty coffee for three main applications. First application: Q-grader certification and calibration. The Coffee Quality Institute includes several triangle tests in its 22 certification examinations, using closely matched coffee pairs (1-2 SCA points apart). A Q-grader candidate must correctly identify the odd cup in 8 out of 10 tests to pass this module. Second application: roastery quality control. When a roaster changes their green bean supplier or roast profile, the triangle test verifies that a perceived difference is real and not a lot-variation artefact. Third application: consumer studies. Coffee brands and distributors use the triangle test to measure the sensitivity of target consumers to a recipe or blend modification.
The probability of success by pure chance is 1/3 (33.3 %) in a three-cup triangle test — considerably higher than a binary test (50 %). As a result, statisticians require a higher number of repetitions to reach statistical significance: an evaluator who correctly identifies the odd cup in 7 out of 10 tests is only statistically significant at the 90 % confidence level, while 10 out of 12 exceeds the 99 % confidence threshold. The value of the triangle test lies in this statistical rigour: it allows differentiation of evaluators with genuine sensory acuity from those performing by chance.
A little-known finding: research published in the Journal of Sensory Studies has shown that some non-expert evaluators outperform Q-graders in triangle tests when the tested pairs correspond to their own cultural aroma categories — for example, a red wine lover may outperform a standard Q-grader in detecting a malic acid difference between two Ethiopian coffees.
Triangle test: results and statistical significance
| Correct identifications out of N tests | Chance probability | Statistical confidence level |
|---|---|---|
| 5/6 | 33.3 % | ~80 % |
| 7/9 | 33.3 % | ~90 % |
| 8/10 | 33.3 % | ~95 % |
| 9/12 | 33.3 % | ~95 % |
| 10/12 | 33.3 % | ~99 % |
| 3/3 | 33.3 % | ~70 % (insufficient for formal evaluation) |