So far in this series, we have covered the basic principles of designing a sensory program and running tests. Here, we’re going to zoom in on discriminative testing—the most important type for identifying differences in samples and a powerful tool in the distillery.
First, let’s refresh our memories: What is discriminative testing, exactly?
The Society of Sensory Professionals defines discriminative tests as those “used to determine if there is a detectable difference between products.” Thus, it focuses on the perceived differences among samples by assessors, and it involves the use of statistics to determine whether or not those differences are important.
Discriminative tests are some of the most typical sensory tests performed in distilleries, and they are often used for quality assurance and control.
Here are four of the most popular discriminative tests used in the distilling industry.
Paired Comparison Testing
Paired comparison tests—which can also be called two-alternative forced choice tests (2AFC)—are some of the simplest and easiest discriminative tests to run. In a paired comparison test, assessors receive two unknown samples and are asked to determine whether there is a difference between them.
Traditionally, a majority of paired comparison tests are performed directionally, which means that the assessors are asked to concentrate their evaluation on a single known characteristic, such as sweetness or acidity. However, the test can also operate nondirectionally, in which case it is commonly called a same/different test. In nondirectional tests, assessors repeat the test multiple times, with the sample order changing a set number of times, to ensure that their responses remain consistent.
The chances of an assessor correctly guessing the answers on a paired comparison test are 50 percent, which means that it is not as statically efficient as other discriminative tests. To run a valid paired comparison test, you need a panel of at least 20 assessors. This puts it out of reach for many smaller craft distilleries; however, for larger distilleries, it is considered a useful test, especially when it comes to recipe development and quality control.
Triangle Testing
The triangle test, or triangular test, is one of the best-known and most-used discriminative tests in sensory science. It is used to evaluate process changes, to determine differences in products, and even to select assessors. Part of the reason for its popularity is that its design is relatively simple.
In a triangle test, assessors receive three unknown samples. Two of the samples are a baseline product, while the third is a product to be tested. The assessors then try to identify which of the samples is different from the others.
When you conduct a triangle test, it is important to represent all possible sample configurations. This helps to minimize bias and ensures the accuracy of results. Customarily, you need at least 18 assessors to conduct a statistically valid triangle test; but, you can conduct the test twice, reducing the required number of assessors while increasing the test’s power and reliability. However, it’s worth noting that to test more than once with the same assessors reduces the test’s power because of the increased chance of bias.
The triangle test is one of the most important for determining a difference in samples because it is statistically powerful. The chances of an assessor guessing the correct answer randomly are only 33 percent, much less than the paired comparison and duo-trio test. That means that the triangle test yields more statistically valid results, and it does so with fewer required data points than many other tests.
Duo-Trio Test
Like triangle tests, duo-trio testing is another common type used to evaluate differences in similar products. In the duo-trio test, assessors receive three samples: one reference sample and two unknown samples. The assessors first familiarize themselves with the reference sample, then they try to identify which of the unknown samples matches the reference sample or which is different.
Duo-trio tests can be run either in a constant-reference or balance-reference configuration. In the constant-reference configuration, the reference sample remains the same in all tests. This is usually done when evaluating the effects of a known process change, such as switching raw-material suppliers or altering a blending regimen. In the balance-reference configuration, the reference sample switches evenly between the two unknown samples. This is more often done when trying to determine whether there is a difference between two like samples. Regardless of the configuration being run, the unknown samples should be presented so that all patterns are equally represented.
Like the paired comparison test above, the odds of an assessor correctly identifying the samples at random are 50 percent. This means that the duo-trio test is not as statistically efficient as the triangle test. However, the duo-trio test is equal in sensitivity to the triangle test, and some consider it superior when evaluating strong-smelling or intense samples because it gives the assessors a reference.
Tetrad Testing
Finally, the fourth discriminative test that is commonly used in sensory science is known as the tetrad test. In the tetrad test, assessors receive four unknown samples. Two of the samples are filled with Product A, while the other two are filled with Product B. The assessors then try to separate the four samples into two groups, placing like samples together.
Tetrad tests can be run either as a specified or unspecified test. In a specified test, those conducting the test know the difference between products A and B. In an unspecified test, however, the differences between products A and B are unknown, which makes the test more suitable for evaluating any effects of changes in process.
Interestingly, the tetrad test is even more statistically efficient than the triangle test, and it needs only six participants to reach a confidence level of 95 percent. However, it’s worth noting that the results from tetrad tests are sometimes inaccurate if fewer than 30 assessors are involved. Furthermore, the tetrad test can cause increased sensory fatigue among assessors, and it is not considered appropriate for strong smelling or intense samples.
Know the Limitations
Discriminative testing can be a powerful tool in a distillery. It can inform you as the distiller how your processes affect your products, and it can help you develop key QA/QC standards. However, it is essential that aspiring sensory scientists understand the limitations of discriminative testing.
First, it must be noted that the numbers given above are for tests that examine the differences between products. If testing for similarity, there are different graphs involved.
Second, and most important, none of the above test are ways to measure preference. Preference testing, sometimes also known as hedonic testing, requires a different methodology—and we’ll cover that in a future article.