The Triangle Test

When you compare two types of food, it's not enough to do a blind taste and pick one out of the two. The correct way to do it is to make three samples, where one sample is from one type of food and the other two samples are from another type of food.

All three samples should be presented to the tasters at once. They should come in the same kind of container. If there is more than one taster, the sequences in which the samples are displayed should be randomized. For instance, the first taster sees sample AAB from left to right, and the second taster gets samples ABA from left to right. Obviously, the tasters should not know which is which at this point.

The tasters are asked to first decide which one sample is different from the other two. If they cannot correctly pick out the sample that is different, the conclusion is that there is no discernible difference between the two kinds of food. If they correctly pick out the different sample, you can then ask them which of the two types they prefer.

This is fine if, for instance, you are just doing a simple taste test yourself to decide if a change to grandma's recipe actually makes a difference. However, if you ever assemble a large taste panel and want to be rigorous, you need to consider the null hypothesis: everybody is guessing. For instance, if around 33% of the tasters correctly identify the one sample out of three samples, you cannot eliminate the possibility that those people just got the right answer by chance. Mathematically, the triangle test is a special case of the Chi-squared test.

Even subjective tests can get hard in a hurry! Now let's move on to some physics and chemistry.