About Triangle Testing

You'll hear us and other taste evaluators refer to a "Triangle Test" constantly. In fact almost all of the experiments we're creating for the site are based around a triangle test evaluation to determine the truth or falseness of the hypothesis. Here's the basics of a Triangle Test. (This content adapted from our book Experimental Homebrewing.)


TRIANGLE TASTING: THE GATEWAY TO VALIDITY

The best way to evaluate a beer objectively is to use what’s known as blind triangle tasting. In a nutshell, have someone else pour two samples of one of the beers you want to evaluate and one sample of another. The objective is to pick out the sample that’s different. If you can’t do that, then you know that whatever you’re testing for doesn’t really matter. If you can pick out the different sample, then you can go ahead and ask additional questions to discover the qualitative impacts the question has on your beer.

Triangle testing is a cornerstone of beer evaluation. In the world of sensory science it is considered a difference test. The taster is given no guidance about what’s different. He or she is just handed three samples and asked which two are the same.
A triangle test can help determine if a difference exists at all. Think about the major debates like: Does decoction mashing make a difference in the flavor of the beer? If it does, the tasters will detect it. If not, then the test tentatively indicates that your uncontrolled variable (decoction mashing versus infusion) makes no perceptual difference.

It can also serve as a test of panelist quality. (For example, a better, more trustworthy panelist will spot the difference.) You can use this during your panels to select judges for further tastings (“Great! Now tell us about the character differences you perceive and evaluate sample B fully”) or to weight their feedback above those who missed the difference.

Keep in mind that this technique doesn’t work for vastly different beers, like a pilsner and a stout. The beers need to be the same beer with a slight difference. Look at the sample process below. The two beers are pilsners with the only difference being the mashing method. It could as easily be two Scottish ales with concentrated kettle runnings in one and the other just plain. Maybe the same wort with two different yeast strains, and so on.

If you want to get all super science-y when you’re using a tasting panel, you may want to use weighting to determine how many panelists are likely to choose the odd beer simply by chance. There are resources available online to help you with weighting.

Cathy Haddock is a sensory specialist at Sierra Nevada Brewing Company in Chico, California. Part of her job is to conduct blind triangle tastings of the beers produced by Sierra Nevada. Sometimes it’s to evaluate new recipes being considered for production, and sometimes it’s for quality assurance, in order to make certain a batch of beer meets Sierra Nevada’s high standards and is consistent with previous batches.

Cathy has a few tips for homebrewers who want to conduct a blind triangle tasting:

Proper protocol needs to be followed in order to trust your results. Proper protocol includes following procedures in which a taster’s response is not biased or influenced due to any psychological factors or environmental conditions. Those psychological/environmental factors that can influences a taster’s response when doing a triangle test are numerous, but I will sound off on a few that I feel are most important:

  1. We will not tell the tasters anything about the samples they are tasting in a triangle test other than the brand so that they do not have any information to bias their response.
  2. We serve the samples in a frosted glass to help eliminate visual cues biasing a taster’s response.
  3. All samples are poured the same amount of beer, with careful attention as to not have one beer more foamy than the others. If that is not enough, I also ask that they not even look at the samples, just simply grab the glass, noting its three-digit code, and evaluate the sample. This way, we can have confidence the tasters are not biased by any visual cues.

We also run the triangle tests in a random balance order to help eliminate First Order effect, which is where the first sample evaluated is perceived stronger—whether negatively or positively—and therefore may be chosen as the odd sample out. This type of presentation format employs that the odd sample out is evenly tasted in the first, second, and third position in a three-sample triangle set. I also allow tasters to retaste if necessary.

Other external controls we employ to help offset bias is that other tasters in the tasting area do not verbalize, whether through speech or body language, any opinions on the samples they are tasting in triangle test.

Sample Triangle Test Procedure

Beer 1: A Pilsner made with a decoction mash
Beer 2: A Pilsner made with a standard infusion mash
Question: Does decoction mashing have a perceivable impact on the beer flavor/aroma?
Needs: 1 coin, 3 numbered slips of paper, 3 frosted/solid cups per taster labeled preferably with non-biased ordered labels (e.g. 'Red', 'Green', 'Blue', 'Circle','Square', 'Triangle', 'Paul', 'George', 'John'). Ideally, you have another person pouring the beers and keeping those notes while you observe the tasting.

PROCESS

  1. Flip a coin. Heads means beer 1, tails means beer 2.
  2. Put three labelled slips of paper in a hat. Pull the slips. The first two pulled are for the beer selected by the
    coin flip. The last is for the other beer.
  3. Pour the beer selected by the coin flip into the two cups selected by the slip pull.
  4. Pour the beer not selected into the remaining glass. You should now have 3 cups of beer with two beers in random pouring order
  5. Serve the beer to the panelists . Give no direction beyond, “Choose the two samples that are the same .” If the test isn’t about the visual quality of the beer, ask the panelists not to evaluate the beer’s appearance . If the panelists choose the correct two samples overwhelmingly, congratulations! You have a determinable difference. Now you need to determine if it’s actually from your process change or because one sample got infected, was colder, and so on.
  6. It's also useful, before revealing the differences being tested to gather the taster's feedback on the different beers and why they chose the samples they did.