This article describes how to conduct your own mead triangle test experiments. This includes recommendations on experiment design and details on best practices. Tools such as surveys and statistical models are provided.
Mead Triangle Tests
What is a Triangle Test?
A triangle test is a method to discriminate meads and is used in sensory science. The test can be used to test for overall differences in meads, changes in processes or ingredients, or selecting qualified participants.
How to implement a Triangle Test?
During a triangle test, a taster is presented with three meads, where two of the meads are the same and one is different. The taster is instructed to identify the odd mead out and record the answer. If the taster is correct, they may be asked to fill out a questionnaire. There are two common ways to present the triangle tests to panelists.
One way, as shown by the pictures below is to put stickers on the bottom of the cups. The lone symbol will be the odd mead out. If you follow this method make sure that the cups are opaque, or at least that the stickers cannot be seen unless the cups are listed. The advantage of using this method is that it is harder for someone to bias the results if they state their preference. It also makes it easier for half the people to get one mead and half the people the other. Merely place sticker 50-50 on the cups and randomly assign the cups. The main disadvantage is that it is a pain to put the stickers on the bottom of the cups.
Another way of conducting the triangle test is to use three different colored cups. For example, you could use a blue, red, and purple cup. High quality 8 oz party cups are often available in a variety of colors and can often be found at dollar stores. You can also use different colored solo cups as used by brulosophy.com. For this method, the odd mead out is added to one color and the other mead is poured in the two remaining colors. The advantage of using this method is that it is easier to pour. The main disadvantage is that someone can bias the results if they state their preference, since everyone can see the colors and the correct answer is the same for everyone. It also means that everyone will get the same mead out, a potential problem if the triangle test is to be considered randomized.
Collecting Data in a Triangle Test
Data is collected in how many tasters identified the odd mead out. Data may also be collected on aroma and flavor differences, preferences, off-flavors, and lots of other potential qualifiers. The survey below is an example of a survey given in the Acid Additions in TANG Cream Soda Mead experiment. The survey can be downloaded here.
Using this survey, Part 1 asks participants for their name, judging status, experience level with mead and their degree of palate fatigue. Participants are asked to identify the odd mead out and write that down in part 2. Once it is written down (it doesn’t matter what they write, left, center, right etc.) they can check to see if they are correct. This can be done my getting them to lift the cups to check the stickers at the bottom (if using that method). If they were correct in identifying the odd mead out, they are asked to fill out part 3, on preference, and aroma, flavor, and overall descriptors. The check box on the right allows for a numerical score of their perception of the quality of the mead.
The sensory analysis can be reported for only those who go the triangle test correct, you may also report the people who got it wrong, as well as an aggregate measure. A very interesting data point is for those who got the test correct, which mead is preferred.
The p-values were calculated from a one-sided tests using the bi-nominal distribution with a null of 1/3, the same as used by brulosophy.com. The p-value calculator comes from onbrewing.com and can be found using the web link. It is as simple as plug-in the number of participants and the number of correct answers. The number of expected correct answers, is a third of the total number of participants. In a triangle test, the probability of a correct answer by chance is 1/3. The probability of choosing an incorrect answer by chance is 2/3.
In addition to the information on how many got the odd mead out descriptors can be recorded and tabulated. List all descriptors used. Try to show how may times a descriptor was used. You can use a table or column for aroma, flavor, overall.
The evaluation of the meads using the check boxes, for perception of overall quality, aroma, off-flavors, etc provide readers of the study a reference point on the quality of the mead being evaluated with an ordinal ranking. Note that inexperienced panelists often have a tendency to score around the mean.
Check boxes for panelists level of experience with meads and palate fatigue are udeful for the understanding which factors make panelsist more likely to be correct. For example, 1 may be no experience with meads, and 5 may be very experienced mead tasters. Averages, standard deviation, median value can then be tabulated. This helps provide a comparison of results across studies.
Optimal experiment design
There are a number of things to consider in designing your experiment to ensure the results are not biased, and get ample data. The following should be considered. If not, it is important to acknowledge these points as conditions on your results when they are presented.
- Only one variable should be changes for each triangle test experiment.
- Try to have the participants blind to the parameter changed. Do not tell them what you are testing for. If people knew the parameter of interest, you should report this with your results.
- All triangle tests should be conducted at the same time, under the same conditions. Parameters such as serving temperature, cup sizes, disturbance of the bottles prior to pouring, and pouring conditions should be held constant. A brew club meeting or a judging event works well for this purpose.
- All three samples should be presented to the participants immediately after pouring or as similar a time from pouring as possible.
- The odd meads out should be randomized by participant. For example pouring from two kegs and half people get two of one sample and the other half randomly get two of the other sample.
- If possible the order of the meads tested should be randomized. There are six possible order combinations for samples A and B: AAB, ABA, BAA, BBA, BAB, and ABB.
- Make sure the participants are instructed to take their time, and that they will be asked about descriptors if they correctly identify the odd mead out. You should be as consistent as possible in your instruction. Report your instruction with your results.
- Supervise the tasting. It may be prudent to ask that a moderator is called over when the participant is ready to make a guess if you are using stickers.
- You should have at least 15 participants for the p-values from the triangle tests to have any validity. Plan to have more than 30 participants to make your study as strong as possible. If there are less participants, the weight of interest will be placed on the qualitative descriptors.
- Don’t tell participants if they got it correct, or the parameters of the study untill all the panelists have submitted their guess.
Finally, keep in mind that these are recommendations on best practices. Even if something goes wrong, or you cannot control a parameter of the study, your results may still be of interest. Just make sure to state whether any of the above factors were different for any of the participants so that the results can be interpreted correctly.