Sensory Panels: How Much Should You Tell Them?

sensory science

I had a very interesting conversation with one of my sensory panellists. She has been on my panel since I started my career as a sensory panel leader and serves on a number of sensory panels. A career sensory panellist, if you will. She told me that she could tell that the differences between the products that she had to taste was not that large and she was afraid of finding differences that didn’t exist.

Before I started the project, I toyed with the idea of telling my panellists what I expected in terms of the differences between the products and have decided not to. The rule of thumb in sensory evaluation is to share as little information as possible in order to avoid bias. But what if a little bit of information can break false assumptions about the products?

I have investigated the effect of divulging product information to sensory panellists before a tasting as part of my PhD research. I was working on rapid sensory profiling of high alcohol beverages and in one part of my study I tested whether prior information about the alcohol content of a product set could influence the way that the panellists evaluated the samples. While my sample was quite small and the results were only indicative, I found that the information that the panel received about the products did influence their evaluation. For example, when they thought the product had a high alcohol content, they put more emphasis on terms that relates to alcohol content e.g. alcohol burn. (1)

Another example of assumptions that panellists make when evaluating a product is their responses in a difference test. The nature of the test is to test for differences, so the assumption that many panellists make is that the difference must be perceivable. In my experience, in a difference from control test, new panellists often show a left skew in their data because they rate differences to be larger than they actually are. That’s why I always insert a blind control. And don’t get me started on triangle tests and panellists insisting that we got the sample coding wrong because they could “definitely taste the difference!”

So where does that leave us? I believe that it is okay to share information if you run a risk of bias due to incorrect assumptions. While I don’t think it is possible to pre-empt every single possible assumption that your panellists may make, it’s a good idea to put yourself in your panellists’ shoes when you plan your project. The information that you give your panellists can influence your results just as much as the instruction on your ballot.