One of the styles of beer that I always reach for is the Bavarian Hefeweizen. In Ontario, it’s a bit of a hit or miss product; the ones that win me over often have more of an emphasis on bananas instead of cloves.
A fellow home brewer once told me that the only way they could ever get lots of banana flavour was with a decoction mash. While an interesting observation, I’ve never had much luck with them. Fast-forward about 8 months to a class presentation from a Lallemand employee on that same subject. We were mostly learning about dry yeast production, but he also presented some interesting research regarding their various strains, including their Hefeweizen strain—Munich. In their research, Lallemand brewed a wheat beer experiment, adding glucose at a concentration of 10g/L in one of segment. Fermentations were set at 10, 15 and 20°C to compare the largest concentration of the chemical responsible for the banana flavours in beer—Isoamyl Acetate.
Data Courtesy of Lallemand
This really piqued my interest and I lined up a Hefeweizen recipe for a future brew. Since my setup makes split batching easier, I decided to brew the base beer and throw some sterilized dextrose into half. I no longer had the class presentation, so I looked around online for a few other dextrose usage rates and decided to use 75g for 11L of wort. Both beers were fermented in the same fridge at roughly 20°C for two weeks. Capitalizing on my classmates basically focusing on beer 24/7, I used our Beer Club as an avenue for analysis.
Pre-Boil Gravity: 1.043 OG: 1.053 FG: 1.013
IBU: 12 SRM: 4.3 Ferm Temp: 18°C
53.3% Crisp Wheat Malt
26.5% Weyermann Pilsner Malt
17.8% Munich Malt
2.4% Unmalted Wheat
6.8g/L Dextrose added to test
12 IBU German Hallertau Mittlefruh @ 60 minutes
2L Culture of White Labs WLP300 split across both batches
Mash Schedule and Chemistry
Target pH: 5.2
Ca: 68ppm Na: 18ppm Mg: 11ppm
Cl: 82ppm SO4: 33ppm HCO3: 119ppm
Ferulic Acid Rest @ 45°C for 15m
Saccharification Rest @ 66°C for 60m
You can see my brew day sheets attached at the bottom of this article. I’ll post up a plain text copy in the near future.
One of my professors has over 20 years of experience from working at Labatts. In that time, he was involved in a lot of sensory panels. After a quick conversation with him he suggested that the flights be randomized in order to reduce bias as much as possible. I took his advice and threw together a quick Ruby script to casually generate any number of flights with a given set of samples. One key piece of advice was to add allow the flights to be generated from four choices, in order to generate a more diverse collection of flights. For example with my test, I used the following as my selection pool for each flight: Dex, Dex, Ctl, Ctl. You can find this code up on Github, and I’ll probably a web interface for easy generation of flights for sensory testing.
14 people signed up for the panel, across three of the four semesters. Each of us have been focusing on beer styles and sensory training for 3–11 months consecutively, and one was a certified BJCP judge. I also collected information about sensitivities to track any heavy blind spots in the test.
During the experiment, I gave everyone flight sheets and explained what I was looking for. I told them is that this was a typical triangle test—a test with three randomly ordered samples, two alike and one different—and I wanted them to pick out which one was different and explain why. I did explain that the flights were completely randomized, so there was a low chance that two flights would be the same. In order to keep bias to a minimum I didn’t tell them which parts of the samples to disregard.
One flaw in my approach was directly pouring from bottles, as I was unable to homogenize them without pitchers. It’s possible that this lead some participants astray because of clarity or carbonation. In the future I’ll be sure to have my own pitchers in order to have more consistent samples for the panelists.
Of the 14 participants, 12 correctly chose which sample was different to give us a p-value of 8.21665×10-5. Many of the participants noted that there was a difference in the ester profile of the dextrose sample too. The notes from the sample have been included in the references of this article.
Several panelists also noted that while both beers tended to have a spicy character to them. I believe this was caused by my attempt at a ferulic acid rest and I plan to completely skip it in the future. Instead I plan to go with a single infusion rest at 68°C to improve the body.
Based on these results, a small amount of dextrose added to wort should provide more banana notes. The noticeable difference made for a Hefeweizen with more banana character and less clove, hitting closer to my favourite Bavarian Hefeweizens.
This was my first stab at doing an experiment and sensory panel, and I’ve already learned a lot. What techniques do you use in order to eliminate bias and while still ensuring that your panelists are analyzing the correct components?
I hope this inspires you to run a sensory panel with members of your brewing community. Check out the tools I’ve put together to generate a set of test flights and distribute them easily.