by Louis K. Bonham (Brewing Techniques - Vol. 7, No.1)
Last year’s unprecedented HBD Palexperiment resulted in a large volume of data that will undoubtedly contribute significantly to the existing body of brewing knowledge. This article is the first of two to discuss the lab results, starting with the IBU and contamination testing.
A common problem for amateur brewing scientists is the limited sample size of their experiments. Unlike a professional scientist or a commercial brewer, amateur experimenters often lack the time or resources needed to replicate their experimental batches more than once or twice. Given the myriad variables presented by a process as complex as brewing, many such experiments (including some of mine) therefore commonly lack the volume of data needed to support any definitive conclusions.
The HBD Palexperiment, described in an earlier article (1), provided a chance to avoid this common limitation. By having more than 40 experienced amateur brewers all carefully brew the same recipe, this venture collected more data then any but the most dedicated single brewer could collect in a year of brewing. Moreover, because the experimental batches were brewed at about the same time, using precisely the same yeast, malt, and hops, the HBD Palexperiment was able to minimize many of the variables that would have faced a solitary experimenter doing sequential test brews. The result of the HBD Palexperiment is an immense trove of raw data on pico-scale brewing and techniques.
In the next two installments of The Experimental Brewer, I will describe the methods that I and a group of intrepid assistants used to lab test most of the beers from the HBD Palexperiment, as well as the results of those tests and some of the conclusions I draw from them. Although space constraints make it impossible to publish all the raw data here, BT will make it available on its website so that readers can not only evaluate the validity of the data, but perhaps also can use this information as a basis for developing new theories about various aspects of small-scale brewing.
General Test Methods
Of the more than 40 participants in the HBD Palexperiment, 35 submitted samples for analysis. Sample bottles were first marked with control numbers, with one bottle of each beer reserved for carbon dioxide testing. The crowns and necks of a test beer bottle were first wiped down with alcohol. Immediately after opening, samples were aseptically transferred to four sterile 50-mL centrifuge tubes, which were premarked with each beer’s control number. The tubes were then sealed and sent to various testing stations we had set up around my house: one for IBU testing, one for contamination testing (and afterwards, pH and ethanol content testing), one for specific gravity testing, and one tube kept in reserve.
Methodology: We assayed the samples for IBUs using the American Society of Brewing Chemists (ASBC) spectrophotometric method (2). Using a pipette pump fitted with a 10-mL volumetric pipette, the tester introduced a minute amount of octyl alcohol into the tip of the pipette, and then used the pipette to transfer 10.0 mL of chilled carbonated beer to each of two 50-mL centrifuge tubes. Using a different pipette pump and a 2-mL graduated volumetric pipette, the tester added 1 mL of 3N hydrochloric acid to each tube and then used an Eppendorf repeater pipettor to add 20 mL of spectrophotometric grade isooctane (2,2,4-trimethylpentane, with A275<0.001 in a 1-cm cuvette) to each tube. The tester then sealed the tubes and vigorously shook them by hand for 10 minutes. (The official ASBC methodology calls for the use of a wrist-action lab shaker for 15 minutes, but I have found that hand-shaking for 8–10 minutes works as well. Of course, this much hand-shaking can be rather tiring, especially if you’re testing 35 beers in one session!) After shaking, we centrifuged the tubes for three minutes to separate the lighter isooctane phase from the beer.
Using a Spectronic 21 DUV spectrophotometer (Spectronic Instruments, Rochester, New York), we then measured the absorbance of this isooctane phase, in which the bittering compounds were now dissolved, at a wavelength of 275 nm. The spectrophotometer was set to read zero absorbance using a quartz cuvette and an isooctane/octyl alcohol solution, with the machine reset in this fashion after every few tests. Before each test, the inside of the cuvette was rinsed with a few milliliters of the isooctane solution being assayed. If the absorbance measurements from the two samples varied by 10% or more, we disregarded the results and repeated the entire test; otherwise, the results were averaged and multiplied by 50 to yield the IBU value of each beer.
Under the protocol of the HBD Palexperiment, brewers were to boil their wort down to below 5 gallons, and then dilute it in the kettle to yield a post-boil volume of exactly 5 gallons. While most of the brewers followed these directions, some of them did not, and they reported final volumes other than 5 gallons. This would, of course, skew the results; a wort with a smaller final volume would have a higher IBU level for the same amount of hops, whereas a larger final volume would have a lower IBU level than if it were concentrated to 5 gallons. I accordingly compensated for these variances by calculating an “adjusted” IBU level for each beer using the formula:
(measured IBUs X final volume) ÷ 5 = adjusted IBUs
Results: After we discarded two anomalous results,* the average adjusted IBU level of 33 beers was 62.1, with the distribution of results producing a bell curve with a peak at about this point (see Figure 1, left). Although more than 80% of the results were within 15% of the average value (and almost half were within 5% of this figure), there still was a swing of almost 25 IBUs between the lowest and highest reliable results. These results aptly illustrate the difficulty in calculating IBUs with a simple formula; such formulas can provide only a rough estimate of IBU levels (3). In this experiment, the hop amounts, alpha-acid content, and boil times were identical, and the wort gravities were comparable — yet the adjusted IBU levels still varied by as much as ±20% from the average value. Thus, factors not usually considered in any conventional IBU formula — water chemistry, boil strength, kettle shape, etc. — clearly play a significant role in hop utilization rates. These results also underscore why you simply must assay your beer if you really want to know your IBU levels.
*Two of the beers measured more than 90 IBUs (yes, we repeated the tests a number of times to confirm these findings). Because these two results differed so dramatically from the other beers, and because the two brewers in question did not report the specifics of their brew cycle (including the final volumes and the timing of the hop additions), I believe it is safe to disregard these values as artifacts.
Although our results demonstrate the limitations of any IBU calculation, most amateur brewers (including me) will still use one in designing a recipe, and thus it is a fair question to ask what the various IBU formulas predicted. One of the first IBU formulas for amateur brewers that I know of was published by Byron Burch (4). This formula predicted an IBU level of 45.6 for the HBD Palexperiment beer. Jackie Rager’s seminal work, which has formed the basis for many subsequent efforts, estimated an IBU level of 53.5 (5). Greg Noonan’s well-regarded methodology predicts 56.9 (6). Mark Garetz’s popular formulation, which he based on Rager’s work, gave 41 IBUs (7).* Glenn Tinseth’s approach, which makes very different assumptions than the other commonly-used formulas, generated an IBU level of 48 (8). Ray Daniels’ approach, which uses the basic Rager formula but with separate and more comprehensive utilization charts for whole and pellet hops, predicted that the IBU level would be 67.9 (3). But the top place in the HBD Palexperiment IBU sweepstakes goes to a fiendishly simple little device created by Randy Mosher known as “Dr. Bob Technical’s Incredible Hop Go Round” (Alephenalia Publications, Seattle, Washington). This slide-rule device estimated the IBUs for the beer in question at 61.† (The Palexperiment hopping schedule is shown on page 21.)
Methodology: Using sterile 1-mL transfer pipettes (used once and then discarded), we aseptically spread about 0.5 mL of each beer on a plate of Lee’s Multi Differential Agar (LMDA) (9). These plates were then sealed with Parafilm and incubated aerobically at approximately 82 °F (28 °C). I periodically inspected the plates for the next week. At the end of the week, I catalase-tested colonies from most of the contaminated plates. I also Gram-stained slides of representative colonies and examined them under the microscope at 1000X. The results of the testing are also posted on the chart on the website (names withheld to protect identities, of course). Where I am not fairly certain as to identity of the contaminating organisms (usually because I encountered a mixture of organisms and I have relatively limited microscopy experience), I have indicated my best guess with a question mark.
Results: When deciding what tests to run, I had originally planned to use the industry standard method of testing finished beer for contamination, which involves filtering 100 mL of beer through a sterile 0.45-micron filter membrane, then aseptically transferring this membrane to a plate of LMDA or Universal Beer Agar (UBA). However, both Katie Kunz and Paul Farnsworth strongly dissuaded me from doing so, in part because in their experience homebrew typically has so many bacteria present that membrane filtration is unnecessary. Our results dramatically confirmed their advice. While about a quarter of the plates yielded no colonies at all — even after one week — most of the beers showed clear and often dramatic evidence of bacterial contamination. By far, the most common contaminant was Pediococcus damnosus (found in about half of the beers), but we also encountered Lactobacillus, Acetobacter, enteric bacteria, and Bacillus.
†I must, however, acknowledge that our data is based on a critical assumption — that the alpha-acid levels reported by the hop merchant are accurate. If in fact the hops used had a higher alpha-acid level, then one of the other formulas probably would have yielded the most “accurate” prediction.*In a private communication that took place before I released the preliminary results of our tests, Mark indicated that if he assumed advanced home brewing techniques, he would have increased the hop utilization rates and estimated the IBU level at 65.
I have subjectively characterized the levels of infection as “clean” (9 samples), “mild” (4 samples), “moderate” (6 samples), or “severe” (15 samples). As I use these terms in this article, “clean” means no growth whatsoever, a “mild” infection means a few (<25) discrete colonies, a “moderate” infection refers to about 25 to 100 discrete colonies or a continuous mass of nonspreading colonies that covered less than about 20% of the plate, and a “severe” infection was more than this. Where the infection was of a “spreader” type colony (for example, enterics, Bacillus), I gauged the degree of infection based on how long it took the spreader colonies to cover the plate.
I must emphasize that these terms are purely relative for this experiment, and that if we were using objective commercial standards the results would be viewed much more harshly. Generally, colony counts of 10 or more are cause for serious concern for the professional brewer, especially if the contaminating organisms are beer spoilers such as Pediococcus or Lactobacillus. Thus, by commercial standards, what I have termed a “mild” Pediococcus infection could in fact be very serious, and a “moderate” level of 100 colonies would cause most professional brewers to reach for the Maalox.
OK, so quite a few of the beers had levels of bacteria that would be unacceptable in a commercial setting. Does it matter? Certainly, some of the beers submitted tasted like they had “gone off,” but most tasted like slightly-better-than-average all-grain homebrew. Some of the beers — even though the LMDA test indicated that they were swimming with Pediococcus — were actually very good (indeed, the beer judged to be the best by the sensory evaluation panel had a moderate Pediococcus infection), while some of the “clean” beers tasted only average to me. And, in fairness to the brewers, all of the beers were shipped during the heat of the summer, and this stress would certainly have exacerbated any contamination that was present.
Nevertheless, the basic findings of our LMDA tests are unavoidable — when compared to commercial beer, most of the test beers had extreme levels of bacterial contamination. Whereas they might not have produced immediate and drastic effects on the beer’s flavor, such infection levels are probably more common than most amateur brewers want to admit. (For those of you who are convinced that your beer meets commercial levels of cleanliness, I urge you to aseptically plate a few drops of it onto some LMDA . The results may surprise you.)
The high degree of contamination we encountered also suggests a potential problem with the pitching rates used. In order to minimize variables, each participant was given a Wyeast XL pack (Wyeast Laboratories, Hood River, Oregon) from the same lot and package date and instructed to pitch directly from that package. Despite the manufacturer’s claims that these packages do not require a starter for a 5-gallon batch, all of the participants reported lag times of 18 hours or more, and these longer-than-usual lag times could have provided bacteria and other contaminants a window of opportunity to establish a significant presence. Because pitching large quantities of healthy yeast minimizes the lag phase and is one of the best weapons brewers have against bacterial contamination, our findings suggest that it would probably be a better practice for brewers to use a starter with the Wyeast XL packs. (See George De Piro’s article on page 48 for information on making a starter.)
The Rest of the Story
In the next issue, I’ll finish describing the lab tests and results from the HBD Palexperiment, as well as the results of some related experiments on determining alcohol and carbon dioxide content. If you have any comments or theories on the results of the HBD Palexperiment, please feel free to contact me — there might be an Experimental Brewer guest column in your future.
All contents copyright 2021 by MoreFlavor Inc. All rights reserved. No part of this document or the related files may be reproduced or transmitted in any form, by any means (electronic, photocopying, recording, or otherwise) without the prior written permission of the publisher.