Distillery Practice—Blending

Follow along: IG @birectifier

A very unique mimeographed manuscript crossed my desk that was penned by a Seagram employee, likely in the late 1930’s or early 1940’s, but before the famous 1943 Seagram text called “Fundamentals of Distillery Practice” written by Herman F. Willkie and Joseph A. Prochaska. I’m not aware of any informally published manuscript like this having been found by the distilling field. It came to me from an individual who wanted to contribute to the education of the new generation of distillers.

The manuscript is a bit scattershot and not exactly in any kind of order and contains a ton of information from the era not seen in the 1943 text. This was likely intended as an internal training document. These were the kinds of jobs where once you were in, you were in for life so they took education seriously.

In the manuscript is a section on blending which is a topic barely covered in any of the known distilling literature. My take is that this represents some of Seagram’s earliest explorations and everything in here was quickly made obsolete as knowledge accumulated. It was known that in the early 1940’s Seagram was exploring sensory evaluation techniques developed by the U.S. Army’s Quartermaster Corps. It is not clear if the Distillery Practice manuscript is before or after their influence.

What is presented raises a lot of questions because the subject is blends that are considered incredibly lame by today’s standards. We see data on preference but learn nothing about the tasters and what acquired tastes they have. Would they prefer straight Bourbon over a blend? I could scrutinize a lot of their ideas here, but what is cool is that we see the construction of small sketches to explore the position of a whiskey blending stock among other components. We also learn a little bit more about historically important blends like Calvert and Seagram 7 which may have had sherry wine or even malt whiskey added. We also see some other things such as a lack of correlation between age and preference which might surprise many.

No chemical data is correlated to any of these basic studies, but we know that is coming next as Seagram conducted many deep investigations that run parallel to those of Rafael Arroyo.

Something else absent, and possibly not part of the whiskey producer’s repertoire, are any mention of the German tests that ranked blending stocks such as heavy rums by persistence tests, evaporation tests or even tests of surface tension. They mention no use of any tools like the birectifier which was used by Karl Micko, well before Arroyo, at the beginning of the 20th century. Much of the whiskey work led by Seagram starts in roughly the same years as Arroyo and there is even an acknowledgement at the end of Studies on Rum to C.S. Boruff, Technical Director for Hiram Walker who once worked alongside Herman Willkie.

Arroyo was interested in what chemical features of a rum could lead to earlier maturation, or even support blending such as the presence of rum oil. Whiskey seemed to go the direction of fusel oil separation at distillation while Arroyo was exploring novel yeasts or fermentation technique and trying to reduce fusel oil formation at the point of fermentation so as not to compromise other high value congeners.

We encounter a tale of the sales department frustrating the blenders, but we are not made aware of any economic necessity at the time for blending. Were spirits like straight bourbon or three chambered rye too expensive for the mass market to drink without blending in substantial neutral spirits or was it a marketing driven ploy to stretch spirits economically to make more money? In this era, where do the boundaries lie between economics and style? Should we count our blessings that the average drinker today can enjoy affordable straight Bourbon whiskey that is aged longer than anything presented here?

A few acronyms tied to the various whiskies likely representing geography for plant names, are used and I’m hopeful that a reader may be able to help us decode them. [We quickly got some clarity where the abbreviations are codes for plant name/location, mash bill, and then yeast strain.]

---

Outline for Some Aspects of Blending
by
Richard R. Slater

[Keep in mind this is just a manuscript and not a completed document. Some of the information here was likely forwarded to the main authors for consideration.]

[There are a few acronyms here for straight whiskeys that may compare somewhat to unique rum marks. We see KDB (30% small grains), KBT, KCB (35% small grains), LBB,LBQ; all Bourbons. I’m hoping some reader will swoop in and tell us the likely names behind the acronyms. The 80/20 mixture we will soon learn is a blend of 80% straight bourbon and 20% three chambered still rye.

[Tyler Derheim lets us know that KBT is Kentucky plant, 36% rye, 60% corn Bourbon mash bill and yeast #110.]

KBB appears later and is 40% small grains.

How do we interpret the claims that for LBB, there was no difference in preference as it aged? Is that because the Bourbon is likely very good or possibly very bad?]

[5 years old appears to be the oldest spirit experimented on.]

Some Aspects of Blending

Definitions of Blending:

(A) Governmental: The bureau of Internal Revenue defines blending by saying “the blending (mixing together) of two or more spirits…. by a rectifier constitutes rectification, and, except as herein-after provided, necessitates payment of the rectifying tax of $0.30 per proof gallon”. From the governmental point of view, blending is a simple process defined only in so far as is necessary to facilitate the collection of certain taxes.

(B) Operational: Operationally, blending in the rectifying plant is characterized by the withdrawal from the warehouse of certain designated types of whiskies and spirits. Simultaneously fresh spirits are withdrawn from the Cistern Room. Upon delivery of these materials to the Rectifying House, the blender calculates the quantity of each which must be released in order to fill the requirements of a certain formula. It is normally the practice to discharge the different items successively beginning with the so-called heaviest whiskies bearing the greatest age and working down the range to the lightest spirit with the least age. From the receiving tank these materials are pumped to a processing tank, and the cistern spirits are added in the necessary quantity. Reduction in proof, and addition of coloring and flavoring material ensue. For all practical purposes, blending, in so far as the plant is concerned, amounts only to this.

(C) Laboratory or Experimental: From a laboratory viewpoint the development of a blend is a complicated and lengthy process. For many years blending was considered and was, for all intents and purposes, an art. It has been the endeavor of the Quality Improvement Laboratory to establish criteria whereby the artisan shall be replaced by the scientist. The advantages of such an alteration are obvious. Once we are in a position to establish critical tables for blending we shall have advanced immeasurably beyond the old methods.

It is our hope to bring forth in this discussion evidence that we are progressing scientifically and to show some of the difficulties that are faced in our endeavors.

Whiskey—Age and Preference:

The first questions which arise when a person begins formulating a blend are: How old must the whiskey be, to be at its best? Is the maturing of whiskey or the relationship of age to preference—a linear, exponential, logarithmic, or some other mathematical function? In an effort to determine the relationship between aging time and preference, we have collected certain interesting data, some of which are tabulated below:

**These particular dates were obtained in an effort to study more closely the maturing process in the early stages.

It will be noticed in Bourbon A the the 34. month old sample is significantly superior to its older brother and that the youngest of the three is just as good preferentially as the oldest of the three.

For Bourbon B it is noteworthy that there is an improvement trend until we reach the oldest sample where digression is noticed.

Bourbon C demonstrates no appreciable difference for the very significant age differences.

We should not jump immediately to the conclusion that aging does not increase preferential values, since for Bourbon D a rather definite improvement is noted with increasing age (with the exception of the three samples between the ages of 12 and 22 months).

Here again, we must not draw erroneous conclusions for were we to base future work on the results of the investigation of the young Bourbon D’s, we would never suspect what is apparent for Bourbon E; namely, that a remarkable linear increase in preference value is found with an increase in time in wood.

What have we accomplished by accumulating such data as above? If nothing else, we have certainly established the necessity for experimental evidence directing the choice of any given whiskey to be used in a blend. How simple it would be to handicap oneself by adhering to the usual obvious beliefs regarding age and preference. We may, at the present time, in the light of these and other data, conclude only one thing: that controlling such variables as we are able to control, age often seems to be less important that certain other characteristics. Apparently, in many cases, the difference in day’s production, barrel position in the warehouse, temperature of the warehouse, or some other variable or group of variables is more important that an additional 3 or 5 or even 12 months of age. House else can we explain the fact that when we have analyzed samples of certain rye withdrawn at the 36, 48, and 60 month age periods for preferential value, there is not statistical differences from one sample to the next? It appears that some factor has displaced the preference ratings of the two youngest ages to that of the oldest, or else some influence has retarded the process of the two oldest samples to give them the same preference value as the youngest. Our next objective then is to determine what these factors are so that they may be controlled.

Optimum Whiskey Concentrations:

Logically, once having chosen a given number of satisfactory whiskies, the next step is to determine the concentrations in which these whiskies should be used with one another. In an effort to establish indicative data, the following experimental work was undertaken. A quantity of 5 year old Bourbon and Charge Still Rye was mixed in proportions of 80% Bourbon and 20% Rye. This mixture was found to be the best combination for these two whiskies when a series ranging from 100% Bourbon to 100% Rye by 5% increments was studied. With this optimum mixture was incorporated a continuous still rye in the following proportions:

[Charge Still Rye would be three chamber still rye.]

When the percent preference is plotted against the percent Continuous Still Rye, the peak of the curve is at 5% LQI. Those samples containing 7.5% and more of these rye are definitely inferior. It, therefore, is indicated that the best combination of LQI with the 80-20 mixture is that which includes 5% of the continuous still rye and 95% of the mixture. It does not follow that the 5% continuous still rye is the optimum quantity to be used with any other combination of the Bourbon and Rye. It is possible by the pursuit of this method to establish definite concentrations at which to use two or more whiskies. Unfortunately, however, the introduction of neutral spirits may alter to some extent the results so obtained.

Proof and Preference:

It is so simple to look upon a relatively small change in proof as being only in the nature of a slight dilution or concentration. Few people actually consider a change of as little as 0.8° of a proof as being of consequence in the rating of a blend. After several months’ work developing a formula for Carstairs “White Seal” at 90° proof, a formula was ultimately established which gave this blend qualities greatly superior to another blend used as a standard. As the time for bottling approached, the Sales Department decided that 86° proof would be better than 90 for reasons of its own. Unfortunately, it was found that the reduction in proof, without any further change, was accompanied by a terrific loss in preference value of 25 as compared to 75 for the 90° proof White Seal. When the White Seal was reduced to 86° proof, the resultant relationship between the two was significantly in favor of the standard blend! We worked along endeavoring to improve the 86° proof blend, incorporating change after change. During this time a further request was made by the Sales Department that the proof be increased to 86.8°. Oddly enough, the changing of the then best 86° proof sample to 86.8° again lowered the preference value. By this time it appeared that we were getting no place and getting there in a hurry. But we were more successful at 86.8° proof than we had been at 86° and in due course of time returned the preference rating of the White Seal Blend to a point above that of the arbitrary standard blend. This indicates the importance of proof changes and proof control to the quality of a finished product.

Optimum spirit Concentration:

Arbitrarily, back labels establish the percent of neutral spirits to be employed in a given blend. Very often the decision as to what percent to use lies in the hands of someone other than the blender. Other than personal opinion, no one seemed to have any definite data concerning the optimum concentration of neutral spirits that should be employed in spirit blends. In an endeavor to find some indication with respect to this problem, a straightforward experiment was devised. A range of samples varying from 100%, 40% small grain Bourbon, 5 years of age, to 100% neutral spirits was set up with 5% intervals. The resultant 21 samples were studied for preferential differences, and it was found that as the concentration of neutral spirits increased, preferential value also increased until the highest preferential value had been reached at 100% neutral spirits! It was found that the greatest contributing factor to preference in this study tactual property exhibited by the whiskey but not by the spirit. The optimum range was found to lie between 55% and 100% spirits. It does not follow that the same range would be found had a different neutral spirit or a different whiskey been employed. As a matter of fact, we have analyzed further, perhaps more specific, evidence which indicates very strongly that as the quantity of whiskey in a blend diminishes with respect to the neutral spirits, it becomes easier to produce a blend of high absolute preference rating.

Optimum Spirit Age:

An established blend was studied to determine what influence increasing age alone of the spirits would have upon the quality of the blend. In order to do this, spirits at 0, 12, 24, and 36 months of age were employed. In so far as was possible, all variables, except age, were held constant. The following data were obtained:

That blend which employed 24 month spirits was significantly the best, while there is no significant difference between any of the other three. The factor influencing the apparent loss of preference value by an increase in age of from 24 to 36 months may well be due to any one of a number of things; for example, difference in production, increased woodiness, etc. The only reason for discussing this work is to point out that just as the problem of age of whiskies has not been settled, so neither has the problem of maturation of neutral spirits been solved. Much more extensive work for both of these projects is planned.

A Blending Study:

Let us assume that having progressed to this point, we have developed a suitable blend (by suitable is meant one which has a pleasing taste and odour and which is acceptable to the public). We should have established the several goods employed in the blend, and also the percentage of each which was to be used. Would it not be interesting to know what some slight variation such as the withdrawal of one of the goods might bring about? Add to this the desire to overcome a common fault. The usual fallacy to which many blenders adhere is that certain whiskies or spirits should be good for a blend, while others are harmful. Quite naturally, therefore, it is suspected that when one merely chooses the whiskies he knows to be of the best type and incorporates with them spirits known to be beneficial, and fortunately this fallacious opinion has been shown in its true light by experimental evidence which follows. But it is entirely unwarranted and without any basis in fact for a person to say: “This whiskey, if used in the blend, will improve its quality; or that spirit, in included in the formula, will most certainly harm the product.” (Obviously, excepting faulty distillates.)

Indicated by X in the first section of the data reading across the page are the whiskies employed with the various spirit combinations, designate by the numbers 1-10 reading down the page; that is, with whiskey combination “X” 10 spirit combinations were employed to producer 10 different blends. Similarly, with whiskey combination “Y” 10 blends were made employing the same 10 spirit combinations as for whiskey “X”. The difference between the whiskey portions “X” and “Y” is noted to be rather a small one. This was intentional, since it was hoped to demonstrate the importance of even relatively small alterations.

[To interpret this, X and Y are proxies for the big personality spirits that drive the blend. They each represent five spirits blended together and are 28.5% of the blend where other minor parts of presented with variations in 1-10. On the far right of the table, you see if there is a preference for X or Y for each given change to the minor parts of the blend. The arrows indicate examples where the preference was dramatic. N/C means new cooperage.]

Inspecting the spirit combinations, we shall note that we have dropped #2 the 2% R/C-B which appeared in #1, and in this place have added 2% W/C-A. Next #3 we added in its place 2% D/C-A. In #4, R/C-F was increased 2% to take care of the elimination of R/C-B. In sample #5 the cistern spirits were increased 2% to accommodate the charge. In all these first five, sample N/C-B was allowed to remain unaltered. Since, however, it is the same type spirit as the one in question and is different only to the extent that it was matured in new cooperage, it was necessary to continue the work with the elimination of N/C-B, also. Therefore, in samples 6,7,8,9, and 10 we note the substitution of N/C-A for N/C-B. This is the only logical substitution which can be made since they are the only two new cooperage spirits in the formula. Having made this change, samples 6,7,8,9, and 10 are employed to indicate the effect of the substitution of the various other spirits for R/C-B as it is dropped from the formula.

Considering R/C-B and N/C-B as the spirits in questions, and X-1 as the control, note in series X that 5 samples are better than X-1 and none significantly poorer. Of these five, four have no B spirits at all, while one has only R/C-B and no N/C-B. Since there is no sample significantly poorer than X-1, one is inclined to conclude that B spirits are no good. But change the whiskey portion slightly as in Y and had this been the whiskey portion to begin with, this half of the experiment would have shown that only one sample was better than the control Y-1, with three samples definitely poorer and five not significantly different in preference rating. Here, than, the use of B spirits is apparently advantageous!

The change in whiskey itself (spirit portion constant) made no significant impression on preference value in six cases, while in three cases changing from X to Y was harmful and in one beneficial.

To summarize the findings: The fact shown here is simply that every item included in a blend is responsible for the influence of every other item on the quality of the finished product. Unless a single item is definitely faulty, it is unfair and dangerous to predict its influence without trial.

Best Whiskey To Use In A Complete Blend:

Having some idea of the age and optimum concentrations of a whiskey, one may have proceeded to the point where a blend had been completed. If they are inquisitive, they will no doubt wonder what would have happened had they used one of the other whiskies against which they discriminated in the earlier part of their work; or they may wonder, and rightly so, if the best whiskey outside a blend is also going to be the best whiskey to be used within a blend. Three Bourbons were studied by themselves with the following results:

[Note the highest preference here goes to the highest percentage of small grains. Could that be a significant variable?]

As whiskies alone, when compared to one another, KBB is obviously the best, although not much more significantly so than KCB. KDB is unquestionably the poorest of the three at this age.

[What I’m curious about is features of a whisky that may make it achieve true maturation faster than other. This document is nearly 80 years old and there may have been more diversity of whiskey features where maturation could appear more varied after the same duration.]

Kessler Private employs in its formula 14% continuous still Bourbon. It is found that when the three Bourbons mentioned above are used to replace that Bourbon and a study of the blends so obtained is made, the following results are obtained:

The difference between these three blends is significant in each case with the highest preferential rating going to the KCB blend and the lowest to KDB. Correlation is noted between the results for KDB, since in each experiment it was the lowest, but unpredictable is the fact that KCB produces a significantly better blend than KBB, although KCB, as a straight whiskey, is not the best of the three. So again, it is evident that one may not assume a dogmatic attitude with regard to the merits of any given item as a blending whiskey.

Flavoring Materials To Be Used:

We often hear opinions expressed concerning the advisability of using so-called blending agents. At the present time such evidence as we have accumulated indicates that blending materials are of great importance to the finished product. We have yet to find a case in which the addition of a suitable flavoring material does not increase the preference value of the blend. As evidence of this, witness the following in which the only variable was the type of flavoring material.

Of course, this is an isolated example; and it does not infer that Scotch Malt at 0.5% is the best flavoring material, but indicates only that this is true under the conditions of this one experiment.

In the above series there was a generic difference between each flavoring material employed. But let us examine a few instances of the profound influence brought to hear on a blend by the use of very similar blending materials. In the following groupings, two blending sherries, domestic (D) and imported (I), were used alone and in combination with one another. It is noteworthy that the difference between these two sherries, as blending sherries by themselves, could not be called marked.

For Calvert Reserve it will be noted that the use of D and I separately in equal quantities was accompanied by no change in preference value, whereas an increase in the quantity of D definitely lowered the preference value. Marked improvement is brought about by the use of 0.1% and 0.2% D in Calvert Special, whereas in Seagram’s 7 Crown, D alone, and the mixture of D and I, were in each case below the I sherry alone. Here, we see that the blending sherry does not bear a constant relationship, regardless of the blend, but deviates with a change in the blend. That the blend itself and all the things in the blend are the important things is again borne out.

Color:

Let us discuss briefly the color at which a spirit blend should be bottled. Arbitrary standards have been set over a range of from 20-25 Lovibond units (1 inch cell). There was no basis other than opinion for these colors. Recently, upon the development of a spirit blend for Carstairs’ Harmony, it was decided to apply psychological methods, to determine the validity of placing the color within the above mentioned limits. The following range of colors and visual preference values was studied and the noted visual preference values obtained:

It is noted that the highest preference value is obtained by a sample of approximately 44 Lovibond units of color. By interpolation it was found that a color of 30 approaches, if we can assume a smooth curve, the preference value of the peak sufficiently to make its use satisfactory. Therefore, it was recommended that the new Carstairs’ Harmony be colored to 30. We shall not discuss the method of obtaining these date other than to say that the results are reproducible.

Conclusions:

It is such evidence as we have been endeavoring to put before you that we continually seek, and it is by the amassing of such data that we hope to be able to class blending eventually as a real scientific procedure rather than one colored by the artistic touch. Blending, as a science, is still in its infant stage of development; but if we have done nothing more, we certainly have laid open for study almost every aspect attributable to blending. There is scarcely a single feature for which we cannot present data showing conceptions, regardless of the school of thought, to be partly or wholly in error. It is our objective to free this phase of our field from the encumbrances placed upon it by opinions, ideas, prejudices,—and wishful thinking. Some day the formulating of a blend may be no more complicated than the development of any straightforward research problem.