BJCP Color Guide

By A.J. deLange

At the Montgomery County Fair I noticed several beer judges were using the new BJCP Color Guide and so I thought Newsletter readers might be interested in knowing how this simple aid was put together. For those who have not seen it, the Color Guide is a 2.5” by 5” card upon which are printed 12 color patches each labeled with a SRM value ranging from 1 to 44. Each patch approximates the color of an average beer with the given SRM as viewed in a 5 cm container by daylight. To describe its evolution we’ll have to discuss how color in general and beer color in particular are perceived and measured.

I certainly can’t give a complete description of color vision here but as most of you will know the eye contains four types of receptors: rods, which are color blind and used in dim light, and three types of cones. Each cone type responds to range of light colors in either the long (red), medium (green) or short (blue) wavelength part of the spectrum. The color we see when looking at an object depends on the level of stimulus of each of these three cone types which depends, in turn, on how the light energy reaching our eye is distributed over the entire visible spectrum (from 380 to 780 nm wavelength). Thus, to determine the color of an object we need to know the spectral distribution of the light reflected from it or, in the case of beer, transmitted through it and that depends on the spectral distribution of the light source and the amount of light of each wavelength which passes through the beer unabsorbed. If we know these and the responses of each type of cone we can calculate a perceived color and express it in terms of 3 numbers. There are several systems for reporting color with R, G, B and L, a, b probably being the best known.

The spectral distributions of standard light sources (Daylight (C), tungsten (A), D50, D65 etc.) are tabulated in standards and texts as are the responses of the cones (color matching functions) so all we need to measure to characterize beer is its transmission spectrum which is the fraction of light that a given thickness of beer passes at each of 81 wavelengths starting at 380 nm (violet) through 780 nm (red) in 5 nm steps. These are measured with an instrument called a photometer or spectrophotometer. To calculate the color of the beer as seen in a particular light we simply multiply the intensity of the source at each wavelength by the transmission of the beer at that wavelength and then by each of the 3 color matching functions for that wavelength. The sums (over the wavelengths) of these products yield the 3 color values in R, G, B; L, a, b or other color systems after a little math. Because, for example, tungsten and daylight sources have different spectral distributions (the former relatively stronger at the red end of the spectrum and the latter at the blue end) beer will appear to be different colors when viewed in these lights. In addition, the amount of light transmitted by a beer at each wavelength depends on the thickness of the beer through which the light must pass. Thus beer viewed through a 1 L mass will look much redder than the same beer viewed through a 2 cL Kölsch stange.

Given that it takes three numbers to describe the color of something how can the BJCP aid use the single SRM value to describe beer color? Given that beer color depends on glass size and light source how can BJCP put colored patches on a card and have them represent what we might actually expect to see? Part of the answer, of course, is that it can’t to a precise degree. The goal is to get close enough that SRM can be estimated to within a couple of units. The nature of beer helps us to do this.

The SRM was conceived by Stone and Miller (Ref 1) in 1949 when they realized that beer spectra normalized by the 430 nm value (readings at other wavelengths are divided by the 430 nm reading) were nearly the same. Conceptually this means that one can reconstruct the entire spectrum of a beer from a single measurement at 430 nm. The normalized spectrum of beer, as averaged over measurements of an ensemble (Stone and Miller used 39 beers) is simply multiplied by the value measured for the beer of interest at 430 nm. From the complete spectrum one can go on to calculate the color of beer in any light and any sized glass. The SRM (Ref 2) number is simply -12.7 times the logarithm of the spectral transmission of 430 nm light through 1 cm of beer.

This would work very well, meaning that the SRM number would be a complete specification of beer color, if it were strictly true that normalized beer spectra were all exactly the same which, of course, they aren’t. Stone and Miller knew this and required, though few seem to remember it, that each beer be tested to be sure it has “average spectral characteristics” before an SRM value could be attached to it. In my own work (Ref. 3) I have examined the deviations of normalized beer spectra from an ensemble average of normalized spectra and proposed that for beers which deviate from the average the deviations, simply coded, be reported along with the SRM value. The full spectrum would then be reconstructed from the normalized average and the deviation values. In the course of my investigation I measured the normalized spectra of 99 beers and averaged them. The result is the normalized spectrum of beers with “average spectral characteristics”. To produce the BJCP aid it was a “simple” matter of scaling that spectrum by each SRM number for which a patch was to be printed, choosing an illuminant (‘C’ i.e. daylight) and path thickness (5 cm) representative of the conditions under which judges view beer, calculating the L, a, b colors for each SRM value and having a printer print patches with the given L, a, b values on a transparency. The color you would see if you held the transparency up to a light source close to Illuminant C in looking at a particular patch would be what you would see if you held up 5 cm of average beer with the SRM value printed next to that patch.

Indeed everything except the last step was simple. While one might think any printer could readily reproduce a given L, a, b specification, it turns out that those who can, charge a great deal more than the BJCP had budgeted for this project. Printing on transparency material added much to the cost and difficulty. Thus we wound up with the white card stock, which turns out to work better than I had expected it would. Though colorimeter measurements of the patches do not match the specified colors as closely as might be hoped, experiment showed that inexperienced people using the card can judge beer SRM to within a unit or so at the low end of the range and 3 to 6 units at the upper end.

To best use the card arrange a light source, preferably daylight or an electric lamp with a color temperature that approximates a daylight (e.g. Eiko SP30/955K) source over your shoulder. Obtain a piece of very white paper and place it in front of you with the beer to be tested in a container whose interior measures 5 cm (many glasses measure about this diameter but a 5 cm square glass bottle is ideal). Look at the white card through the beer (use the center of a glass i.e. the path which is 5 cm long) and make sure that the card behind the container is directly illuminated by the light source (not shadowed by your body or the container). Place the guide adjacent to the container and pick the patch which most closely matches the color of the white card as viewed through the beer. Interpolate between patches if you feel you can.

Ref. 1: Stone, I and Miller, M.C., “The Standardization of Methods for the Determination of Color in Beer”. Proc. Am. Soc. Brew. Chem. Pp 140-147, 1949

Ref. 2: American Society of Brewing Chemists, Methods of Analysis, “Beer-10A Spectrophotometric Color Method”. ASBC, St. Paul, 2006

Ref. 3: deLange, A.J. “The Standard Reference Method of Beer Color Specification as the Basis for a New Method of Beer Color Reporting”. J. Am. Soc. Brew. Chem. 66 (3):143-150, 2008