Critical Reviews in Food Science & Nutrition

ISSN: 0099-0248 (Print) (Online) Journal homepage: http://www.tandfonline.com/loi/bfsn19

Colorimetry — methodology and applications

F. M. Clydesdale & E. M. Ahmed

To cite this article: F. M. Clydesdale & E. M. Ahmed (1978) Colorimetry — methodology

and applications , Critical Reviews in Food Science & Nutrition, 10:3, 243-301, DOI:
10.1080/10408397809527252

To link to this article: https://doi.org/10.1080/10408397809527252

Published online: 29 Sep 2009.

Submit your article to this journal

Article views: 77

View related articles

Citing articles: 85 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=bfsn20
 December 1978 243

COLORIMETRY — METHODOLOGY AND APPLICATIONS*

Author: Fergus M. Clydesdale

Department of Food Science and
Nutrition
University of Massachusetts
Amherst, Massachusetts

Referee: E.M. Ahmed

Department of Food Science and Human Nuirition
University of Florida
Gainsville, Florida

INTRODUCTION ity control. The food service industry is not as

The term food quality, as frequently spoken
of, includes three principal areas: nutritional
value, acceptability, and safety. Acceptability
includes a large array of attributes, among
them visual appeal, aroma, flavor, texture,
mouthfeel, convenience, and cultural appropri-
ateness. Thus, acceptability is, in fact, what
makes eating more than just a mathematical ex-
ercise.
A sound argument may be made for the fact
that color is perhaps the most important sen-
sory or acceptability attribute of a food mate-
rial. Certainly, no one can argue the fact that
if a food material is not the right color, all other
sensory attributes fall by the wayside, since it
will never enter the mouth, and of course, when
this happens, both nutritional and safety as-
pects of quality also may be disregarded.
Accepting this, it is almost inconceivable to
realize that the food industry lags far behind in
other industries (such as paints, textiles, and
paper) in the application of colorimetry to qual-
far behind in utilizing the psychological effects
of color to enhance its products. They have
taken advantage of such effects in many ways.
The fast food service outlet where one may pick
up a quick coffee and donut in the morning is
quite often painted an almost bright bilious yel-
low color. This color imparts an impression of
cleanliness but also imparts a feeling of psycho-
logical activity which really does not enhance
the mood for sitting around and enjoying one's
coffee. However, it does provide a motivating
force for removing oneself from such an atmos-
phere as quickly as possible. High volume is the
secret in this case, and the color provides the
impetus for movement along with a feeling of
cleanliness. If one looks further at food service
areas, one can recall other eating establish-
ments where the bright yellows are toned down
by blues and browns. These would be the cafe-
terias where the meals are slightly more expen-
sive, and yet, it is imperative to move people
through in a fairly rapid fashion. Then we have
the very expensive "exclusive" type of food

* Paper No. 2197 from the Massachusetts Agriculutral Experiment Station, University of Massachusetts at Amherst. This
work supported in part by Experiment Station Project No. 192 and the Glass Container Manufacturers Institute.
244 CRC Critical Reviews in Food Science and Nutrition
service establishments which use deep reds,
deep browns, along with subdued lighting,
which is nearly always incandescent or candle-
light. There is a very good reason for this, since
it provides not only an atmosphere aesthetically
pleasing to the senses, so people may linger over
cocktails or after-dinner drinks (which of
course creates a profit), but in addition, it is
known that cosmetics look particularly good
under candlelight or incandescent light.
In certain aspects of our industry where mys-
ticism has built up, color is also included. I can
remember very vividly going to a restaurant for
dinner with a gentleman who considered him-
self a wine connoiseur. We arrived and sat
down, when the wine steward appeared at the
table and a wine of a certain vintage was or-
dered. With a great deal of fanfare, the wine
was produced, and the steward made the mis-
take of placing the glass in front of me which I
quickly drank and said was fine. This, of
course, shocked the gentleman I was dining
with who insisted on tasting this mystical de-
light prior to approving it. Having tasted the
wine, he proceeded to argue fairly violently
with the wine steward as to the quality of that
bottle of spirits. The argument continued on
for some 15 min, at which point I was becoming
incredibly hungry; finally, my dinner compan-
ion, in utter exasperation, took out his hand-
kerchief, twisted it around, shoved it into the
wine and then held this soaked handkerchief up
to the light and said to the wine steward,
"See!" The wine steward backed off and said,
"Sir, you are right." The wine was quickly re-
moved and replaced by another bottle. Mysti-
fied, I asked my friend what he had done. He
said he had no idea, but he knew that the mys-
ticism involved with wines would allow him to
get away with such a ploy, and he was tired of
arguing.
It would be rather facetious to state that
some foods or beverages are measured in our
industry in this manner, but at times the tech-
niques used are not too far off. There has been
an exceedingly slow acceptance of color tech-
nology by the food industry, and a certain de-
gree of apparent antagonism towards the use of
the sophisticated technology which is available.
It is hoped that this will be a "how to" article
as well as one which produces some thought on
the possible advantages of applying optical
techniques not only for measurement but for
optimization of food color. There is some dis-
agreement in this area, both in the measure-
ment and optimization usages of colorimetry,
but at least the disagreement is honest and per-
haps some of the things said in this article might
create thought in this area. Unfortunately, one
cannot write a "how to"-type of article without
first utilizing an introduction into what color is
and what it means. This will be kept to a mini-
mum, but it is essential for an understanding of
what follows. Therefore, the first section of this
manuscript will deal very briefly with general
color theory. Other references, such as the re-
cent books by Francis and Clydesdale21 and
Hunter,30 which deal in detail with color theory
and in particular with food color measurement,
should be referred to by the interested reader.
THEORY OF COLOR SPECIFICATION
AND MEASUREMENT
In discussing colorimetry, the first concept
that must be stressed is that the color specifica-
tion of a material is simply the specification of
a point in three-dimensional space. This three-
dimensional space, whether visual or mathe-
matical, has become known as a color solid.
One of the earliest and most successful attempts
to develop a visual color solid was that of A.
H. Munsell in the early 1900s. Munsell was
born in Boston, Massachusetts, in 1858 and
spent most of his lfe as a teacher at the Massa-
chusetts Normal Arts School. We can imagine
his frustration in attempting to convey the
meaning of color to the students. His deep con-
cern for a meaningful description of color lead
to the publication in 1905 of the first edition of
A New Classification of Color."" Munsell con-
tributed materially to the concept of a color
solid. He developed names to describe the three
attributes of color. If we can picture an orange,
the vertical core would represent Munsell value
ranging from lightness to darkness. That is,
going up or down the core of the orange would
describe one of the three-dimensional qualities
of color, that of lightness or darkness. Each
segment of the orange would represent Munsell
hue, that is green, yellow, red, etc. Hue is the
single function of color which is commonly
used to subjectively describe a total "color".
Thus, as we go around the parameter of the or-

ange and go from such colors as green, yellow,
red, etc. we are really describing only one of the
three parameters of color that is hue and not
the total color of the entire system. The dis-
tance from the core of the orange to the skin
could represent Munsell chroma or intensity of
color. Thus, each point within the boundaries
of the orange or the color solid as represented
by the orange would represent one, and only
one, color. This humble beginning lead to the
present Munsell Book of Color which was first
published in 1929. Since that time, the Munsell
Color Company Baltimore, Maryland has pro-
duced innumerable versions and many types of
colors on different surfaces are available. This
is probably the most successful of the visual
color solids used in the food industry today.
From this it follows that colorimetry, both vis-
ual and instrumental, is simply a means
whereby the location of a point may be de-
scribed in terms of numerical values. A com-
mon question which arises in this area of instru-
mentation is: why do we have to use
instruments and numerical values to describe a
color? After all, everyone who has normal
color vision can use his or her eyes and make a
judgement, why is there a need for
instruments?
Unfortunately, eyes must be present within a
living body in order to make a subjective judge-
ment and all the biases, personality differences,
and associated human frailties present subject
the color judgement to many different interpre-
tations. This is one problem associated with a
color judgement, but another one which is
SCREEN
FIGURE 1. Color matching by the addition of lights.
December 1978 245
equally important is the near impossibility of
describing a color in subjective terms; what do
the terms dark red, musty yellow, or lavender
pink really mean? This problem may be exem-
plified by a quote from Robert Louis Stevenson
when writing from Samoa in 1892 to a friend
in London:
I should rather like to see some patterns of unglossy — well,
I'll be hanged if I can describe this red — it's not Turkish
and it's not Roman and it's not Indian, but it seems to par-
take of the two last, and yet it can't be either of them be-
cause it ought to be able to go with vermillion. Ah, what a
tangled web we weave — anyway . . . . send some —
many — patterns of this exact shade.
Obviously, there is a need for some sort of
objective measurement which would be a func-
tion of what the eye sees yet provide a solution
to some of the problems which have been
enumerated.
In the field of instrumental color measure-
ment, we are much more fortunate than those
workers in other sensory areas who are attempt-
ing, with some success, to create objective mea-
surements. This is due to the fact that people
with normal color vision see each and every
color in almost an identical manner. Moreover,
it is known that with the appropriate combina-
tion of red, green, and blue lights shone on a
screen an observer can match such colors. This
may be done simply as shown in Figure 1, as
first described by Newton" and Grassmann."
The color to be matched is projected onto a
screen, and in addition, each of three projectors
shines red, green, and blue primary light, re-
COLOR
PROJECTOR
246 CRC Critical Reviews in Food Science and Nutrition
spectively, on the other half of the screen.
There is a complete overlapping of the three
primaries to create one spot whose color is
changed by changing the intensity of red (R),
green (G), or blue (B), light striking the screen.
Thus, any color may be specified mathemati-
cally in terms of the intensities of the three pri-
maries required to create a match. Unfortu-
nately, it was found experimentally that all
colors cannot be matched by the addition of
three primaries, even if these primaries are
spectrum colors. However, this problem can be
overcome simply by adding one of the primar-
ies to the test color and matching the combined
color by the addition of the other two primar-
ies. Then, for purposes of describing the test
color, the light added to it may be thought of
as being subtracted from the other two primar-
ies. That is, the test color can be described with
a positive amount of two primaries and a neg-
ative amount of the third.
In order to develop this concept into a work-
able system of color measurement, it was nec-
essary to reach agreement on two separate is-
sues. The first of these would be to establish the
amounts of the three primaries required to
match any color throughout the visible spec-
trum in terms of a representative population of
people with normal color vision. The second
would be to arrive at a suitable mathematical
system in order to express the coefficients of the
primaries in a manner which facilitated their
manipulation as color units.
The first problem resolves itself into the def-
inition of a standard observer's response to the
visible spectrum which, as stated previously, is
similar for all people with normal color vision.
Referring back to the type of equipment illus-
trated in Figure 1, one may see how such stand-
ard observer responses may be obtained. If the
test light represented a certain wavelength of
the visible spectrum, then an observer with nor-
mal color vision could match that color by
means of additive or subtractive mixing of the
three primaries R, G, and B.
This could be done for each wavelength, and
the coefficients or tristimulus values of the pri-
maries could be plotted at various wavelengths
to obtain standard-observer curves for the visi-
ble spectrum. The tristimulus values of the
standard observer are represented by the sym-
bols r, g, and b. By using a random sample of
persons with normal color vision, average
standard-observer curves may be developed
which represent the physical stimuli causing
each color to be perceived by the brain.
The Commision Internationale de 1'Eclairage
(CIE), which is the most prestigious body con-
cerned with colorimetry, defined color match-
ing characteristics of the standard observer on
this basis in 1931. Several modifications in the
definition have occurred since then, but the
techniques used remained fundamentally the
same.
Having defined the standard observer curves,
the CIE had to consider the second problem
which was to define appropriate reference stim-
uli or tristimulus values. As stated previously,
the use of R, G, and B involves the use of neg-
ative quantities. Negative values of color were
unlikely to be appreciated by those who were
going to use the system. In addition, these neg-
ative values would complicate the mathematical
manipulation of color units and tend to hamper
the design of photoelectric instruments
equipped with mechanical integrators.
For reasons such as this, the CIE decided
that, although the R, G, B system was appro-
priate for defining standard responses, another
system of reference stimuli should be devel-
oped. Thus, X, Y, and Z were chosen by the
CIE as imaginary primary lights for the descrip-
tion of colors. These primaries were selected so
that:
1. All possible real colors could be
"matched" by positive amounts of the pri-
maries;
2. A relatively large range of colors in the yel-
low-red region could be "matched" with
only two primaries (note in Figure 2 that
the z curve ends in the yellow region, and
colors beyond this may be matched by x
and y only); and
3. The intensity (luminosity or lightness) of
the light needed to make the "match" is
specified by the Y primary alone.
Figure 2 shows the standard-observer curve
in terms of x, y, z. These curves were developed
from r, g, b standard-observer curves, and it
should be emphasized that although x, y, z may
be calculated rom r, g, b experimental data, the
X, Y, Z values cannot be produced experimen-
tally.
Having discussed the rationale for the devel-

400 500 600 700
Wavelength - nm
FIGURE 2. Standard-observer curves based on the CIE
X,Y,Z system.
opment of instrumental color measurement, it
is now appropriate to describe the general types
of instruments and their theoretical basis. As
stated previously instrumental measurement
may be divided into three parts: (1) spectropho-
tometry, (2) tristimulus colorimetry, and (3)
visual colorimetry.
Consider spectrophotometric measurement
of color using the CIE system. An instrumental
measurement must measure the physical stimuli
in terms of how the eye sees color. The instru-
ment achieves this by using either one of two
different arrangements of standard source, test
sample, and photodetector.
One is called 0 to 45°-viewing and is achieved
by measuring the light which reflects off the test
sample from the source at an angle of 45°. This
is known as diffuse reflection and is mainly a
function of color.
The other arrangement involves the use of an
integrating sphere, which is a hollow metal
sphere painted white inside. An integrating
sphere collects all the light reflected from the
surface of a sample placed against an opening
(commonly called a port) in its side. The light
collected for measurement may or may not in-
December 1978 247
elude specular reflection (mainly a function of
gloss) along with the diffuse reflection. Both of
these methods are used, but unfortunately they
do not give identical results for all types of sam-
ples.
Readings are obtained over the visible spec-
trum and a spectral response curve for the test
sample is obtained. This response is symbolized
byR.
The monochromatic light used is a CIE
standard source that has a defined spectral
curve of its own. The energy elicited at every
wavelength of this curve is symbolized by E.
Thus, the total energy received by the photo-
tube is the product RE.
In order to obtain values analogous to what
the eye sees, this energy function RE must be
multiplied by the functions x, y, and z, of the
standard observer curve, respectively. The mul-
tiplication may take place at intervals across the
visible spectrum, so that at any one wavelength,
X = REx, Y = REy, and Z = REz.
If the wavelength interval across the visible
spectrum was d, then the color of the object
may be specified by integrating between 380
and 750 nm as follows:
X
= 380* 750 RE
*d*
Y
- 38oS 750 R E
^
Figure 3 shows this operation in a diagram-
matic manner.
The integrals could be calculated manually;
however, normally, instruments have a me-
chanical integrator which carries out these op-
erations so that a direct read-out of the tristim-
ulus values X, Y, and Z is obtained.
In order to obtain the tristimulus coefficients
or chromaticity coordinates, these values are
merely expressed as fractions of their total so
that:
X
X +Y + Z
Y
y ~
X 1• Y + Z
Z
r = -
X +Y +Z
248 CRC Critical Reviews in Food Science and Nutrition
FIGURE 3. A diagrammatic illustration of the calculation of the X, Y, and Z
primaries.
I
Surface
to be
measured
Tristimulus
filters
FIGURE 4. A simple tristimulus colorimeter.
The second type of instrumental color meas-
urement is that obtained using a tristimulus col-
orimeter. In this case, the integration procedure
using standard observer curves is replaced by
using filters which simulate the x, y, and z
curves, respectively, as shown in Figure 4.
The third type of instrumental color meas-
urement involves the eye as a photocell. In this
sense, the method is subjective; however, math-
Photo-
cell
Meter
ematical specifications are still provided by the
instrument, making the description of the color
objective. These instruments are widely used in
certain areas and may run the gamut from disc
colorimetry to color comparators to the use of
standardized glass filters whose combinations
form the units which specify the color to re-
fined spectrometers or colorimeters using the
eye as a detector.

FIGURE 5. The CIE x y chromaticity diagram showing
percent Y.
There are many different types of instru-
ments based on the preceeding principles. As
well as different instruments, there are a large
number of different color solids. All such solids
represent three-dimensional color space and are
mathematically interconvertible. The choice of
a given solid is dependent upon the final use
and fit of the data to visual judgements as well
as the degree of uniform chromaticity desired.
The CIE system represents one of the most
common scales or solids, and one method of
visualizing this is shown in Figure 5. Here the
spectral colors are plotted on x, y coordinates
to obtain the CIE horseshoe-shaped spectrum
locus. All colors fall within this locus but, in
order to obtain a point in three-dimensional
space, °7oY (specified as the brightness or lumi-
nosity function) is also plotted. This means that
any color may be specified in three-dimensional
space by a statement of x, y, and °7oY.
Another simple set of scales is that in which
a pure white sample (MgO or BaSO4) is taken
to read 100 when viewed through each of the
filters. Readings obtained are often called G for
green, corresponding to CIE tristimulus value
Y, R for red, related to the large peak of the X
curve, and B for blue, related to the Z curve.
In an instrument of this type (three filters),
values of X, Y, and Z can be approximated by
adding a fraction of the B reading to the R
reading in order to compensate for the small
peak of the X curve. Then the modified R read-
ing plus the G and B readings may be multiplied
by appropriate factors to obtain tristimulus val-
ues.
December 1978 249
CIE Y for MgO is 100 so that when the G
value is set at 100, it becomes equal to Y by
definition. Adding a fraction of the B reading
to the R reading assumes that the small peak of
the X curve is the same shape as the Z curve.
This is not strictly correct; therefore, some in-
struments use a fourth filter, R, to stimulate the
small peak of the X curve. Tristimulus values
are then calculated in the same manner as
above.
Another common set of scales developed by
Hunter35 are used on some instruments. These
produce results in terms of L, which is a light-
ness function and simulates Y; a, which pre-
dicts redness if positive and greenness if nega-
tive; and b, which predicts yellowness if
positive and blueness if negative. A diagram-
matic representation of the Hunter solid is
shown in Figure 6A. All conversions described
above are calculated within the instrument, so
that direct read-outs are obtained.
Figure 6A also shows the three functions
which make up the Munsell solid mentioned
previously. This is perhaps the best known of
all the color order systems. It is a system which
attempted to create painted colors to represent
equal intervals of visual perception of color dif-
ferences between adjacent samples and also to
specify these colors in terms of its three coor-
dinates: hue, value, and chroma. Hue describes
what the average person thinks of when he
speaks of color, i.e., red, yellow, green, blue,
etc. Value describes the lightness or darkness of
the color. Chroma describes intensity.
As may be seen in Figure 6B, hue is repre-
sented on the horizontal circumference of the
solid, value is the vertical central axis, and
chroma is described in units measured outward
from the central axis.
There are many other scales or solids, too nu-
merous to mention in this discussion; however,
a review on this subject has been written by
Francis and Clydesdale.22
GENERALIZED APPLICATION OF
COLORIMETRY
The application of color measuring tech-
niques may be utilized in a predictive sense for
not only an estimate of visual appearance but
also as a tool for chemical analysis. Little45 re-
cently pointed out correctly that:
250 CRC Critical Reviews in Food Science and Nutrition
L.a.b. COLOR SOLID
L:100
FIGURE 6. The Hunter (A) and Munsell (B) solids.
Once we accept, however, that color belongs to the realm
of sensory perception, we must also accept that it can only
be measured directly in psychological terms. From physical
measurements, nevertheless, we can obtain data which pro-
vide the basis for establishing psycophysical scales, from
which we can predict visual appearance.
This emphasizes the fact that the measurement
of appearance is the measurement of a sensory
phenomenon irregardless of the use of instru-
mentation. Therefore, it is necessary to under-
stand the physical, physiological, and psycho-
logical aspects of the stimulation which
ultimately results in visual perception of the
color of an object.'
Also, when color measurement is used to pre-
dict chemical changes in a food, it is necessary
to be aware of the optical classification of the
food, the light-object interactions, and the type
of measurement technique in order to fully op-
timize the results. Generalizations from these
results may be made to other foods, but such
foods must fall into the appropriate classifica-
tion.10
In regard to the assessment of color changes,
Little47 has written an interesting paper which
presents a procedural framework in an attempt
to provide a basis for developing solutions for
specific commodities and for specific purposes.
The essentials of this framework include the de-
Value
Chroma
The Munsell Solid
velopment of model systems constructed to
vary in a known and controlled fashion along
the requisite dimensions of color; the determi-
nation of optimum conditions for analyzing
samples precisely and reproducibly for changes
in physical and chemical properties which un-
derlie changes in perceived color; the definition
and construction of psychological scales which
provide data relevant to the case in point; and,
finally, the demonstration of a highly signifi-
cant relationship between visual evaluations
and physically determined quantities. Thus, for
different products, the same conceptual model
is used, yet the procedural details may differ
widely, depending on the constraints imposed
by physical and chemical characteristics of the
sample and the specific requirements of the
case.
Little's paper47 points out that the concept of
the following basic model was developed from
observations made during the course of study
of a number of transparent, translucent, and
opaque systems. The evolution of these studies
ended in a proposed model for color evaluation
of foods and is reproduced47 as follows:
The Proposed Model for Color Evaluation of
Foods
The basic model for exploring the interrela-

tionship of the light-modifying properties of
foods (and, by extension, biological systems in
general) to visual color perception and to chem-
ical and physical properties of the system is as
follows:
1. The object
2. The synthetic process
3. The analytical process
4. The analytical process as a predictor of the
synthetic process
The object (the sample) on which our interest
is focused may be opaque, translucent, or
transparent with properties of absorbing, re-
flecting, transmitting, and scattering incident
light. Pigment distribution can range from a
single homogenously distributed pigment sys-
tem to a random distributed multicomponent
system. The degree of complexity is a major
factor in determining the conditions of sample
preparation for presentation to the instrument.
The synthetic process is affected by the ob-
server who, by virtue of the retinal receptor sys-
tem, optic nerve, and cerebral cortex complex,
translates the qualitative and quantitative dis-
tribution of light activating the receptor system
to visual color perception. Observer judgments
are evaluated by psychological scaling tech-
niques, among which ranking, rating, and
paired comparisons have provided useful infor-
mation. Independent information regarding
color vision and color discrimination ability is
obtained from screening tests, the HRR-PIC
and the Ishihara Plate Tests, and color discrim-
ination and color aptitude testing procedures,
notably the F-M 100 Hue Test and the ISCC-
CAT, currently being evaluated for effective-
ness in predicting the ability of observers to
make required visual judgments.53
The analytical process involves the determi-
nation of spectral reflectance and/or transmit-
tance characteristics of the sample as prepared
for precise, reproducible analysis.11-50 A suc-
cessful resolution of the problem of defining
the analytical procedure is facilitated by inde-
pendent knowledge of the chemical composi-
tion and physical characteristics of the system
obtained through construction of model sys-
tems or by alternative analytical procedures. In
turn, the determination of reflectance or trans-
mittance characteristics can be used to deduce
December 1978 251
certain chemical and physical properties of the
system. Model systems may be prepared by
combining materials of known properties in
various combinations to simulate a food; by the
addition of controlled increments of pigment to
a food; by blending two components by system-
atically changing the composition of the mix-
ture; by controlled oxidative degradation of a
pigment system; by addition of light scattering
particles to a nonturbid medium, etc.
The analytical process as a predictor of the
synthetic process is defined through the deter-
mination of a highly significant relationship be-
tween objectively determined quantities and
visual evaluations, i.e., the psychophysical re-
lationship. The objectively determined quan-
tities can range from reflectance or transmit-
tance measurements across the visible spectrum
reduced to XYZ or RGB, to isolated portions
of the light stimulus (such as single or weighted
combinations of broad- or narrow-band mea-
surements), to levels of nonvisible components.
The latter is particularly useful in the instances
where changes in the levels of ultraviolet ab-
sorbing components can be shown to parallel
exactly the development of color through the
nonenzymatic browning reaction, as demon-
strated with roasting of coffee.
Just as it is important to gain information re-
garding the performance characteristics of the
human eye when used as an instrument to make
critical color judgments, so too must the per-
formance characteristics of the spectrophoto-
meter or colorimeter be known.
A model such as this is important, not be-
cause it is the only way to approach color meas-
urement, but because it specifies all the param-
. eters which must be considered in a color
measuring system before experiments can be
designed and conclusions drawn. Little45 reiter-
ates this model in a more recent publication and
provides several examples which clearly indi-
cate the viability of the approach.
Generally.the many different ways in which
color may be measured have been categorized
by Hunter30 into seven different groups:
1. By spectrophotometric curves (physical
analysis)
2. By visually equivalent additive mixtures of
red, green and blue
3. By location in a three-dimensional color
252 CRC Critical Reviews in Food Science and Nutrition
solid; for example (a) CIE trilinear coeffi-
cients x, y, %Y and (b) opponent colors
with gray in the center (L, a, b, or
U*V*W*)
4. By densities of three substractive-primary
inks or dyes required to represent the color
(color photography and color printing)
5. By location with respect to visually system-
atic arrangement of color chips (Munsell)
6. By location, with respect to an array of
chips, of systematic substractive mixtures
of white, black, and colored pigments of
each separate hue (Ostwald system). These
substractive-mixture dimensions of each
color are familiar to the formulator or dyer
who is experienced in obtaining product
colors by eye
7. By ingredients required to obtain the color
in a given product: (a) derived by color for-
mulator's estimates; (b) derived by optical
model of product, optical constants of in-
gredients, and a computer
There can be, of course, a certain degree of
overlapping of these groups in order to accom-
plish a specific measurement. Furthermore,
when considering instrumentation, it is also ap-
propriate to categorize the object in an optical
sense as well as categorizing the type of meas-
urement technique.
Hunter30 has pointed out that a number of
things happen when light falls upon an object.
First if the object is opaque, light will be re-
flected by it. If it is transparent or translucent,
light will pass through the object, or be par-
tially reflected or transmitted by the object pri-
marily in one of two manners: specularly or dif-
fusely. Specularly reflected light is that which
(as with shiny objects) is concentrated in the
mirror direction of reflection, as a continuation
of the incident beam. The straight-through
transmission of the light by a transparent object
can also be thought of as specular, although
this admittedly is taking some liberties with the
literal meaning of the word. Diffusely reflected
or transmitted light, on the other hand, is that
which leaves the object uniformly in all direc-
tions. Similarly, a white translucent film trans-
mits light, uniformly in all directions; that is,
the source of light is not recognizable through
it as the source.
The foregoing analysis provides the basis for
an arbitrary classification of objects into four
groups, based on whichever of the four man-
ners of light projection is dominant:
1. Opaque, nonmetallic, such as a yellow ce-
ramic vase.
2. Opaque, metallic, such as a metal bell.
3. Translucent, such as a plastic tumbler.
4. Transparent, such as a bottle of clear yel-
low liquid.
Reflecting objects fall in Group 1 or 2; trans-
mitting objects in Group 3 or 4. Objects which
reflect or transmit light primarily in a specular
manner fall into Groups 2 and 4. Groups 1 and
3 are made up of objects that reflect or transmit
light in a diffusing manner.
Having mentioned briefly the four subgroups
in the object mode, it is now necessary to con-
sider the object-light interactions which affect
measurement. Such interaction may produce
the following effects which have been discussed
in some detail by Clydesdale."
1. Reflection from the surface
2. Refraction into the object
3. Transmission through the object
4. Diffusion
5. Absorption within the object
Thus, Hunter's" categorization explains in
more detail the first or object section of
Little's47 model, and Clydesdale's" comments
expand on the third or analytical process sec-
tion of that model. Therefore, as mentioned
previously, it is essential to take into account
not only the model described by Little4' but also
the optical classification of the food, the light-
object, interactions and the type of measure-
ment technique as summarized by Clydesdale.'"
The third section of Little's model involves
the analytical or measuring process itself, which
involves instrumentation to determine the spec-
tral reflectance and/or transmittance character-
istics of the sample. For those interested in
color-measuring instruments and their selec-
tion, the Inter-Society Color Council (ISCC)
has released a report entitled A Catalog of
Color Measuring Instruments and a Guide to
their Selection."This report was prepared due
to the fact that in the last few years, there has
been a marked proliferation in color-measuring

instruments. So many varieties are now avail-
able that it has become very difficult for the
neophyte to select the equipment most suited
for use in solving, problems. The existence of a
relatively simple but complete catalog of avail-
able instrumentation would help to alleviate
this situation.
A problems subcommittee of the Inter-Soci-
ety Color Council was established to compile
and issue such a catalog. The purpose of this
document was to list commercially available in-
struments for measuring color and color differ-
ence, together with their associated data-reduc-
tion accessories. It was understood that color
measurement of a reflecting or transmitting
material consists of the determination of mea-
surements which can be directly related to as-
pects of color perceived by a specified observer
in a specified illuminant and condition of view-
ing.
The information tabulated for such instru-
ment includes salient design features such as
character of illumination, viewing geometry,
sample size, etc. The subcommittee did not un-
dertake comparative testing of instruments.
Sufficient editorial comment is included, how-
ever, to aid the prospective user in analyzing the
nature of their measurement problem in order
that they can select appropriate instrumenta-
tion. However, this catalogue was published in
1970 and is therefore out of date. I would
strongly recommend that a prospective buyer
both write and talk to the color equipment sup-
pliers, not only to find out what is available,
but also to see how specific equipment fits user
needs.
UTILIZATION OF COLORIMETRIC
DATA
The literature abounds with color measure-
ment data utilized in a raw fashion, as read di-
rectly from the instrument. This is done in
many cases in such a frivolous manner that it
appears that at times color measurement has
been "tagged-onto" an otherwise fine piece of
research to round it out. Unfortunately, such
data, when used in this manner, is meaningless
and denotes a lack of understanding on the part
of the users. Not only does this usage not con-
sider the basic theory and application of such
theory to measurement systems, but it also
December 1978 253
shows a lack of understanding of the meaning
of the tristimulus values which are produced
from instrumental measurements. Further, this
problem has often been compounded by inter-
conversion of data between different systems
and/or the reduction of data without a clear
understanding of such reduction or experimen-
tal evidence to justify it.
Yeatman" has recently correctly stated that:
Plant product color evaluation has been reduced to what-
ever colorimetric response correlates best with visual judg-
ment without due regard for the three-dimensionality of
color. For example, when spectrophotometric measurement
of samples of processed tomato juice indicated a significant
difference at specific wavelengths (546 and 640 nm) instru-
mentation was created to isolate the response at these wave-
lengths. There are many such instruments in food industry
use today. Response to the red, green, and blue portions of
the spectrum, referenced to the standard observer, was not
considered, nor was the concurrent expression of tridimen-
sionality — in terms of X, Y, Z; R, G, B; Hunter L, aL,
bi, or any other system of color which dealt with response
— closely related to what the eye sees.
The gamut of data transformation has been extensive.
Color data have been transposed into such mathematical
functions as aL/bL, tan"1 a/b, and 2000 cos9/L (the USDA
Tomato Color Index) in order to obtain the highest corre-
lation to a psychophysical standard.
For these reasons, it is appropriate to review the
use of certain color functions in order for the
reader to evaluate more clearly the intended
and appropriate usage for such functions.
Clydesdale and Francis12 investigated the va-
lidity of using converted and/or reduced data
in formulating a prediction of visual judgment
on processed spinach purees. The color conver-
sions tested were known to be valid with
opaque color standards and the purpose of this
investigation was to check their validity with a
translucent food sample such as spinach puree.
Further, it should be pointed out that just as
one cannot make assumptions between opaque
color standards and a food system he also can-
not make assumptions that particular data ma-
nipulations which provide adequate correla-
tions to visual data cannot be used for another
different food system, unless adequate testing
is carried out. It would be of interest to review
the procedures used by Clydesdale and Fran-
cis12 in order that such comparisons may be
made prior to incorrect usage.
In this work, freshly processed purees and
stored processed purees were mixed in propor-
tions varying from 0 to 100% in increments of
254 CRC Critical Reviews in Food Science and Nutrition

10% to provide a series of 11 samples. These i.e., blue. Apparently blue is less important in
samples simulated the range of color values ac- color judgments of processed spinach puree
tually obtained with samples in storage after than the other two tristimulus values. The mul-
processing. With each set of experiments, a dif- tiple correlations (Table 1) are higher than the
ferent stored processed puree was chosen to simple correlations, but the improvement is not
vary the range of colors slightly. A total of marked.
three different sets of samples were prepared The transformation of instrumental tristim-
and examined in three different experiments. ulus values to the other systems resulted in very
Visual judgments for each set of samples different absolute values from those obtained
were performed with 25 panel members. All from direct instrumental read-out. However,
panel members were instructed to rank the sam- although the absolute values were changed
ples from 1 to 11, the rank of 1 being assigned drastically, the values in relation to ranking of
to the sample with the most acceptable color for color were not affected to the same degree in
a commercial spinach puree and 11 assigned to most cases. This may be seen in Tables 2, 3,
the least acceptable. Controlled illumination of and 4, which show correlations between raw
the samples was obtained by the use of Mac- and transformed data vs. visual ranking for
Beth Examolite with horizon sunlight. General Electric X, Y, Z; Hunter L, a, b; and
Instrumental color data were obtained with Colormaster R, G, B, respectively.
three instruments: a Hunter-lab Model D25® In general, the correlations with calculated
Color Difference Meter; a Colormaster® Dif- values were not as good as the values obtained
ferential Colorimeter, Model V; and a General from the instrument. This shows one of the
Electric (GE) Recording Spectrophotometer®. dangers implicit in using transformations on a
Each set of tristimulus values obtained in- food material when visual judgments are not
strumentally was converted by the standard available.
conversion equations to the other tristimulus The correlations of raw and transformed
values by means of the computer program de- Hunter L, a, b values in Table 3 show that when
veloped by Clydesdale and Podlesny". In ad-
dition, instrumental and calculated tristimulus TABLE 1
values were reduced to such color functions as Correlations of Visual Rankings vs. Instrumental Tristim-
a/b, tan"' a/b; (a2 + b1)1'2; G/R; and X/Y. In ulus Values for Processed Spinach Purees
the case of R, G, B, a conversion to Adams L,
a, b tristimulus values was carried out, since Correlation, visual ranking
there is no instrumental method for obtaining with Correlation coefficients
Adams L, a, b. These values were considered
Instrument Value I II III
as reduced data. All values were correlated
against visual rankings to investigate their rela- (r)
tionship to the subjective evaluations. Delta E Colormaster G 0.959 0.980 0.986
values, a measure of total color difference, R 0.956 0.979 0.987
were also calculated to determine the distance B 0.879 0.968 0.979
Hunter L 0.961 0.959 0.992
between the most and least acceptable samples a 0.945 0.993 0.976
in color space. b 0.979 0.830 0.994
Table 1 shows the correlation obtained be- GE X 0.960 0.979 0.990
tween instrumental tristimulus values and vis- Y 0.968 0.980 0.989
Z 0.948 0.876 0.909
ual ranking for each of three separate experi-
(R)
ments. As expected, data from all the Colormaster GR 0.960 0.981 0.989
instruments produced high correlations in all R.G.B 0.975 0.984 0.989
three experiments. The third tristimulus value Hunter a,b 0.986 0.993 0.995
(Colormaster B, Hunter b, and General Electric L.a.b 0.986 0.993 0.996
Z) was lower in at least one of the experiments. GE X,Y 0.976 0.983 0.990
X.Y.Z 0.976 0.989 0.997
This was probably because this tristimulus
value is a measure of the energy of the visible Note: I, II, and III refer to three separate experiments,
spectrum concerned with lower wavelength, each having 11 samples.
 December 1978 255

TABLE 2

Correlations of Visual Rankings vs. CIE X,Y,Z Data Obtained with a General Electric Recording Spectro-
photometer ® and as Calculated From Other Systems

Correlation coefficients
Calculated from
GE recording Calculated from Hunter D Colormaster® Model V
Correlation spectrophotometer 25® L,a,b values R,G,B values
of visual
ranking with: I II HI I II III I II III

r
C1EX 0.960 0.979 0.939 0.947 0.991 0.944 0.979 0.987
CIE Y 0.968 0.980 0.957 0.956 0.989 0.959 0.980 0.987
CIEZ 0'948 0.876 0.848 0.695 0.976 0.883 0.968 0.976

CIEX.Y.Z 0.976 0.989 0.987 0.988 0.996 0.984 0.984 0.989

Note: I, II, and III refer to the average of duplicates of each of three separate experiments, r = simple
correlation; R = multiple correlation.

TABLE 3

Correlations of Visual Rankings vs. Hunter L.a.b Obtained from the Hunter D25® and as Calculated from
Other Systems

Correlation coefficient
Hunter L,a,b calculated
Hunter L,a,b calculated from GE recording
Hunter L,a,b read-out from Colormaster® Spectrophotometer®
Correlation of from the Hunter D25 Model V R,G .B X.Y.Z
visual ranking
with: I II III I II III I II III

r
Hunter L 0.961 0.959 0.992 0.964 0.985 0.990 0.959 0.982 0.996
a 0.945 0.993 0.976 0.741 0.611 0.972 0.924 0.983 0.997
b 0.979 0.830 0.994 0.972 0.987 0.991 0.963 0.989 0.999
R
L,a,b 0.986 0.993 0.996 0.986 0.988 0.991 0.979 0.993 0.998

Note: I, II, and III refer to the average of duplicates of each of three separate experiments, r simple correla-
tion; R = multiple correlation.

Hunter a was calculated from R, G, B data in
Experiments 1 and 2, the correlations with vis-
ual ranking were decreased to an appreciable
extent. Hunter a is a common function used to
describe green vegetable color, since it is a
measure of greenness (-a, green; + a, red); ob-
viously, in this transformation from other sys-
tems, its value in measuring visual judgments
has been lost.
In many cases the incorporation of two tri-
stimulus values into one function will provide
a single measure of color that simulates visual
judgment. This has been termed reduction of
data, and the results of such treatments are
shown in Table 5.
In the case of G/R and X/Y, the result was
to produce correlations with visual ranking
which were not as good as the single tristimulus
values alone. Obviously, there is no advantage
to using this type of reduction for the samples
used in this study.
Single Adams tristimulus values are also pre-
sented. These values do not strictly fall under
the definition of "reduced data", but they are
included in this table, since there is no instru-
mental method to measure Adams L, a, b. Only
256 CRC Critical Reviews in Food Science and Nutrition

TABLE 4

Correlations of Visual Rankings vs. Colormaster® R.O.B Obtained from a Colormaster Model V and as Calcu-
lated front Other Systems

Correlation coefficient
R,G,B calculated from GE
Recording
R,G,B read-out from R,G,B calculated from Spectrophotometer®
Correlations of Colormaster Model V Hunter D25® L,a,b X.Y.Z
Visual Ranking
with: I II III I II III I II III

r
Colormaster G 0.960 0.980 0.986 0.957 0.956 0.989 0.956 0.980 0.987
R 0.956 0.980 0.987 0.941 0.945 0.992 0.939 0.980 0.991
B 0.879 0.968 0.979 0.848 0.686 0.975 0.933 0.878 0.907

R,G, 0.975 0.984 0.989 0.986 0.987 0.996 0.980 0.988 0.998
B

Note: I, II, and III refer to the average of duplicates of each of three separate experiments, r = simple correlation;
R = multiple correlation.

TABLE 5

Correlations of Visual Rankings vs. Reduced Data for Processed Spinach Purees

Correlation coefficient

Correlation, Visual Ranking with: I II III

Colormaster® G/R 0.809 0.744 0.937

Adams L (from Colormaster data) 0.965 0.986 0.925
a (from Colormaster data) 0.769 0.491 0.981
b (from Colormaster data) 0.968 0.968 0.993
a/b (from Colormaster data) 0.772 0.584 0.901
tan"1 a/b (from Colormaster data) 0.772 0.580 0.991
(a2 + b')"' (from Colormaster data) 0.969 0.969 0.992
Hunter a/b 0.918 0.995 0.989
tan"' a/b 0.919 0.995 0.989
(a2 + b')1" 0.976 0.840 0.979
vGE X/Y 0.841 0.857 0.964

Note: r = simple correlation.

Adams b and L showed a good correlation with
visual judgment. It was known from other mea-
surements in this study that L (lightness or
darkness) was a good measure of visual prefer-
ence for the samples under study; however, b
was not as good. This indicates that the Adams
transformation caused b to correlate more di-
rectly with visual preference than it actually
should.
Adams a value showed a significantly poorer
correlation in two of the experiments than any
other functions associated with an a value.
Thus, the transformation lowered the correla-
tion of a, which is normally quite a good indi-
cation of visual preference for green food sam-
ples.
Adams a/b and tan' 1 a/b also showed low
correlations with visual judgment. The Adams
functions are valuable tools when used prop-
erly; however, they did not seem to produce

valid results with the samples used in this study.
Since they have been used widely, it should be
stressed that each individual type of colored
material should first be ranked visually to es-
tablish good correlation with Adams data be-
fore transformations are used to predict visual
preference. .
Adams (a2 + b2)1'2, which is a function of
chroma, showed high correlations with visual
rank and apparently has solved some of the
problems associated with the a value.
The Adams a, a/b, and tan"1 a/b correlated
well with visual judgments (Table 5) in only one
of three experiments. The series that showed
the poorest correlation (Experiment 2) had a to-
tal color difference between the first and last
samples of 4.2 as measured with D25 and cal-
culated by
AE = ( ALJ + Aa2 + A b ' ) " ]
The series with the intermediate correlation
(Experiment 1) had a AE of 6.0, whereas the
one that correlated most highly (Experiment 3)
had a AE of 6.2. Apparently, with spinach
puree, as the total color difference decreases the
validity of the transformation decreases. As
was expected, since Hunter a and b values were
high in correlation with visual ranking, the re-
duced data showed the same type of correla-
tion, with the exception of that for (a2 + b2)2'2,
which was slightly lower in one of the three ex-
periments.
Obviously, this study showed that there are
dangers in using the manipulation of color data
without first justifying it with experimental evi-
dence. There is no need to use such a detailed
screening procedure, as was used in this study,
in order to evaluate the applicability of a spe-
cific function for a specific food. However,
prior to using an untested function to define a
particular color in a food, it is essential to:
1. Understand the theoretical basis behind the
use of the function
2. Evaluate the function by use of a visual
panel prior to adapting it
3. Use the function only in the color range
tested and on samples which are physically
similar to those used in Number 2
Reduction of data is a powerful tool and one
that should be employed whenever it is appli-
cable. The major reason for its use is to reduce
December 1978 257
the number of parameters required to define a
quality attribute (i.e., color) so that control and
definition become simpler. However, simplicity
cannot be gained through a decrease in validity.
Specific Examples of Data Reduction
One of the most common ways in which
workers have attempted to reduce colorimetric
data has been the expression of a color differ-
ence between two colors by a single number.
This single number will express the distance in
three-dimensional space between two colors
and thus provide a nonvector specification,
since distance but not direction is defined. The
manner in which such a color difference is cal-
culated will depend not only on final usage but
largely on which color scales are used to meas-
ure the color. Previously, it was noted that
there are many color scales in existence, the
more common being the CIE X, Y, Z; the
Hunter L, a, b; and the G, R, B. The major
reason for the development of colorimetric
scales has been the desire to express color in a
visually uniform solid. Ideally, this would mean
that a color difference in the red region would
be the same as a color difference in the green
and blue region. This, however, was not the
case with the CIE system, and although im-
provements have been made, it is not the case
in any system today.
If the CIE system simulated the eye exactly,
then one would expect that equidistant colors
representing a circle in the Munsell system
would also form a circle in the CIE system. Un-
fortunately, when the tristimulus coefficients x
and y of the Munsell colors are calculated and
plotted on a chromaticity diagram, as in Figure
7, the resulting plot is not a perfect circle.
Many systems have been developed in an at-
tempt to improve the visual spacing of the CIE
system. One of the first was the Uniform Chro-
maticity Scale (UCS System) of Judd.41 Breck-
enridge and Schaub8 developed this into the rec-
tangular UCS system (RUCS). MacAdam52
developed the u, v system and later another sys-
tem (MacAdam51) for the calculation of small
color differences. The system developed by
Hunter" is closely related to the read-outs of
Hunter instruments.
Other systems, by various methods, have at-
tempted to simulate the visual spacing of the
Munsell system. A partial list might include the
Adams Chromatic Value System,1 the Omega
258 CRC Critical Reviews in Food Science and Nutrition

5GY
50

40 5YR

30

20
20 30 40 50

FIGURE 7. A portion of the CIE chromaticity diagram, showing the locations of

the Munsell colors at the ten major hues, Value 5, and Chroma 8.

Space of Moon and Spencer,54-55 the Saunder-

son-Milner Zeta Space," the Din System," the
Glasser System,24 and the Adjusted Hue System
of Billmeyer.5 None of these systems were com-
pletely successful, and there is some question as
to the real advantage of having visual uniform
spacing as long as small color differences can
be specified. Color-measuring,instruments, in
fact, are best used to predict color differences
between samples and are misused when they are
employed to measure the absolute color of a
sample.
The nonuniformity of the CIE system was in-
vestigated in detail by MacAdam 51 in 1943. He .2 -
had an observer make over 25,000 judgments
of least perceptible chromaticity differences. .1 -
Fortunately, the observer was typical, as later
observations confirmed the data. The data
which was obtained was used to plot least per-
ceptible differences and resulted in what came FIGURE 8. The MacAdam ellipses plotted on a CIE x, y
to be known as the MacAdam ellipses as illus- chromaticity diagram. The diameter of the ellipses in ten
trated in Figure 8. Every radius on every ellipse times the value of the just noticeable differences of chro-
has the same visual magnitude, but obviously maticity from 25 typical standard colors.
they differ on the x, y plot. More simply, this
confirms the fact that equal visual distances are
not equal in every part of the CIE solid. More tempts to obtain visual uniformity up to 1973
detail on the historical background of the at- may be found in Francis and Clydesdale.21

More recently, a CIE technical committee
has proposed a study of two color spaces and
associated color-difference formula to replace
the CIE 1964 U*V*W* space which was origi-
nally developed by Wyszecki" as follows:
W* = 2 5 Y " 3 - 1 7
U* = 1 3 W * ( u - u 0 )
V* = 13W* (v - v 0 )
where u0 and v0 are the chromaticity coordi-
nates of the light source. For CIE source C:
u0 = 0.2009 and v0 = 0.3073
4X _ 4x
X + 15Y + 3E -2x+12y
6Y 6y
X + 1SY + 3Z - Z x + 12y + 3
w = 1 - U- V
and the color difference
AE = ([AU*] 2 + [ AV*] 2 + [ A W * 1 2 ) " 2
The new spaces are as follows.70 The first color
space is a modification of the CIE 1964
(U*V*W*) space. It is produced by plotting in
rectangular coordinates the quantities L*, u*,
v* defined by:
L* = 25 (100Y/Y0)'/J - 16, 1 < Y < 100
u* = 13L*(u'-uo0
v* = 13L*(v'-vo0
with
4X 9Y
X + 15Y + 3Z X + 15Y + 3Z
4X
0 ... . _
V
X o + 15Y 0 + 3Z 0 O
The tristimulus values Xo, Yo, Zo define the
color of the nominally white object-color stim-
ulus. Usually, the white object-color stimulus is
given by the spectral radiant power of one of
the CIE standard illuminants, for example, D65
or A, reflected into the observer's eye by the
perfect reflecting diffuser. Under these condi-
December 1978 259
tions, Xo, Yo, Zo are the tristimulus values of
the standard illuminant with Yo equal to 100.
The total difference A ECIE (L*u*v*) between
two colors, each given in terms of L*, u*, v* is
calculated from:
A E C , E ( L * u V ) = [( A L*)2 + ( A u*)2 + (A v * ) ] ' ' 2
The second uniform color space is a simpli-
fied version of the Adams-Nickerson (AN)
space. It is produced by plotting in rectangular
coordinates the quantities L*, a*, b* defined
by:
L* = 25 (100Y/Y 0 )' / l - 16, 1 < Y < 100
a* = 5 0 0 ( [ X / X 0 ] " 2 - [Y/Y 0 ]" 2 )
b* = 2 0 0 ( [ Y / Y 0 ] " 2 - [ Z / Z o j " 2 )
The tristimulus values Xo, Yo, Zo define the
color of the nominally white object-color stim-
ulus. Usually, the white object-color stimulus is
given by the spectral radiant power of one of
the CIE standard illuminants, for example, D65
or A, reflected into the observer's eye by the
perfect reflecting diffuser. Under these condi-
tions, Xo, Yo, Zo are the tristimulus values of
the standard illuminant with Yo equal to 100.
The total difference A BCIE (L*a*b*j between
two colors, each given in terms of L*. a*, b* is
calculated from:
A ECIE(L*aV) ([AL*] 2
More recently in 1976, Pauli58 discussed an
extension of the second proposed space. Since
the coordinates on an approximately uniform
color space were defined70 by:
L* = li
a* = 500([X/Xo]»'3-(Y/Yo)"3J (1)
™0
3Z0 b* = 200[(Y/Y0)I"-(Z/Zo)"3l
with the restriction
Y/Yo > 0.01 (2)
X, Y, Z are the tristimulus values of a color and
Xo, Yo, Zo define the color of the nominally
white object-color stimulus with Yo equal to
100. In this color space, the total color differ-
260 CRC Critical Reviews in Food Science and Nutrition

ence E*o6 between two colors, each given in
terms of L*, a*, b*, is calculated from
AE* b =
The values of E*.b correlate with the color dif-
ferences E(AN40) calculated using the original
AN formula2 only if all three ratios X/X o , Y/
Yo, and Z/Z o , respectively, are greater than
0.01. For most object colors, the condition ex-
pressed in Equation 2 is satisfied. But Z/Z o of
deep yellows, oranges, and reds may assume
values below 0.01, causing marked deviations
of A E*ot from AE(AN40). Values of X/Yo and
X/Xo also below 0.01 are found if the transmis-
sion color of nearly opaque samples is mea-
sured.
It is desirable, therefore, to extend the CIE
recommendation to values of X/Xo, Y/Yo, and
Z/Z o down to 0.00 by linear extrapolation.
Writing the formulas:
L = 116/(Y/Y0)
a*= 500(/[X/X01-/[Y/Y01)
b* = 200(/[Y/Y0)-/[Z/Z0])
with the function f (q) defined as:
3
/ (q) = q " - — for q > 7
116
/ (q) = <*q for q < 7
(A Eot)** the total color difference calculated
with a and y of Equation 7.
In the case where:
(3)
X/Xo > 0.01, Y/Yo > 0.01, Z/Z o > 0.01 (8)
Equations 1 and 4 are identical.
However, as of this writing, these recommen-
dations have not been finalized.
Having discussed the historical background
and complexity of stating uniform color differ-
ences, it might be illuminating to consider one
of the more commonly used expressions for
color differences, E = ( L2 + a2 + b2), and its
uses and limitations.
Hunter31 has provided an excellent discussion
of color differences which is reproduced here
by permission of Hunter Associates Labora-
tory, Fairfax, VA.
A measured value of color difference is intended to be a
quantitative representation of the ease with which the dif-
ference between the colors of two specimens can be seen
with the eye. Numerical color differences may be derived
from measurements of two specimens using any of the color
scales shown in Figure 9. Since color is three dimensional, .
(4) color differences are widely represented in three dimensions
such as L, a, and b.
A measured color difference can also take the form of a
total color difference unit, often designated as E. These sin-
gle-number color difference values are derived by means of
algebraic expressions relating to the geometry of the differ-
ences between two specimens in three-dimensional space.
Figure 10 shows the relationship of two specimens in color
space and one of the ways the distance between is measured
to arrive at E, total color difference.
(5) ftELATIOKSHtPS OF COLOR SCALES

If we choose for the constant y the original SP£CTR0(*OTO*T£R

TMST1IU.US
REflECTOHETER

value and require the function (q) to be contin-

uous over the whole range, we have:
ADAMS
7 = 0.01 CHROMTIC
HUNTER
VALUES TABLES ADO oitta KAoirn,
COMPUTATION COMPUTATION
16 (6) GLASSER-
a = 100 ([0.01] " 3 - ) =7.751 REIU.ET
CUBE ROOT COMPUTATION
CIE
X.T.I DIREa READIW.
HUNTER
L.i.b
116 COMPUTATION

MACADAM HUKTER
But if we require both the function /[q) and its me COMPUTATItM OIRECT RCAOIIK. I'l'b"
COMPUTATION
first derivative to be continuous we get:
COWUTATIOI

7 =(_*£)» = 0.008856 OONHUKT

116 WVELEKCTH
CftAPHS
CIE
COMPUTATION
CIE U«V^(«
AMO PURITT

a = 1/3 ( H±) 2 = 7 . 7 8 7 (7) WAPHS

24
MUNSni
To distinguish these two cases, we denote by (A REMIT ATIOK

Ea»)* the total color difference calculated with

the narameters a and v of Eauation 6, and by FIGURE 9. Relationship of color scales.

AL
AE = /AL 2 +Aa 2 +Ab 2
FIGURE 10. The calculation of color difference.
Major Uses of Color-difference Measurements
Precise analysis of small color difference is
used in research and technology to:
1. Compare industrial production specimens
with standards to test these specimens for
conformity to color-tolerance specifica-
tions. This is probably the most widely
used type of measurement in all appearance
technology.
2. Determine nature and magnitude of color
adjustment required to convert an unac-
ceptable color to a product having accept-
able color.
3. Measure color changes of products result-
ing from their exposure to weather, light,
laundry, or other real or simulated usage.
The major advantages of instrumental color
difference measurements are that they are
quantitative, repeatable from time to time, and
reproducible from one laboratory to another.
The disadvantages are that they provide less
than perfect correlation with visual estimates,
especially if a single scale is used, and that there
are a variety of different color-difference scales
resulting in much confusion.
Total Color-difference Scales
The confusion arises in the use of total color-
difference units in industry, since the psycho-
December 1978 261
The concept of threshold (identifying the
smallest differences perceivable) difference was
based by MacAdam on his original 1942 chro-
maticity discrimination experiments. The unit
selected was intended to represent the smallest
perceptible difference. MacAdam's experiment
involved chromaticity differences only, but the
third dimension of lightness was added by
Brown and MacAdam some years later. Al-
though the equations defining the MacAdam
perceptibility ellipses are readily handled by
modern computers, this was not so earlier.
Consequently, graphical techniques were devel-
oped by Simon and Goodwin in 1957, and Fos-
ter in 1966, to aid in computation of color dif-
ference in MacAdam units from CIE Y, X, Z
data.
The use of MacAdam total color-difference
units is further complicated by a series of mod-
ifications made in the 1960s by Friele, followed
by MacAdam and Chickering. Friele used Mull-
ers theory to assign equations yielding oppo-
nent-color scales to the original MacAdam
data. Consequently, the term "MacAdam" can
now be used in reference to any one of several
slightly different MacAdam color-difference
units.
logical basis for the visual estimation of the
color difference is not the same in all cases.
Two color difference concepts are in general
use in industry today:
1. The largest difference considered to be ac-
ceptable as a commercial match (maximum
acceptable).
2. The smallest difference capable of being
perceived (minimum perceptible.
The NBS unit of total color difference origi-
nally represented the maximum acceptable dif-
ference between two colors in a commercial
match in a dye house in 1939. The NBS unit
generally referred to today is the 1942 Hunter
rectangular coordinate version of Judd's equa-
tion for the original NBS unit. This unit has
been used in terms of a number of the oppo-
nent-colors scales. In 1950, Nickerson proposed
the use of the Adams opponent-colors scales
for the measurement of color differences and,
by applying appropriate factors, made the mag-
nitude of the Adams unit as close as possible to
that of the NBS unit of color difference.
262 CRC Critical Reviews in Food Science and Nutrition
Psychological Validity of Color-difference
Scales
The validity of total color difference as a
meaningful measure is under question from
several points of view. If it is to be used to
measure acceptability in commerce, there is
doubt that acceptability correlates with percep-
tibility, specifically with threshold differences.
Dr. Friele, a leader in the field, has raised the
question of whether there is a mechanism in the
brain for taking the square root of the sum of
the squares and thus giving measurements cor-
relating with distance in color space as repre-
sented by Figure 10. Further, it is an established
fact that estimates of a magnitude of color dif-
ference change significantly with changes of ob-
serving conditions such as surroundings, level
of illumination, and optical character of speci-
men. Thus, one should be warned to question
the universal applicability of measures of color
difference to commercial applications of color
acceptability and the like. Nevertheless, when
used with caution, measurements of total color
difference are valuable since they are quantita-
tive, repeatable, and reproducible.
Conversions Among Color-difference Scales
There are significant quantitative differences
between the units of total color difference char-
acteristic of the various scales. In the past, there
have been some suggestions that one could use
a simple multiplying factor to convert color dif-
ference units in one system to those in another
system. One such suggestion is that if the num-
ber of NBS (commercially acceptable) color-
difference units is multiplied by 3 it will approx-
imate the number of MacAdam (just percepti-
ble) units that would define the same color dif-
ference. Actually, the multiplication factor
must change with the position and orientation
of the color difference in color space. A one to
three relationship between an NBS unit and a
MacAdam unit of color difference may obtain
at one point and direction in color space, and
not hold true at some other point or direction.
It is generally not possible to convert from one
system to another with a single conversion fac-
tor.
Conversions of measured values of total
color difference from one set of scales to an-
other can be made when one has the values for
the colors, not just the color differences in-
volved. Transformations can then be made by
converting the color coordinates by way of CIE
specifications from one set of coordinates to
the other and from there to the color difference
unit. High computational precision is necessary
where the color differences involved are small
and near the limit of commercial acceptability.
To aid in these conversions, there are available
graphs, differential formulas, and conversion
factors (for single colors). The use of com-
puters has considerably lessened the labor in-
volved, and has opened up new possibilities in
this area.
Using Color Differences in Practice
Most of the published discussion is con-
cerned with the use of total or single-number
color-difference units. As noted, color meas-
urement scales can be used to give color differ-
ence by subtracting one value from another in
each of the three dimensions. When one looks
at two samples in considering the visual differ-
ence between them, one most naturally thinks
in these three-dimensional terms, i.e., "this is
lighter but is also redder and yellower, etc."
Therefore, color differences in terms of A L, A
a, A b are often more useful than a total color
difference in indicating what to do about the
perceived difference.
The significance of the total color-difference
unit such as A E is less readily apparent. Where
sometimes reductions to a single number can be
more convenient as a unit for describing color
tolerances, it can also be more confusing, since
a single number tells only the size of the differ-
ence, rather than the "direction" or nature of
the difference from standard. Because of the in-
complete information given by a single number,
the trend in recent years has been to examine
the character in each color difference by record-
ing its components separately. For use with
three-component color-difference measure-
ments, a complete color specification would
contain nine numbers: the three numbers that
describe the desired color, and the six numbers
that describe the individual plus and minus tol-
erances.
Color-difference tolerances should be de-
signed as boundaries in color space within
which acceptable colors must fall. The bound-
aries do not necessarily correlate with percepti-
bility of difference but rather with the limits of

acceptability. The standard color, furthermore,
may not lie at the center of the circle, or ellipse,
described by the tolerance boundaries. When
color-difference measurements are used to
identify acceptible product colors, the mea-
sured magnitudes of permitted color difference
from standard very often do not compare to
perceptibility of these differences. For example,
where subsequent yellowing may occur, the tol-
erance for the yellow-blue dimension might be
+ 0.1, -0.8 units, indicating that virtually any
yellowing at all is unacceptable.
Further material on the use of color differ-
ences for setting specifications and tolerances
may be found in Francis and Clydesdale21 and
Hunter.30
Reduction of Data to a Function of Hue
Hue is the single color function which de-
scribes what most people think of as color.
That is, at a constant degree of lightness and
saturation, hue describes the shift from red to
orange to yellow to blue to violet. Also, it is a
known psychological fact that hue often corre-
lates with visual acceptance of many foods. For
these reasons, hue has often been chosen as a
single function to represent color, and, mathe-.
matically, colorimetric data has been reduced
to one function which describes hue. A case
may be made for the assumption that the ratio
of any one part of the spectrum to any other
part is a function of hue. Thus, a shift in color
may be described by a shift in such a single
number ratio. For this reason, a ratio of two
wavelengths or ratio of any tristimulus values
(such as X/Y, X/Z, G/R, a/b, etc.) may be ar-
gued to represent hue.
However, in a geometric sense, not all the
above ratios can be defended. The geometry of
the Hunter L, a, b system is such, however, that
both geometrically and conceptually certain
functions satisfy a specification of hue. For this
reason, the function a/b became a widely used
single-number function to describe hue in foods
many years ago.
If we consider a typical Hunter a, b plot at
constant lightness (as shown in Figure 11), it
can be seen that the point B represents an or-
ange color with an a/b ratio equal to unity. The
fraction + a/ + b is actually the tangent of angle
OBA, which in this case would be equal to 45°.
As the hue becomes more yellow and moves to-
December 1978 263
ward point C, the tangent of OBA (OBA =
BOC) approaches zero and the ratio BC/CO
(a/b) approaches zero. In the second quadrant,
the ratio a/b would represent the yellow-green
colors. The point E, represented by angle COE,
would have a tangent - a / + b, which would be
equal to —1. The point F, represented by angle
COF, would have its tangent equal to infinity.
In the third quandrant, the blue-green color H,
represented by angle HOK, would have its tan-
gent —a/—b equal to + 1. Similarly, the red-
blue color M, represented by angle KOM,
would have its tangent + a/—b equal to — 1.
It is obvious from Figure 11 that the a/b ratio
is a tangential function not a linear one. If the
ratio is in the vicinity of 1, where the angle is
approximately 45°, the departures from linear-
ity are not too great. It was in this area with
tomatoes that the a/b usage becomes firmly en-
trenched. However, if one attempts to use the
a/b ratio in a portion of the color solid where
it approaches infinity, the concept obviously
breaks down. In such cases, it is preferable to
use the actual angle that the point makes with
the vertical axis, rather than the ratio itself. A
rough rule of thumb would be that if the a/b
ratio is within 0.2 and 2.0, it probably is satis-
factory by itself. If the ratio is outside these
limits, the angle should be used, not the ratio.
The first use of the hue angle was suggested
by Judd in 1950 in an NBS circular.40 This con-
cept was probably known to colorists at the
time but was not widely known to food scien-
E c B
1
\
7
u /
+b
1 \ >
1
\ /
f\ G-a >/ IA
IN
1
\
Y
FIGURE 11. Hunter, a, b plot showing how the a/b ratio
and the tan1 a/b values vary with the color.
264 CRC Critical Reviews in Food Science and Nutrition

tists where the field of colorimetry was just be-
ginning. This hue angle is the angle 8 shown in
Figure 12. Shortly, thereafter in 1952 Francis20
published work with apples utilizing tan"'a/b,
which is the angle 90-0 in Figure 12. In the case
of apples, this was a sound concept since nega-
tive values of a in the yellow-green quadrant
would produce angular values from 0 to more
negative values, thus providing negative values
to describe greenness. Similarly, positive values
of a in the yellow-red quadrant produce angular
values of o to more positive to describe redness.
Thus, positive angular values describe increas-
ing redness and negative angular values de-
scribe increasing greenness for apples. This
concept breaks down in the blue-green quad-
rant (-a/-b) which would be positive; how-
ever, practically no foods are blue-green in
color.18
There has been some controversy in the use
of angular functions which describe hue culmi-
nating in two recent publications by Little46 and
Francis.18
Since these angular functions represent one
of the best and most widely used data reduction
systems, it would be worth while to look at
these two viewpoints in more detail. In Little's46
words:
Although many positive advances have been made, de-
spite the rather haphazard approach to the formulation of
principles of colorimetry of foods, misapplications or un-
conventional interpretations of basic concepts can and have
occurred; and once incorporated into the literature, they
then tend to gain acceptance through repetition.
I am concerned here not with exploring differences of
opinion that can be discussed on a philosophical basis, but
with a particular difference of opinion based on a conven-
tional vs. a nonconventional interpretation of a well-estab-
lished mathematical relation. This has to do with expressing
colorimetric data in the Hunter L, a, b system to provide
information that can readily be conceptualized as hue. One
does not have to search very hard for references where hue
is expressed as a/b with the hue angle unconventionally de-
fined as tan"1 a/b. It is not my purpose, to catalog the ref-
erences; it is far more interesting to document the history
of the origin and development of this misconception.

YELLOW
I.STOGr)

- < • - .

GREEN. RED
(180*. 3.14160 i -o (360", 6.28321)
\
\
\

\
-b

• - * -

BLUE
(270*. 4.7I24f)

FIGURE 12. Depiction of hue angle and saturation index on the Hunter a,b dia-
gram. Point P, located at (a,b) in rectangular coordinates and (0,r) in polar coordi-
nates; 9, (hue angle) = tan"1 b/a; Saturation index = (aJ + b 1 )" 1 . The broken curve
with arrows pointing in counterclockwise direction indicates the direction of change
in hue from red to yellow as the slope increases from 0 to °», from yellow to green,
as the slope decreases from °° to 0, etc. (Reprinted with permission from Little, A.
C , J. Food Sci., 40, 410, 1975. Copyright © by Institute of Food Technologists.)

Further, she describes the historical evolve-
ment of these functions including the develop-
ment of the indices of lightness, hue, and satu-
ration as follows:
1. Lightness index: L = Y"2
2. Hue angle: H = tan"1 b/a
3. Saturation index: S = (a2 + b 2 )" 2
Again quoting Little:46
Consider, now, the a,b diagram shown in Figure 1. The
point Pj, represents the plot of sample S, designated (a,b)
in rectangular or Cartesian coordinates, and (r,0) in polar
coordinates. According to accepted tenets of analytical ge-
ometry, the superimposition of polar coordinates on the
Cartesian coordinate grid places the pole at the origin and
the polar axis identical with the x-axis (called the a-axis in
this special case). Thus, for point P., the following relations
pertain between the Cartesian and the polar coordinates:
r = (a2 + b 2 ) " 2
tan 6 = b/a
6 = tan" 1 b/a
and to determine the quadrant of 8:
© = cos-i •
(a» + b 2 ) " 2
It is immediately apparent that 8 and r are identical with
hue angle, H, and saturation index, S, respectively, and,
thus, that Hunter was simply making use of well-known
geometric relations in proposing these indices.
How, then, can we account for the persistent upside
down relationship which in fact describes the tangent of 90°
— 8, the angle complementary to 8, i.e., cot 6? I think the
answer is to be found in the papers by Robinson et a l . " "
where a/b was proposed as a color index in Hunter Meter
measurements for tomatoes, with a/b shown to be directly
related to dominant wavelength. This concept apparently
gained immediate acceptance, and in the 1954 Symposium
proceedings, Color in Foods," reference was made repeat-
edly to hue defined as a/b. On page 53 of the proceedings,
one reads that
"a £ /bi is related to the slope of a straight line which
would pass through the origin (a,. = 0, bL = 0)."
The key words here are related to. Delete them and the
statement changes significantly. If the statement had read
that
"aL/bL is the reciprocal of the slope . . . "
these comments might be unnecessary today. One must sur-
mise that a/b was considered preferable for tomatoes for
the values to increase with increasing redness.
It should be apparent, however, that we are dealing not
with linear but with trigonometric functions, i.e., the tan-
gent of the hue angle, b/a, and its reciprocal, the cotangent
a/b, which vary from 0 to 1 to » and from °° to 1 to 0,
December 1978 265
respectively, as the angle increases from 0 to rr/4 to rr/2
radians (0° to45° to 90°).
It was by chance alone that the a/b ratios for tomato
products fell within a range of steadily climbing but not
rapidly accelerating values for a/b, ca. 1.4 to 2.0. Francis20
recognized the pitfalls inherent in expressing hue range in
terms of a/b for other products, such as apples, which
cover a wide range in color. He therefore suggested the
adoption of an angular function, tan"1 a/b, which he de-
scribed as the angle contained between the constant hue line
and the vertical axis. This means, in fact, that he proposed
a novel and unorthodox relation between rectangular and
polar coordinates with the pole at the origin, but the polar
axis now coinciding with the b-axis, and the angle of rota-
tion reversed from counter-clockwise to clockwise. Thus,
he arbitrarily changed a well-understood, universally ac-
cepted coordinate system, merely to maintain a positive re-
lation between increasing redness and the numerically ex-
pressed hue function.
Further, and in conclusion Little46 states:
Although Hunter1* now acknowledges the ubiquitous na-
ture of the a/b ratio by defining the hue angle, 9 = cot"1
a/b, from a conceptual point of view, the b/a ratio is the
expression of choice. It bears the same relationship to the
a,b diagram that the line of constant dominant wavelength
does to the CIE x,y chromaticity diagram. I like to think
of the function b/a as a hand sweeping counter-clockwise
on a dial, starting at Or (red), to nr/2 (yellow), to nr (green),
to 3nr/2 (blue) and at 2nr back to red with its length (a2 +
b2)1'2 providing a measure of saturation. This measure is,
of course, the correlate of excitation purity obtained from
the x,y diagram. Let us stay with conventional relations,
they work very well and serve to reduce ambiguity.
I believe that some confusion has resulted
from these points of view. Certainly some am-
biguity may develop if terms are not strictly de-
fined and understood. However, that is the case
when using any colorimetric function. For in-
stance, the work with apples, as explained ear-
lier, utilized a fortuitous positive-negative con-
cept. Still, if ambiguity can be lessened, I would
agree with Dr. Little and recommend the hue
angle as cot"'a/b.
Likewise, Francis" has stated:
It was fortuitous that the a/b function was appropriate for
tomatoes. This is merely because nearly all colors of tomato
products are in the area where 8 = 45 to 60. The a/b ratio
at 8 = 0 is zero, at 8 = 45 is 1, and at 8 = 90 is infinity.
Obviously, values approaching infinity are useless. We had
a rule of thumb that a/b can be used as a rough index of
hue only between 0.2 and 2. Otherwise, an angular function
should be used. With many foodstuffs, the a/b ratios lie
outside the above range; consequently, the tan"1 a/b has
become well entrenched in the food literature. However, in
the last few years, there seems to be an increasing tendency
266 CRC Critical Reviews in Food Science and Nutrition

to use the hue angle (0) concept. When the tan"' a/b con- the Hunter solid. If, in fact, one of these func-
cept was originally published in 1952, I was under the tions should be found to correlate better with
impression that this was an original concept. It was only
several years later that I discovered that a similar concept,
visual rank than any other function, there
namely the hue angle, had been published in the NBS cir- would be no harm in using it as long as it was
cular.™ The content of this circular may have been well understood that it did not describe chroma.
known to colorimetrists at that time but it was not well Generally, it is best to use functions which have
known to food scientists. a theoretical basis for their use; however, when
There is no doubt that the use of two systems to describe necessary, this rule may be stretched.
hue in the Hunter system has created some confusion. The
hue angle as described by Hunter is probably the simpler Another function which is often used to sim-
of the two to understand and has no technical drawbacks ulate chroma is the "excitation purity" or "pu-
even in the blue-green quadrant. For the sake of uniform- rity" function of the CIE system. This function
ity, it may well be that we should recommend that the hue may best be explained by a short review of the
angle, or 6, should be used in the future.
CIE chromaticity diagram or spectrum locus
shown in Figure 13.
Reduction of Data to a Function of Chroma
Chroma or saturation may be defined45 as the In addition to defining the area over which
proportion of chromatic content in the total physical stimuli must be located, the spectrum
perception; it is also the degree of difference locus provides a greater meaning to the color
from the neutral or gray of the same lightness quality associated with any point within it.
value. Saturation scales, according to these two Thus, in Figure 13, by drawing a straight line
definitions, differ from each other. The Mun- from the illuminant chromaticity point (CIE il-
sell chroma scale is based on the second defini- luminant C) through the point representing the
tion. Another way to picture chroma is to think chromaticity of the sample color to intersect the
of a color solid as an orange as described ear- spectrum locus at A, we know from the geome-
lier. In this case, the distance from the core to try of the chart that a suitable mixture of A and
the skin could represent Munsell chroma or in- the illuminant will match the sample color. If
tensity of color. the hues of the spectrum are familiar to the co-
lorimetrist, the value of A will approximate the
Due to the nonrectangular shape of many of hue of the sample, while the distance along the
the color solids, it is difficult to describe line from the illuminant towards A through the
chroma in geometric terms such that one single sample approximates its chroma or intensity.
function may be used. In fact, the only system The point A is known as the dominant w'ave-
which lends itself to this calculation is the
Hunter L, a, b system. Referring to Figure 12,
it may be seen that the hypotenuse of the trian-
gle may be described by the simple geometric 520
function (a2 + b 2 )" 2 . Upon further inspection,
it becomes obvious that the hypotenuse of the
triangle in fact describes the degree of differ-
ence from the neutral or gray of the same light-
ness value which is the definition of saturation
or chroma. The chroma function, represented
by (a2 + b 2 )" 2 , has been used successfully to
describe the color of an object. In fact, Clydes-
dale and Francis12 showed that this function
calculated from both Hunter and Colormaster
data provided an excellent correlation with vis-
ual ranking for spinach puree (Table 5). At
times in the literature, other functions such as
(G2 + R 2 )" 2 , (X2 + Y 2 )"\ etc. are used to sim-
ulate chroma. It should be stressed that these
functions do not geometrically provide a meas-
ure of chroma, since their respective color sol- FIGURE 13. A CIE chromaticity diagram showing dom-
ids cannot be defined in as regular a manner as inant wavelength, A, complementary wavelength and pu-
rity.

length of the sample color, and the ratio of the
distance (a) from the illuminant to the sample
over the distance (a + b) from the illuminant
to the spectrum locus at the dominant wave-
length is known as the purity (excitation purity)
of the color of the sample.
This description of dominant wavelength and
purity relates only to spectral colors and not to
the nonspectral colors. In the case of nonspec-
tral colors, a complementary wavelength (Ac) is
specified rather than a dominant wavelength.
This complementary wavelength is obtained by
drawing a line from the sample chromaticity
through the illuminant and beyond until it
reaches the spectrum locus at Xc (see Figure 13).
Also, for nonspectral colors, the line used to
compute purity is not a line of dominant wave-
length, but one extending from the chromatic-
ity point of the illuminant through the chro-
maticity point of the sample; it intersects the
chromaticity points of the long- and shortwave
ends of the spectrum. Therefore, the purity of
a nonspectral color is the ratio of the distance
(e) from the illuminant to the sample divided
by the distance (e + d) from the illuminant to
the nonspectral boundary of the spectrum lo-
cus.
Another type of function occasionally used is
colorimetric purity. This is, by definition, the
ratio of the luminance of the spectrally pure
component to the luminance of the matching
mixture. This function is not too important
from our point of view and should not be con-
fused with excitation purity, normally referred
to simply as purity.
The functions defined above are analogous
to certain functions which have been discussed
previously. However, they are not the same.
That is, hue is analogous to dominant wave-
length, chroma is analogous to purity, and
value is analogous to luminosity. However, it
should be stressed again that these are analogies
and are not exactly the same: they behave dif-
ferently in different sections of the color solid.
Nevertheless, purity may be used as an
expression of chroma to describe color if good
correlations with visual rank are obtained and
the system is understood.
Both purity and dominant wavelength has
been used to describe visual assessment in the
literature. This is a valid technique, but the au-
thor believes that a more reasonable approach
December 1978 267
is to mathematically reduce data directly as in
cor'a/b or (a2+ b 2 )" 2 rather than plot data
and introduce a greater potential for error.
For instance, in a study of rose wines, Josyln
and Little39 noted that when the chromaticity
coordinates x and y were calculated and con-
verted in terms of dominant wavelength and
purity that the change in color of wines on stor-
age could be evaluated. Further, they found
that a decrease in dominant wavelength indi-
cated pigment loss and an increase in purity in-
dicated browning.
However, as pointed out by Johnson et al.,37
the design of the experiment carried out by Jos-
lyn and Little3' did not allow for a statistical
evaluation of the results, as only two samples
of each wine were rated subjectively. In fact, in
the study by Johnson et al.,37 it was shown that,
over a wide range of pigment degradation in
cranberry beverages, dominant wavelength did
not -provide an accurate description of pre-
dicted visual ranks. However, results may de-
pend on the type of product being examined.
This work37 emphasized the need for analysis
of the efficacy of any colorimetric function and
an understanding of the theoretical basis for the
measurement prior to adopting the function for
use.
Data Reduction Systems Involving Kubelka-
Munk Analysis
The Kubelka-Munk colorant layer concept
was developed from turbid-medium theory. A
turbid medium is a substance which both scat-
ters (causes a change in the direction of travel)
and absorbs (converts to heat) part of the light
falling on it. This definition includes all trans-
lucent and opaque materials.3
As early as 1905, an astronomer named
Schuster introduced the concept that the pas-
sage of light through a turbid medium may be
represented by two diffuse fluxes, traveling in
opposite directions, each losing some of its light
to the other by scattering in a backward direc-
tion.3 In 1931, Kubelka and Munk published
what is probably the best-known and most used
treatment of turbid-medium theory. Their
treatment uses Schuster's concept of two dif-
fuse fluxes and is based on the "infinite-sheet"
approximation, which means that the sample is
assumed to be a sheet with plane parallel faces,
infinitely large but of finite thickness. Thus, the
268 CRC Critical Reviews in Food Science and Nutrition
effects of the edges can be neglected.3 The infi-
nite layer under consideration is small com-
pared to the total colorant layer but large com-
pared to the diameter of the pigment particles.
Kubelka and Munk assumed that the optical
properties of the materials could be described
(at any wavelength of light) by a scattering
coefficient S and an absorption coefficient K.
These two coefficients describe the amounts of
light involved in scattering and absorption per
unit length of travel of the light beams. As each
flux travels a small distance inside the sample,
it is assumed that each loses an amount of light
proportional to its own energy and to K by ab-
sorption, and an amount proportional to its
own energy and to S by scattering in a back-
ward direction. At the same time, each flux
gains an amount proportional to S and the en-
ergy of the second flux by the backward scatter-
ing of the other flux. These conditions can be
expressed mathematically by differential equa-
tions. It is necessary to make additional as-
sumptions to describe the sample — for exam-
ple, that S and K are the same throughout the
entire thickness. The assumption that all the
light is completely diffuse both within and out-
side the sample boundaries is the most limiting.
In spite of these assumptions, the Kubelka-
Munk theory and equations have found wide-
spread use based on approximate, if not exact,
description of turbid media.3-4*
Many modifications of the Kubelka-Munk
theory have been made since its introduction.
However, for opaque samples, the differences
between the Kubelka-Munk treatment and the
more extensive theories have been shown to be
small.2 One of the difficulties in applying more
complex theories is the evaluation of the reflec-
tion coefficients describing what happens at the
boundary of a real sample. It is especially dif-
ficult to measure the fraction of the light re-
flected when a diffuse beam hits the surface
from inside.3
Kubelka-Munk analysis has been used widely
in the paint industry for color-matching prob-
lems. It has been shown for opaque samples
that the ratio of light absorption (K) to light
scattering (S), at a given wavelength, is propor-
tional to the concentration of colorant in a sam-
ple, as described by the following equation:
K/S = kC
where K = light absorbed; S = light scattered;
C = concentration of colorant; and k = a con-
stant.
Further, the value of K/S has been shown to be
related to reflectance at a given wavelength by
the following equation:
K/S = (1 - R«,)2 /2R,»
where R<» = reflectance of an infinitely thick
sample.
In practice, it is not necessary to solve this
equation, since tables are available which tabu-
late K/S vs. R°° at a given wavelength, making
it very convenient to obtain such values.
This type of analysis has been used widely
and effectively in the paint industry for many
years but is only beginning to be used in foods.
Some of the problems have been reviewed by
Francis and Clydesdale21 (cited in Little)48 as
follows:
In the case of a translucent or turbid food material, we
are interested in the manner in which color is perceived by
the eye. This perception is dependent upon what happens
to the light incident on the surface of the sample. Such light
is reflected, scattered, absorbed, transmitted, and reflected
from layers below the surface. With classical spectropho-
tometry or reflectometry, a measure of the energy which
reaches the photosensitive element is obtained from the tur-
bid material. However, light is scattered and lost, and the
energy measured is not a true measure of that which reaches
the eye. The instrument cannot evaluate the composite vis-
ual appearance, whereas the eye does not distinguish be-
tween the relative contributions of light scattering and ab-
sorption.
The Kubelka-Munk colorant layer concept
has been proposed as a solution to these prob-
lems because it takes into account reflection,
transmission, and absorption from subsurface
layers as related to the internal scattering of
light.
A new problem now results involving the ma-
nipulations of the translucent food for meas-
urement and the mathematical treatment of
these data according to Kubelka-Munk theory.
The sample generally is placed in an optical
glass cell, forming a translucent layer, and three
instrumental measurements are taken: R, Ro,
and R,. R is the reflectance of the layer with a
white background. Ro is the reflectance of the
layer with a black background. Rg is the reflec-
tance of the white background used behind the

sample to obtain R. From these values various
quantities may be calculated:
a = (S + K/S) = 1/2 (R + [ Ro -R + Rg] /RQRg)
b = (a2-I)"2
Re = a-b
R00 is that reflectance creating the visual impact
the eye would perceive when viewing an infi-
nitely thick layer. Having calculated R°°, sev-
eral other functions which describe the compos-
ite effect of scattering and absorption may be
calculated:
K/S = ( l - R ^ ^ R ^
where K = the coefficient of absorption, S =
the coefficient of scattering, and the total func-
tion K/S represents composite effects. From
the foregoing, it may be seen that:
K = S ( a - 1)
with S derived from the expression:
SX = 1/b Arctgh([l-aRo]/bRo)
where X is the thickness of the layer and Arctgh
is the inverse hyperbolic cotangent. These solu-
tions are discussed in more detail by Judd and
Wyszecki.42
From these quantities, Ti, the internal trans-
mittance of a layer, may also be calculated
from the formula:
T; = (R - Ro) (l/Rg - Ro)
Having calculated T, in this manner, a function
can be obtained for turbid media which can be
handled in a manner comparable to the absorb-
ance of a clear solution. This function is ob-
tained by taking the logarithm of the reciprocal
of T,-, i.e., log 1/T,. This function represents a
measurement directly proportional to changes
in turbid media and therefore, is analogous to
optical density (OD) in true solutions; that is,
in true solutions the Beer-Lambert law, which
relates OD directly to concentration, applies. In
this case, log 1/T, is directly related to total vis-
ual impact or the composite effects of scatter-
ing and absorption.
Mackinney et al.53 and more recently Huang
et al. 2 8 " and Gullet"-27 have shown that inter-
esting results for the measurement of translu-
December 1978 269
cent food materials may be obtained by substi-
tuting tristimulus values for reflectance values
in the Kubelka-Munk equations. In place of R,
Ro, and R, for a specified wavelength, for ex-
ample, tristimulus values (such as X, Y, Z, or
L, a, b, etc.) would be substituted. Values of
X, Xo, and X, (analogous to R, Ro, and R,) are
obtained for a layer of sample with a white
background, a layer of sample with a black
background, and for the white background
alone. These may be substituted as follows:
a = 1/2 (X + [Xo - X + Xg] /X o X g )
= (a5-I)1'2
b
X«, = a - b
KS = (1 - X ^ ' ^ X ^
Similarly, K/S values may be calculated from
Y, Yo, and Y,; Z, Zo, and Z,; L, Lo, and L, or
for any other tristimulus value.
As well as calculating K/S values in this man-
ner, Ti, the internal transmittance of the layer,
may also be calculated, where:
T^ = (X-Xo)(l/Xg-Xo)
Similarly, Tiy and Ti,, Tu, etc. may also be cal-
culated. The log of the reciprocal of these Ti
values may be calculated, providing a function
based on tristimulus values which can be han-
dled in a manner comparable with the true ab-
sorbance of a clear solution.
Although the preceding mathematical treat-
ment might seem laborious, computer pro-
grams are available to make these manipula-
tions. Also, it should be remembered that once
a single function has been found which statisti-
cally correlates with visual judgments for a
given product, then other functions and manip-
ulations may be ignored. A further, more de-
tailed discussion of Kubelka-Munk theory in
foods may be found in Francis and Clydes-
dale.21
There has been valid criticism aimed at the
use of the Kubelka-Munk analysis for the ap-
pearance of food. This criticism is based on the
fact that Kubekla-Munk theory is based on a
single wavelength concept and, therefore, does
not and should not necessarily correspond to
visual appearance. Moreover, the trend to use
tristimulus values rather than single wave-
270 CRC Critical Reviews in Food Science and Nutrition
lengths is not a valid theoretical approach. As
Billmeyer4 recently pointed out:
The fundamental physical parameters describing the ap-
pearance of foods are, in my opinion, the reflectance and
transmittance of the sample or the Kubekla-Munk scatter-
ing and absorption coefficients S and K, each as a function
of the wavelength. But these parameters are not simply re-
lated to perceived color and other appearance aspects. The
derivation of new quantities with these properties is well
understood and the methodology and fundamental data re-
quired are documented. The parameters themselves — CIE
tristimulus values and chromaticity coordinates, a variety
of opponent color scales of lightness (L), redness-greenness
(a), and yellowness-blueness (b), color differences, and spe-
cial food color quality indices — are well described in the
literature.
It is important to stress the ways in which the fundamen-
tal and derived parameters are appropriately used, the lim-
ited extent to which they can be interchanged, and the place
of each objective methods for food evaluation.
The fundamental parameters (reflectance, transmittance,
K, and S as functions of wavelength) contain all the physi-
cal data necessary to describe the food, but they do not
themselves correlate well with color or appearance. In my
opinion, it is fruitless to attempt to establish such correla-
tions, as for example between reflectance or K/S at one or
two wavelengths and scales of color or desirability of a par-
ticular product. The literature documents many abortive
attempts of this sort in others areas.
The derived parameters (color, color difference, and spe-
cial scales) abridge the fundamental data in ways that are
specifically designed to correlate with color or other ap-
pearance attributes. It is these quantities which should be
used whenever it is desired to provide instrumentally de-
rived equivalents (as nearly as is possible) to scales of per-
ceived color and appearance.
However, it is not appropriate to use the derived param-
eters where the fundamental ones should be used because
of the abridgement involved. For example, Kubelka-Munk
K and S should never be computed from abridged coordi-
nates such as tristimulus values or colorimeter readings. Ex-
cept in cases of totally nonselective absorption and scatter-
ing, quite rare in foods, wrong answers will inevitably result
and false conclusions can easily be drawn.
Procurement specifications can be written around either
the fundamental or the derived parameters, as long as the
writer is aware of the distinctions between them and has
chosen the type consistent with the purposes of the specifi-
cation. Thus, if it is desirable to specify the appearance of
the product, appropriate derived parameters must be used.
If, however, the purpose of the specification is analytical,
such as to ensure adequate strength or concentration of one
or more components, the use of properly selected funda-
mental parameters is necessary.
What Billmeyer is stressing is that the Ku-
belka-Munk theory is based on the characteris-
tics of particulate scattering at a particular
wavelength or particular wavelengths. There-
fore, substitution of tristimulus values, which
are based on the entire spectrum plus the psy-
chophysical observer curves, are not theoreti-
cally valid. The appropriate use of Kubelka-
Munk, then, is for measurement of concentra-
tion and not visual acceptance.
I would agree with this opinion. However,
some literature"'"-48 has shown that such a
treatment fortuitously provides high correla-
tions with visual acceptance. This is not the
method of choice, and it should be understood
that it is not a theoretically valid treatment.
For a further discussion of Kubelka-Munk
theory and light scattering in foods the reader
is referred to Law and Norris44 and an excellent
article still in press by Birth.6
DEVELOPMENT OF MEASUREMENT
SYSTEMS
Having discussed and reviewed the general-
ized theory and applications of colorimetric
data, it might be interesting to look at certain
selected systems which have been developed to
measure the color or some color quality func-
tion of specific foods or food groups. Ob-
viously, not all foods can be covered and, there-
fore, only selected examples will be given.
Hopefully, these selections will provide exam-
ples of the significant potential and use of co-
lorimetry and its applications in food.
Translucent Foods: Orange Juice
It is reasonable to state that with existing
technology, it is quite a simple task to specify
the color of a material if that material is ho-
mogeneous, nonscattering, and transparent or
opaque. Unfortunately, most foods do not fall
into this category but are translucent. This
means that, when light strikes them, some and/
or all the following phenomena occur: reflec-
tion, refraction, transmission, diffusion, and
absorption.
Compounding this complexity is the fact that
translucent foods do not have a constant or re-
producable physical or chemical make-up.
Being biological material, they are subject to
the many environmental vagaries which pro-
duce more or less pigment and scattering mate-
rial from sample to sample. Citrus fruit in gen-
eral and oranges and orange juice in particular
exemplify these problems.
Pigment content varies from variety to vari-

ety as well as within a given variety. Also, pulp
content and size may vary due to biological dif-
ferences as well as processing procedures. Thus,
all the factors mentioned are effected to a
greater or lesser extent depending upon the pulp
and pigment present.
Most objective methods for measuring citrus
fruit color owe their origin to the work con-
ducted by Wenzel, Huggart, Barron and others
at the Agricultural Research and Education
Center, Lake Alfred, Florida, and the literature
is replete with their investigations.
However, much of this work culminated in
1963 when the Department of Citrus requested
R. S. Hunter of Hunter Associates Laboratory,
to develop a prototype citrus colorimeter to du-
plicate the visual judgments of the graders. It
was to be used primarily for orange juice, but
there were obvious implications that it could
probably be used for other juices, such as gra-
pefruit, lemon, lime, tangerine, etc.
Hunter33 developed an instrument to measure
citrus juice colors using a 1-in diameter test
tube as a sample holder and USDA plastic
standards to calibrate this instrument. The best
single-number read-out for a simple instrument
would be Hunter a as judged from data ob-
tained with the Hunter Color and Color Differ-
ence Meter. This meter measures redness by the
following equation
a = 175/ y (1.02X-Y)
where X and Y are CIE tristimulus color values
and fy is nonlinear function of Y. Because of
the nonlinearity, f, is not a simple function to
measure from phototube signals X and Y. An
expression for chromatic redness, which is
much easier to measure and correlates closely,
is the ratio X/Y. Edwards et al.17 derived a
regression equation relating visual score to
Hunter a as follows:
Visual score = 39.44 + 0.87A
From this equation, the relationship between
X/Y and visual score could be calculated (Fig-
ure 14). The second curve of this figure, also
derived from the same data, shows the relation
of 0.8(A/Y) to the visual score. In this equa-
tion, A refers to the amber reflectance, that is,
the long-wave portion only of the X tristimulus
function. The violet part of X is omitted. The
use of the A/Y function provides a steeper
December 1978 271
slope to the curve in Figure 14 and hence better
separation of samples. The filter is also easier
to build than one fitting the complex X func-
tion. The A function was adopted for the Citrus
Colorimeter which was actually graduated to
read:
CR = 200 (A/Y-1)
Citrus redness (CR) would read colorless sam-
ples as zero, and the factor 200 was added to
give better spacing between samples. It is pos-
sible to convert Color Difference Meter values
to CR by the equation:
143a 72b _ 2.55X-0.32Z
CR :-200
/yY f Y Y
where
21 + 0.2Y
fy = 0.51
1 +0.2Y
and X, Y, and Z for the white MgO standard
are 98, 100, and 118, respectively.
A companion scale, citrus yellowness (CY),
was also developed based on the ratio Z/Y:
CY = 100 (1 - 0.847Z/Y)
A white or gray juice will measure zero and in-
crease with increasing yellowness. Color Differ-
ence Meter values can be converted to CY by
the equation:
143b
CY =
Corresponding scales for greenness and blue-
ness are obtainable if one adds the necessary
switch positions from the equations:
CG = 200 (1 - A/Y)
CB = 100 (0.847Z/Y -1)
The term greenness may be a slight misnomer
because grapefruit, lemon, and even lime juices
will all measure plus in CR. This is due to the
use of the A rather than the X function. When
either the X-Y difference or the X/Y ratio is
used to measure redness, samples with zero red-
ness are close to pure yellow, that is, they are
neither green nor red. With the selection of the
A/Y ratio in the Citrus Colorimeter, samples
that are visually slightly greenish will neverthe-
less measure as plus CR.
The conversion of Citrus Colorimeter data to
 272 CRC Critical Reviews in Food Science and Nutrition
1.000
.950
.900
.850
FIGURE 14. The X/Y and 0.8 (A/Y) function plotted against USDA visual score
for the 1954-1955 color data for orange juice.
L, a, b and vice versa via the preceding equa-
tions should be regarded as merely approxima-
tions, since the geometry of measurement for
the two instruments is different. The chromatic
scales of the Citrus Colorimeter differ from the
Color Difference Meter in two ways: (a) The
use of A rather than X tristimulus values, (b)
The use of Y rather than L in the denominator.
This choice was deliberate because the scale ex-
pands as the colors become darker. It was felt
that the CR and CY scales would give color in-
tervals throughout the resulting color solid
which would be more in agreement with visual
ratings of colors of liquids than were the older
a and b scales, which were designed for the
color of reflecting surfaces.
Not only was the development of this color-
imeter an excellent example of the appropriate
use of color theory; upon inspection, the scales
also provide a potential manner in which the
product may be improved. On examination, it
may be seen that citrus red (CR) equals 200 (A/
Y-l) and citrus yellow (CY) equals 100
(1-0.847 Z/Y). This means then that CR is a
function of A/Y and, therefore, to increase the
CR score, we must increase A and/or decrease
Y. A in this case represents the amber portion
of the X standard observer curver, i.e., the red
portion, and Y represents the lightness function
of the CIE color system. Therefore, to increase
CR, one must increase the pigment, which is the
red carotenoids represented by A, and decrease
Y, which may be decreased by the addition of
more pigment, by the use of large diameter pulp
particles, by a lesser number of small diameter
pulp particles, or by the use of extremely small-
diameter pulp particles.
The question might be asked as to the reason
for worrying about pulp, since one need only
increase the pigment content. This is not the
case. In fact, both of these, as related in the
citrus red equation, are very important and cer-
tainly should be investigated. One of the optical
characteristics which affects amount of reflec-
tion, lightness, or darkness is the internal dif-
fusion of a substance, and this is affected, in
turn, by the particle size of the material within
the media, in this case pulp. As particle size be-
comes smaller, diffuse reflection increases
down to a limiting size. In fact, as the particle
diameter decreases, the scattering, or diffusion,
of a material increases inversely with the square
of the particle diameter to a maximum. This
maximum occurs when the particle diameter
reaches approximately 0.1 pm which is about
one quarter the wavelength of light. At this di-
ameter, the particles become too small to reflect
and refract in the usual manner, and reflectance
decreases with approximately the cube of the
particle diameter.
The paint industry takes advantage of this
fact by attempting to maintain pigment particle

diameter at this size. The food industry should
take note of this, particularly the citrus indus-
try. In the citrus industry, a fairly low Y value
or at least an optimum Y value is desired in or-
der to produce a large CR reading. Therefore,
research into finding the optimum particle size
and the optimum amount of pulp which will
give the best possible color reading and color
assessment in conjunction with the amount of
carotenoid pigments present is a potential an-
swer. In turn, if one is dealing with the amount
of carotenoids necessary to add to a given juice,
it is essential that the Y function become con-
stant. That is, the particle size of the pulp and
the amount of pulp in the juice must be kept
constant and the carotenoid pigments varied.
This, then, allows a CR reading to be taken
which is dependent only on the A function or
the carotenoid pigments, since the Y function,
the pulp concentration, and particle size re-
mains the same throughout.
The CY function, which should be high, is
dependent upon the function — , that is, blue-
ness or yellowness over lightness or darkness.
Again, in this case, all parameters should be op-
timized to give high CY readings.
This kind of experimentation is a good ex-
ample of optical characterization. The concern
is with the relative contribution of pigment, and
thus, absorption of light, and pulp which pro-
vides scattering of light. The systems can be dis-
sected in order to find out which parameters
should be changed in order to optimize the total
effect which finally reaches the eye and im-
presses upon the brain the color which is per-
ceived.
More recently Eagerman14 examined the use
of general purpose tristimulus colorimetry in
orange juice measurement. Five general-pur-
pose colorimeters from three different manu-
facturers were tested to determine whether
these instruments could be used in place of the
special purpose Hunter lab Citrus Colorimeter
(CC). A specific sample presentation method
was found that provided a correlation coeffi-
cient of 0.96 or better for each of the instru-
ments with the CC. Comparison of the various
instrumental methods with visual panel scores
showed no significant differences between in-
strumental methods.
It should be stressed that techniques em-
ployed in this study must be utilized for trans-
December 1978 273
lucent materials. That is, simple presentation to
any instrument will not provide satisfactory re-
sults. For methods of sample presentation, the
reader is referred to Francis and Clydesdale,21
Chapter 18, on sample presentation.
Another approach to citrus color measure-
ment has been taken by Petrus and
Dougherty.59 In this study, they evaluated the
spectrophotometric characteristics of orange
juice and pulp. Following this investigation,
Petrus et al.60 found that the correlation of the
sum of carotenoid absorbance with color score
was high regardless of the extractor type or set-
ting and finisher setting. Samples of varying
"Brix, percent sinking pulp, maturity, and va-
riety did not appear to effect the correlation.
The coefficient of correlation was r2 = 0.973
or greater with a coefficient of determination
of r2 = 0.947 or greater. Although this is not a
true colorimetric technique, it has a great deal
of potential for on-line color control.
Translucent Foods: Tomato Products
Tomato products represent another excellent
example of translucent foods which require an
acceptable color and flavor in order to achieve
maximum economic potential from the har-
vested and/or processed product. Due to the
large economic incentive, a great deal of work
has been done on the color measurement of to-
matoes and tomato products. Research dates
back to the 1920s, yet there are still problems
as pointed out by Yeatman.78
It is amazing that raw tomato grading with-
stood the test of time for so many years when
one considers the description of raw tomatoes:
A U.S. #1 tomato had to be "well colored"
with 90% of the flesh having a good red color.
A U.S. #2 tomato had to be "fairly well col-
ored" with 66 Vi°Ia of the flesh having a good
color. In light of this description, instrumental
techniques were evaluated to replace this sub-
jective grading system. A detailed account of
this investigation has been written by Francis
and Clydesdale.21
Two areas of research are of interest and
serve to underline the proper use of data reduc-
tion and the need for a thorough understanding
of color theory. These areas are disc colorime-
try, using the MacBeth-Munsell Disc Colorim-
eter® (MMDC) upon which current grades
have been established, and, secondly, the devel-
274 CRC Critical Reviews in Food Science and Nutrition
opment of the tomato colorimeter.
During the late 1950s the color group at the
USDA in Beltsville, Maryland, were working
on an objective color score for tomato juice.
They obtained visual scores from the MMDC
which they compared to objective L, a, b mea-
surements. It was found that graders were judg-
ing color acceptability of tomatoes not by one
function alone but by all three color functions.
In order to test the correlations of objective
parameters to visual judgments, a series of sam-
ples were evaluated, and it was found that the
function a/L would differentiate between
some, but not all, visual judgments. Further ex-
amination showed that the inclusion of the
chroma function (a2 + b2)1'2 might improve the
correlation of objective data to totally subjec-
tive judgments. As a result, the following color
function was developed:
a/LX l/(a2 + b 2 ) " 2
which produced a correlation of 0.97 with
MMDC data.
Upon examination of Figure 12, it can be
seen that a/(a2 + b 2 )" 2 = cos 8 and, therefore,
the formula could be simplified to cos 0/L.
Hunter and Yeatman36 developed an instru-
ment to read out tomato color (TC) = cos 0/L
or in the CIE system:
TC = 21.6/Y1'2 - 3.0/Y1'2 (Y-Z/X-Y)
where X, Y and Z are measured under illumi-
nant C.
Yeatman has discussed food color standards
TABLE 6
Representative Appearance Standards for Products Under FDA Regulation or USDA Inspec-
tion"
Agency Product Specification
FDA Canned tomatoes Munsell no-
tation; Vo
components
of 5R 2.6/
13; 2.SYR
5/12; Nl or
N4
USDA Fresh tomatoes for USDA To-
processing mato Color
Index TCI
in general" and, more recently, has written an
excellent paper78 on the need for revision of the
current tomato color standards. This paper
shows the need for careful examination of data
expression and the problems which result when
such expression is loosely applied.
Table 6 lists the methodology for canned to-
matoes and fresh tomatoes specified by the
Food and Drug Administration (FDA) and the
USDA grade standards.71
Upon examination of these standards, it is
found that the FDAs standard of quality for to-
mato products, 21 CFR 53.41(b)(2), specifies in
part, when speaking of matching colors by disc
colorimetry, that minimum red color is "the
strength and redness of . . . not less than that
of the blended color of any combination of the
color discs . . . " The USDA, in a similar fash-
ion for inspection of canned tomatoes,72 speci-
fies that:
'A of the disc must be of the specified red color; 'A or less
of the disc must be of the specified yellow color; and the
balance of the disc color (which may be as much as Vi when
no yellow is present) may be all black, all grey, or any com-
bination of these two colors. Thus, the USDA says, "many
combinations of these four colors [may] result in the "Min-
imum Red" color . . . .
Further, the USDA instructions recognize an
important point:78
If the full 'A of yellow and 'A of black (without the grey
color) is used, the same degree of redness (33fo) is present,
but the composite resultant color (upon spinning the disc)
may appear to be a deeper and duller red than if, for ex-
Determination Reference
Visual compari- 21 CFR£543.41(a)l
son; disc color- and (b)
imeter; mini-
mum color
USDA Tomato 7CFRf51.3317(a)
Colorimeter;
nondeaerated
raw juice

ample, equal parts of grey and black are exposed on the
disc. If in this combination only Vi grey is
used . . . color . . . appears . . . a lighter red, but . . . the
same degree of redness is present. If the maximum permit-
ted yellow (Vi) is not used and either the black or grey discs,
or combinations thereof are substituted . . . the resultant
color will not appear to have a red hue but rather a dull
reddish-brown effect. In using the Munsell color disc . . .
for comparison as to compliance with "Minimum
Red" . . . the maximum permitted yellow ('A of the disc
area) is recommended to avoid comparison with a color
that is not typical (dull reddish-brown effect).
This observation is important in visually as-
sessing product color.
Yeatman78 has pointed out that:
The FDA's "Minimum Red" for tomato products purports
to be a single color, when in fact it is many colors. Figure
15 illustrates the impact upon color standards when any
combination of specified colors is permitted. The descrip-
tion of "Minimum Red" color permissable in both the
FDA and USDA standards contains a multiplicity of colors
which are so widely divergent as to be unrealistic as stand-
ards for tomato products.
Further:
Since the USDA grade standards for processed tomato
products may serve as the basis for correlation of instru-
mental measurement to score points assigned by experi-
enced inspectors, a scoring system which would avoid ex-
trapolation beyond the grades set for A and C quality
would be beneficial.
FIGURE 15. Color designations for "minimum red"
plotted in three-dimensional aspect in CIE color space.
Darkened area at top of each column is the actual point.
(Reprinted with permission from Yeatman, J. N., Food
Technol. (Chicago), 30(11), 50, 1976. Copyright © by In-
stitute of Food Technologists.)
December 1978 275
The USDA grade standards specify grade A (45 score
points) and grade C (40 score points) in addition to the
"Minimum Red" promulgated by the FDA. Looking at
these color standards in three-dimensional aspect (Figure
16), we see a displacement in color space from the FDA's
"Minimum Red". "Minimum Red" in Figure 16 is desig-
nation No. 1 in Figure 15.
Based on these facts, Yeatman78 has sug-
gested two new color designations for USDA
tomato products grade standards:
I. A lower-limit disc designated C - (35 score points), con-
taining the following percentage components of 1972 Mun-
sell production paper: 41 % red, 36.5% yellow-red, and
22.5% black; and having the following chromatic values:
%Y = 9.5, x = 0.5525, y = 0.3717; with Munsell notation
= 1.5YR3.7/9.7.
2. An upper-limit disc designated A + (50 score points),
containing the following percentage components of 1972
Munsell production papers: 78% red, 13% yellow-red, and
9% black; and having the following chromatic values: %Y
= 6.8, x = 0.5712, y = 0.3446; with Munsell notation =
9.8R 3.1/9.7.
A single color designation for "Minimum Red" is also
suggested, to avoid the multiplicity of colors which now
dominate the federal standards. The color suggested should
contain the following percentages of 1972 Munsell produc-
tion papers: 33% red, 33% yellow-red, and 34% black; and
have the following chromatic values; %Y = 8.6, x =
'tf/. /
16 18 20 22 24 26 23
FIGURE 16. Color designations for USDA grades of re-
constituted tomato paste plotted in three-dimensional as-
pect in Hunter color space. Darkened area at top of each
column is the actual point. Columnar letters are USDA
grades; graded samples are at the center and lower
left.(Reprinted with permission from Yeatman, J. N., Food
Technol. (Chicago), 30(11), 50, 1976. Copyright © by In-
stitute of Food Technologists.)
276 CRC Critical Reviews in Food Science and Nutrition
0.5448, and y = 0.3716; with Munsell notation = 1.6YR
3.5/9.0.
Use of these additional guidelines regularly spaced
should improve both visual evaluation and instrumental
measurement, providing a firm base for inspection and
quality control.
Visually Transparent Foods: Beverages
Earlier it was stated that color measurement
was rather a simple exercise if the sample was
transparent. Like all simplistic generalizations,
exceptions seem to constitute a majority of the
generalized cases. Such is the problem with
many beverages due to degree of darkness or
perhaps minute amounts of particulate scatter-
ing material. In order to illustrate some inter-
esting colorimetric work in this area, certain
beverages will be chosen to represent this class
of food materials.
Although the discussion thus far has mainly
centered around the use of tristimulus colori-
metry for the measurement of color and the
possible application of such techniques to pig-
ment analysis, it would be counterproductive to
ignore any useful chemical techniques for such
analysis. A method for the determination of to-
tal anthocyanin (T Acy) and a degradation in-
dex (DI) for cranberry juice has been proposed
by Fuleki and Francis." If natural anthocy-
anins (Acy) were to be used as a colorant, one
of the problems which would arise is the deg-
radation of such pigments during storage and
the accumulation of brownish degradation
products. Although these degradation products
have an absorption maxima around 415 nm,
they also absorb, to some extent, at 510 nm."
This would introduce an error if a simple direct
T Acy determination method, similar to that
developed for cranberries, was used.23
Therefore, Fuleki and Francis" described the
development of an improved T Acy determina-
tion which minimized the interference due to
these brown-colored degradation products.
This method involves the measurement of the
absorbance at 510 nm on samples diluted with
pH 1.0 and 4.5 buffers. The pigment content is
calculated in absolute quantities with the aid of
extinction coefficients established for the cran-
berry Acy dissolved in the buffers. The data ob-
tained from such a method can also be used for
the calculation of an index, which is indicative
of the proportion of degraded Acy in the sam-
ple. Such an index is quite useful, particularly
in cases when the original Acy content is not
known, and a measure of the state of Acy deg-
radation is needed.
The degradation index is calculated by utiliz-
ing a T Acy determination based on the single
pH method (pH = 1) as well as the pH differ-
ential method (pH = 1.0, pH = 4.5). Thus, the
DI is calculated as follows:
DI T Acy by the single pH method
T Acy by the pH differential method
In most cases, particularly where the Acys
are unidentified or E values are not known, a
simplified DI (DI simp) can be obtained using
the following equation:
DI simp = TOD pHl.O/OD
Where TOD is the total optical density.
This proposed DI is based on the principle
that the measurement obtained at pH 1.0 will
include the absorption due to the degraded as
well as nondegraded Acy, while the difference
in absorption between the pH 1.0 and 4.5 media
will be due only to the nondegraded Acy. Sam-
ples not containg degraded pigment or other
brownish-colored compounds, should give al-
most identical results for T Acy with the single
pH and pH differential methods. However, as
the degradation proceeds, the amount of red-
dish-brown pigments will increase. Conse-
quently, the T Acy determined with the single
pH method and the DI will be higher than 1.0.
The simplified DI value will be greater than 1,
even for fresh samples not containing degraded
pigment, because the Acys have a small absorp-
tion at pH 4.5.
Fuleki and Francis" point out that the DI
should give a better measure of the actual color
of the sample than either the T Acy content, the
TOD or the AOD value. The principle use fore-
seen by these authors for the proposed DI
would be similar to that of absorption ratios.
A limit value could be established for the DI
above which the degraded color becomes visible
and the product should be rejected. It might
also be possible to establish a direct correlation
between DI and the percent of the original pig-
ment which is still present in the sample.
Another objective spectrophotometric
method was developed by Watada and Abbott74
for determining the color grade of grapes. This
method utilized the USDA Spectrophotometric

Difference Meter® (SDM) developed by Birth
and Norris7 at two wavelengths, 630 and 690
nm. Most of the absorbance at 630 nm was as-
sumed to be due to Acys. Other factors which
absorbed at 630 nm were assumed to be absorb-
ing at all other wavelengths, thus the difference
in absorbance at 690 nm from that at 630 nm
was used as the SDM reading of the Acy con-
tent.
Acy-containing beverages, including wines,
have provided an interesting substrate for a
great deal of colorimetric work. Francis and
Clydesdale21 have reviewed the measurement
area, but some specific comments concerning
data manipulation are appropriate to this dis-
cussion.
An interesting method of blending wines to
color has been proposed by Little and LiaW'
based on the construction of blending dia-
grams. The reciprocal log functions of the mea-
sured red, green, and blue transreflectometric
readings are plotted against a sliding-scale ab-
scissa, from which can readily be determined
the composition of blends from different start-
ing wines. Tristimulus colorimetry has long
been used in the paint and textile industries for ' and saturation are fixed by both the L and the
blending color, and it is refreshing to see food
materials being treated in a similar scientific
fashion.
Van Buren et al.73 presented a method for ex-
pressing the color of different Acys under dif-
ferent conditions of acidity. In this work, the
importance of environmental conditions to
color should be noted. Too often measurements
are assumed to be reproducible under varying
conditions; generally, this is not the case. In
part, this study found that the variation of A
max with pH was less pronounced at low con-
centration of Acy. At the higher concentra-
tions, shifts as great as 9 nm were seen as the
pH was changed from 3.0 to 3.8. These shifts
may be related to the increasingly greater pro-
portion of Acy pigment present in the carbinol
and anhydro forms as the pH goes higher. The
A max for absorption by Acy solutions occurs
at only slightly shorter wavelengths than the
555 nm A max for the action spectrum of the L
color component. Therefore, solution A-max
shifts to longer wavelengths would be expected
to result in a more complete overlap with the L
function and would decrease the L correspond-
ingly to a given absorbance. Thus, cyanidin-3-
December 1978 277
glucoside solutions, with their lower A max, had
a higher L at a given absorbance than did solu-
tions of the other Acys. When Acy solutions
were examined at different pHs and a constant
absorbance, it was found that the longer A max
at high pH was associated with a lower L value.
A convenient method of presenting Hunter
tristimulus color values is by plotting chromat-
icity factors, such as a, b, 9 (hue angle) and sat-
uration, against L values.64 This not only places
the color in a definite location in the color
space, but it also provides an indication of the
effects of concentration of the perceived color.
In summary, Van Buren et al." found that
solutions of a pure Acy give a range of colors
depending on concentration and the pH of the
solution. While there are marked differences
between Acys regarding the range of colors ob-
tainable, they impart a similar dark red color
to concentrated solutions. The L color function
increases as concentration decreases, while at a
given concentration, the lower the pH (in the
3.8 to 3.0 range) the lower the L. The relation
between L and absorbance is different for dif-
ferent Acys. Color chromaticity dimensions, 6,
pH of the solution. Furthermore, the 0 and sat-
uration show maximum values in the middle L
region. These maximum values are related to
the structure of the Acys.
Little45 has presented the application of her
proposed model,47 which was discussed previ-
ously, to wines. Two dark red wines were di-
luted with equal volumes of 12% aqueous al-
cohol to bring them into range for
spectrophotometrically analyzing for Acy con-
tent. The diluted samples were also measured
with thin-layer transreflectometry (Little and
Liaw49) in terms of R, G, B and converted to
CIE and Hunter coordinates. In Table 7, it was
noted that the essential but important diference
in evaluations involves the estimation of change
in dominant wavelength (DWL) and hue angle
6. According to the CIE data, upon acidifica-
tion, Sample 1 showed a large shift in DWL and
Sample 2 only a small one, while the Hunter
data indicated equal changes in 0 for the two
samples.
Previously, Joslyn and Little3' had shown
that changes in DWL in rose wines at their nat-
ural pH, under conditions favoring Acy degra-
dation, correlated with loss of Acy, as mea-
278 CRC Critical Reviews in Food Science and Nutrition

TABLE 7

Comparison of Shift in Chromaticity of Red Wine with Decrease in pH, in Hunter and CIE Coordinates

CIE Hunter

Sample pH Y X y DWL EP L a b 8 SI

I 3.3 56.1 0.3406 0.3250 605 11 74.9 9.1 6.8 36.8 11.4
1.2 40.1 0.3840 0.3023 493 20 63.3 32.8 5.4 9.4 33.2
Difference — -16.0 + 0.0434 -0.0227 + 106.5 +9 -11.6 + 23.7 -1.4 -27.4 + 21.8
2 3.3 51.1 0.3614 0.3505 584 22.5 71.5 6.5 15.2 66.8 16.5
1.2 44.7 0.3804 0.3386 594 24 66.9 17.0 13.9 39.3 22.0
Difference -6.3 + 0.0190 -0.0119 + 10 + 1.5 -4.6 + 10.5 -1.3 -27.5 + 5.5

Note: DWL = dominant wavelength; EP = excitation purity; SI = saturation index.

From Little. A. C. and Liaw, M. W., Am. J. Enol. Vitic.,25, 79, 1974. With permission.

TABLE 8

Estimation of Anthocyanin Content of Red Wine by Comparing the Difference in Absorption with the Magnitude of
the Shift in Chromaticity in Hunter and CIE Coordinates

Brown Chromaticity shift

pigment
Sample PH O) CIE(x 10J) Hunter

1 3.3 0.075 0.051 0.024 30 4.90 23.7

1.2 0.355 0.355
Difference — 0.280 0.304
2 3.3 0.056 0.024 0.032 57 2.24 10.6
1.2 0.170 0.170 2.24
Difference — 0.114 0.146
Ratio (1:2) — 2.5 2.1 2.2 2.2

From Little, A. C. and Liaw, M. W., Am. J. Enol. V7r/c.,25, 79, 1974. With permission.

sured analytically. Therefore, there was reason
to predict that the Acy content of Sample 1 is
considerably higher than that of Sample 2,
since both DWL and 6 suggest it to be redder
than Sample 2, and it did in fact appear redder.
Further, it could be predicted that, upon acidi-
fication, the shift in hue or its psychophysical
correlate DWL would be greater for Sample 1
than for Sample 2 — a prediction borne out, as
already mentioned, by the CIE but not by the
Hunter data. The relatively higher purity at pH
3.3 and the smaller change upon acidification
for Sample 2 suggest increased browning in
Sample 2.
Confirmation of these predictions is found in
Table 8, which gives analytical data on Acy
content by a modification of the Swain-Hillis"
method.
For Sample 1, the contribution of brown pig-
ment at 520 nm amounted to 30% of the total,
while for Sample 2, the proportion of brown
was 57%. This is fully consistent with the ob-
served differences in purity. The chromaticity
shifts, expressed in both CIE and Hunter coor-
dinates, showed a twofold difference between
Samples 1 and 2, in agreement with the absorp-
tion differential. In this example, although a
measure of magnitude-of-chromaticity shift in
both coordinate systems worked equally well,
the CIE system provided additional valuable in-
formation regarding changes in dominant
wavelength (DWL) and excitation purity (EP)
upon acidification.
As mentioned previously, Johnson et al."
• found a discrepancy in the correlation of DWL
with visual assessment; however, this may have
been due to the systems studies. Also, it might
have been advantageous to use the pigment
analysis described by Fuleki and Francis23
which overcomes some of the shortcomings of

the earlier Swain-Hillis69 method. Nevertheless,
this work is an excellent example of how data
reduction can be evaluated in a model.
Robinson et al.,64 in a study with red wine,
found that both hue and saturation increased
with decreasing lightness to a maximum and
then decreased. Van Buren et al." found the
same relationship with different types of Acys
in solution. This is not an unexpected result,
since darker solutions allow less light through
and, therefore, less energy reaches the photo-
tube. However, the complete answer is not as
simple as this. As a result of this perception,
Eagerman et al. is investigated the use of differ-
ent color scales for dark-colored beverages. The
color scales currently employed on a number of
colorimeters were developed for reflectance
measurements. This is true even if the instru-
ment is equipped to make transmission mea-
surements. Since transmission data may be ac-
quired with these instruments, the applicability
of these so-called "reflectance scales" was in-
vestigated. This research involved examination
of several reflectance or surface color scales
and transmission color scales to determine their
usefulness for transmission colorimetry.
The main problem involved in making trans-
mission measurements on samples is the diffi-
culty of photocells adjusting to low-luminosity
situations in a manner similar to the human
eye. With lightly colored clear liquids there is
no problem. However, with darker-colored liq-
uids, the photocells do not adequately measure
the visual chroma or hue of the sample. Thus,
the measurement may be more of a lightness or
darkness measure than a color judgment. Little
work has been done on the application of trans-
mission scales to dark transparent food prod-
ucts.
Several colorants, both FDC dyes and a nat-
ural Acy, were evaluated at different concentra-
tions with different color scales. In this study,
some new terminology was defined in order to
explain the results, and it is appropriate to de-
fine these terms in this discussion.
Area of confusion of inversion area — The
region where changes in the pigment concentra-
tion of the samples fail to correlate with
changes in the scale readings. This is illustrated
graphically in Figure 17.
Color scale — A scale consisting of three pa-
rameters which completely defines or locates a
December 1978 279
color in a particular color solid. A color scale
may be transformed mathematically to another
system which utilizes a different color solid.
Luminous-transmittance parameter — The
one parameter of a color scale which defines
lightness or darkness of a sample. The most
commonly used luminous-transmittance pa-
rameters in this investigation are CIE Y and
Hunter L.
Chromaticity parameters — The two param-
eters of a color scale which define hue and
chroma at a constant luminous transmittance.
Compound functions — Those color-quality
parameters whose conversion formulas possess
a luminous-transmittance parameter in both the
numerator and the denominator of the conver-
sion equations.
Simple functions — These color quality pa-
rameters whose conversion formulas do not
have a variable quantity in the denominator of
their conversion equations. These are the equa-
tions for the derivation of CIE X and Z and
Colormaster R and B.
Root functions — Those color quality pa-
rameters whose conversion formulas are simple
but have the variable terms taken to some root.
Examples of these are the Glasser and Adams
Scales.
Data from both the General Electric Record-
ing Spectrophotometric (GERS) and the
Hunter D25 were converted into the 11 remain-
ing scales. It was seen that with most scales
there was some area of confusion dependent
upon the value of the luminous-transmittance
parameter. This means that the chromaticity
parameters increased with pigment concentra-
tion to a certain level of luminous transmit-
tance. Below this level, the same parameters be-
gan to decrease. The area of confusion, or
inversion area, refers to this change in the chro-
maticity values when a certain level of luminous
transmittance is reached.
Since chromaticity parameters are analogous
to hue and chroma, it would be desirable, if, as
the colorant concentration increased, such val-
ues would increase linearly, no matter what the
luminous transmittance. This does not hold,
basically due to lower energy levels being trans-
mitted through darker samples. The photocells
then collect a weaker signal, which is translated
into a smaller chromaticity-parameter reading
and also due to the shape of the color solid
280 CRC Critical Reviews in Food Science and Nutrition

which is nonlinear in hue with a linear value
progression. The luminous-transmittance val-
ues, where areas of confusion occur for differ-
ent color scales, are shown in Table 1. The lu-
minous-transmittance parameter used was
Hunter L for all scales, including X, Y, Z, in
order to simplify interpretation. The results
from the two different sets of data agree quite
closely.
Figure 17 illustrates the terms "area of con-
fusion" or "inversion area". This is a graph of
the Hunter L, a, b values for cyanidin-3-gluco-
side (Cn-3-G). The horizontal axis is the ap-
proximate pigment concentration and the ver-
tical axis is the value for the respective
parameters. In this graph, it may be seen that
luminous transmittance decreases proportion-
ally with increased pigment concentration, but
both chromaticity parameters show an inver-
sion in certain luminous-transmittance areas.
Some scale parameters did not show any in-
version. These included all of the luminous-
transmittance parameters and those chromatic-
ity parameters whose conversion formulas were
simple functions, as defined previously. Those
parameters which showed areas of confusion
were the compound or root functions. Further
inspection of the results showed that nearly all
of the compound and root functions had areas
of confusion at almost the identical range of lu-
minous transmittance (Table 9). The only pa-
rameters which had inversions at some other lu-
minous transmittance were the two scales
specifically proposed for transmission work,
the Hunter L, a', b' and the Citrus scales. Also,
the Rd, ARJ, bKd scale, which was designed for
reflectance measurements, showed an inversion
area at a different luminous-transmittance
level. However, its inversion area was at the
highest luminous-transmittance level of any of
the scales tested. All of the other scales are con-
sidered reflectance scales. Although each gives
different numerical tristimulus values for the
samples, the overall results, with regard to
areas of confusion, were approximately the
same.
Further work in this study indicated that
many commercial dark juices when measured
in 10-nm cells are at or below the areas of con-
fusion for all scales except the Hunter L, a', b'
scale. Thus, care should be taken when express-
ing quality standards for these products.
Eagerman et al.15 explained the inversion the-
oretically as well as providing empirical data to
support their thesis.

70

L values
a values
b va/ues

t r 1o
PIGMENT CONCENTRATION (mg/ m |)
FIGURE 17. L, a, b values vs. pigment concentration (Cn-3-G). (Reprinted with
permission from Eagerman, B. A., Clydesdale, F. M., and Francis, F. } . , J. Food
Sci.. 38,1051, 1973. Copyright © by Institute of Food Technologists.)
 December 1978 281

TABLE 9

Luminosity Levels at Which Areas of Confusion Occur for the chromat-

icity Parameters of Various Color Scales Calculated from GERS X, Y,
Z and D25 L.a.b Values

Luminosity value (in terms

of L) where area of confu-
sion occurs
Chromaticity
Color scales parameter GERS data D25 data Ref.

L.a.b a 36.8—42.9 45.9 67a

b 51.9—55.8 48.3
L.o,/? a 36.8 34.4—37.0 35
P 46.8—48.8 44.7
L'.a'./r a' 36.8 34.4—36.5 35
P' 46.8—48.8 44.7
L.a.b a. 36.8—42.9 40.3 24
bc 42.9 40.8
L,. a,, b. a. 36.8 34.4—40.3 24a
b. 42.9 40.8
L., a., b. a. 35.4—42.9 40.3—40.8 1
b. 40.3—42.9 40.3
U*. V*,W* U* 40.3—42.9 44.0—44.7 75
V* 48.8—53.2 48.3
%Y,x, v X 25.5—29.8 29.9—34.4 8a
y 53.2—57.1 48.5—53.4
R-, a,, b. a,, 63.0 62.8 35a
b. 59.0—61.2 53.4—61.7
L, a, b' a' 19.0—21.9 19.0—21.9
b' 35.4 29.9—37.0 32
Citrus scales CR 19.0—21.9 19.0—21.0 34
CY 35.4 29.9—34.4

Note: Pigment used was cyanidin-3-glucoside.

From Eagerman, B.A., Clydesdale, F.M., and Francis, F.J., J. Food

Sci., 38,1051,1973. With permission.

The conversion formulas for the Hunter L, a, b scale from
X, Y, Z are:
L =1 0 Y " 2
a = 17S(1.02X-Y)
Y"J
. _ 70(Y-0.847Z)
The factors 175 and 70 were derived experimentally to in-
crease the fit of the spacing of Hunter coordinates to Mun-
sell coordinates. On the other hand, 1.02 X and 0.847 Z are
factors required to place Illuminant C at the origin of the
Hunter a, b space. The X, Y, Z scale is a simple function,
having no areas of confusion and all parameters increase
as measurements go from the darkest to the lightest solu-
tions. Figure [18] is a rectangular coordinate plot of the X,
Y, Z raw data for cyanidin-3-glucoside dilution. It is a plot
of parameter readings versus pigment concentration, and
will be referred to throughout the following discussion.
Because Z is constant for a good portion of the dilution
curve, it is easier to begin the mathematical discussion with
the parameter involving Z. The reason for Z being constant
can be seen in Figures [19] and [20] which are representative
spectra of the systems tested. It may be seen that until the
overall solution becomes light enough, there is not suffi-
cient magnification of the X' peak to be measured by the
photocell. X' is approximately one-sixth the size of the X
peak and is located in the Z region of the spectrum. With
a red material, X' should be a very high percentage of the
total Z reading. When the X, Y, Z values are converted to
the " b " parameter of the L, a, b scale, Z is only involved
as a term which is subtracted from Y. If Z is constant, while
Y is increasing, the numerator of the formula will increase
and the overall function will increase. However, as can be
seen from Figure [18], once Z starts to rise, it does so at a
much faster rate than does Y. This will cause the numerator
to decrease, while the denominator increases, thus causing
the function to decrease. The area of confusion is the tran-
sition period where the numerator is increasing only
slightly, so that the increase in the denominator causes the
overall function to remain approximately constant and then
decrease.
282 CRC Critical Reviews in Food Science and Nutrition

70

X
Y vt/ues
~Z vtltes

1-
I"
10

2 4 6 B 10
PIGMENT CONCENTRATION (mg/ml)

FIGURE 18. X, Y, Z values vs. Cn-3-G concentration. (Reprinted with permission

from Eagerman B. A., Clydesdale, F. M., and Francis, F. J., J. Food Sci., 38,
1051, 1973. Copyright © by Institute of Food Technologists.)

4001

90

400
WAVELENGTH (nm)

FIGURE 19. Spectral curves for Cn-3-G solutions.

Curves represent 0.10, 0.17, 0.69. 1.73, 3.45, 6.90, and 17.3
mg of pigment per milliliter. (Reprinted with permission
from Eagerman, B. A., Clydesdale, F. M., and Francis, F.
J., J. Food Sci., 38, 1051, 1973. Copyright © by Institute
of Food Technologists.)
 December 1978 283

WO]

80

UJ

UJ
OL

400 £00 600 700

WAVELENGTH (nm)

FIGURE 20. Spectral curves for solutions of F,D, and C.

Red No. 1. The curves represent 0.05, 0.10, 0.30, 0.70, 3.0
and 10.0 mg of pigment per milliliter. (Reprinted with per-
mission from Eagerman, B. A., Clydesdale, F. M., and
Francis, F. J., f. Food Sci., 38, 1051, 1973. Copyright ©
by Institute of Food Technologists.)

With "a" the situation is similar. Here, the numerator is
1.02 X-Y, and while neither of the variables is constant, the
relative rates of increase do vary. In the lowest luminous
transmittance range, X increases faster than Y, then in the
middle and upper regions, Y increases faster than X. So,
here again, the numerator first increases, then decreases,
with the denominator increasing constantly.
The other compound scales show that the chromaticity
parameters for the two Judd-Hunter scales, Citrus Scales,
R<, a»w, ba* and Hunter's transmission scale all have nu-
merators similar to the L, a, b system in that variables with
different rates of increase are subtracted from each other,
while the denominator increases constantly. The two other
compound function scales, Wyszecki (1963) [Reference 75]
U*, V* and the chromaticity coordinates have conversion
formulas which differ in form from the others. These have
just one variable in the numerator, X, Y or Z, and all three
in the denominator added together. The inversion of these
parameters is dependent upon the denominator. Here
again, it is the relative rates of increase that control the
inversion. Chromaticity coordinate "x", whose conversion
formula is: X/(X + Y + X), will serve as the example. At
lowest luminous transmittance, X increases the fastest; then
later Y and Z start increasing faster than X, so the overall
denominator increases faster than the numerator. Thus, the
parameter first increases then decreases. The analogy holds
for the the other parameters of this type.
Root functions, such as the Glasser and Adams scales
have variables taken to some root and subtracted from each
other. It appears that relative rates of increase control scale
inversion here also, as the formulas, without taking the
equation to the root, are similar to the form of the com-
pound function numerators. Also, the areas of confusion
for these scales are in the same luminous transmittance re-
gion as those for most of the compound functions.
Simple scales, such as the CIE X, Y, Z scale, do not have
any areas of confusion, nor are they linear with pigment
concentration. The rectangular coordinate graph of X, Y,
Z readings versus pigment concentration [Figure 18] resem-
bled the shape of a semilog plot. Therefore, a plot of the
reading versus the log of the pigment concentration was
made, as shown in [Figure 21.] The plots of X and Y, the
only ones of major concern in a red system, closely approx-
imate a straight line for one log cycle, but veer off slightly
in opposite directions on either side of the straight line por-
tion. The linear portion of X corresponds to luminous
transmittance levels of Y = 13 to 55 or Hunter L of 37 to
75. This does not include the luminosity levels of the dark
colored juices. The linearity of the line, as described by lin-
ear regression analysis, showed a coefficient of determina-
tion equal to 0.964 for the linear portion and 0.931 for the
entire range. The coefficients for Y over its linear region
were 0.986 and 0.875 for the linear portion and the entire
range respectively. However, because the plots are not lin-
284 CRC Critical Reviews in Food Science and Nutrition

90
X values
Y values
Z values

.06 •/ 1.0 10.0

LOG PIGMENT CONCENTRATION

FIGURE 21. X, Y, Z values vs. the log of Cn-3-G concentration. (Reprinted with
permission from Eagerman, B. A., Clydesdale, F. M., and Francis, F. J., J. Food
Sci., 38, 1051, 1973. Copyright © by Institute of Food Technologists.)

ear from the initial concentration upward, this scale would

not be useful for the prediction of pigment concentrations.

Reduced data was also tested and explained

in an analogous manner. This work should be
fully investigated by those interested in color
measurement of dark-colored beverages.
In this study it was noted that the two scales
having inversion areas at the lowest luminous
transmittance were those developed by Hunter
3204
for use with liquids. They are the L, a', b'
scales and the Citrus Colorimeter scales. These
differ mathematically from the Hunter L, a, b
basically by the denominator of the conversion
equation. The transmission scales have Y in the
denominator as opposed to Y" 2 for the reflec-
tance scales. The logic used in changing the de-
nominator was to expand the chromaticity pa-
rameters at lower luminous transmittance
relative to their values at higher luminous trans-
mittance. Figure 22 shows a graphical represen-
tation of the expansion at lower luminous -a
transmittance levels of the L, a', b', labeled
"a", scale relative to the L, a, b scale at the FIGURE 22. A representation of L,a'% b' and L.az.b,
same luminous transmittance. For the range of color scale expansion. (Reprinted with permission from Ea-
L values, it is shown what a would be relative german, B. A., Clydesdale, F. M., and Francis, F. J., J.
Food Sci., 38, 1056, 1973. Copyright © by Institute of
to constant a of 1 or - 1 . The two-dimensional Food Technologists.)
graph for b and b' would be identical. The rea-
son why a scale was proposed which expands at cannot. This expansion is presumed to be a
lower luminous transmittance is that, although mathematical adaption mechanism for the pho-
the eye is able to adapt to lower luminous trans- tocells. The proposed scale (Figure 22) partially
mittance and give color judgments, photocells accomplished the desired effect. The area of
 December 1978 285

confusion was lowered, but grape juice, fruit
punch syrup, and prune juice, in 10-mm cells,
all produced readings around the area of con-
fusion for these scales.15 It appeared that more
expansion of the chromaticity parameters was
necessary.
Therefore, Eagerman et al.16 developed a set
of new scales, based on this reasoning, designed
for transmission colorimetry.
It was known that greater expansion of the
parameters at lower luminous transmittance
than that given by Y in the denominator was
necessary. Employing this logic, Y2 was tried as
the denominator, while leaving the numerator
unchanged. Figure 22 shows the expansion
caused by a function with Y2 in the denomina-
tor (labeled "a 2 "). Using Y2 as the denominator
eliminated the inversion of the parameters and
created scales which were close to being linear
with pigment concentration when plotted on
rectangular coordinate paper. Figure 23 is a
plot of the data from four different colored sys-
tems, cyanidin-3-glucoside, Food, Drug, and
Cosmetic (FDC) Red #1, Red #2, and Yellow
#6. Each is a plot of: a2 = 170(1.02X-Y)/Y2.
However, each colored system produced lines
of different slopes, even though all lines were
reasonably linear.
Observing that despite the differing slopes,
all the curves had similar shapes, a graphical
solution was attempted. Using a computer
(CDC 3600®) to do both the calculations and
the graph plotting, various exponents and ad-
ditive constants were used in several attempts
to find formulas which most closely approxi-
mated linear relationships between colorant
concentration and parameter values. The for-
mulas developed will be called a* and b* to de-
note the relationship to the Hunter color solid.
It was hoped that one formula for a* and one
for b* would linearize all of the colored sys-
tems, but that did not happen. Because of dif-
fering slopes, the multiplying factors straight-
ened each curve to a different extent. As a
result, different formulas had to be found for
each colorant system.
Table 10 illustrates the results of this com-
puter technique and shows the many possible
combinations of exponents and additive con-
stants which give plots very close to linear and
coefficients of determination better than
0.9990.

70

60

50

40

• Red N« 2
20
o Cn-3-G
* Yellow N* 6 (XIO!
10
o Red N*I

0
2 3 4 5 6 7 8 10
COLORANT CONCENTRATION (mq/ml)

FIGURE 23. Plot of a2 for each of the colorant systems. The coefficients of deter-
mination for the Curves 1, 2, 3, and 4 are 0.9865, 0.9937, 0.9724 and 0.9901, re-
spectively. (Reprinted with permission from Eagerman, B. A.. Clydesdale, F. M.,
and Francis, F. J., / . Food Sci., 38, 1056, 1973. Copyright © by Institute of Food
Technologists.)
286 CRC Critical Reviews in Food Science and Nutrition

TABLE 10

Formulas and Degree of Linearity for Color Quality Parameters Which are Linear to
Colorant Concentration

Colorant or Parameter Coefficient of

pigment (a* o r b ' ) Formula Slope determination

F.D.C a* 2.50 0.9995

Red No. 1
2.43 0.9994
i 2.36 0.9994
2.45 0.9993
F,D,C a» 7.61 0.9999
Red No. 2
7.57 0.9998
7.94 0.9996
9.32 0.9993
Cyanidin-3-glu- a* 18.53 0.9996
coside
22.25 0.9992
20.55 0.9990
F.D.C a* 0.0127 0.9992
Yellow No. 6
F.D.C b* -6.47 0.9996
Green No. 3
F.D.C b* 147.6 0.9994
Blue No. 1

Reprinted with permission from Eagerman, B. A., Clydesdale, F. M., and Francis, F.
J., J. FoodSci., 38, 1056, 1973. Copyright © by Institute of Food Technologists.

This process, which relates color scales to
pigment contents, bears some relationship to
the use of broad-band filters in absorptimetry
prior to the present popularity of grating mon-
ochromators. The important difference is that
these equations utilize most of the visible spec-
trum instead of a small portion. It remains to
be seen whether this approach will provide a
more sensitive mechanism for pigment systems
which extend over a broader portion of the
spectrum than that employed with an absorpti-
meter. This distortion of the conventional color
scales, which would occur when the coefficients
are changed, will obviously affect the relation-
ship to visual appearance. However, if this dis-
tortion is too large, the colorimeter could be
programmed internally to read out any desired
scale, and, with a switch, converted back to
normal operation. This principle is employed in
the latest model of the Tomato Colorimeter®.
The ramifications of such a system are quite
immense. Colorimetric techniques normally re-
quire only a few minutes to produce data
whereas chemical techniques generally require
a minimum of several hours. This concept pro-
vides a possible rapid, accurate, and efficient
method for the determination of pigment con-
tent within a defined system. This is a use of
tristimulus colorimetry which has not been
taken advantage of in the past.
In order to test these scales, which Eagerman
et al." developed with model systems, Johnson
et al." investigated the use of this concept on a
food product. Cranberry juice cocktail was
chosen as a typical red fruit juice which is
known to show color degradation on storage.
It was decided by Johnson et al.37 that a good
test of this concept would be to attempt to de-
velop scales which would not only follow pig-
ment degradation but which would also predict
visual color assessment. This would provide a
simple colorimetric technique which would pro-
vide not only a measure of pigment content but
also visual assessment.
Instrument readings and reduced data were
correlated with both pigment concentration and
visual ranking. Data from the CIE system were
converted to the L, a*, b* system using the
computer program developed by Eagerman et
al. 15-16 Data obtained from this program were
 December 1978 287

then correlated with pigment concentration and
visual ranking.
Table 11 shows the correlations of Acy con-
centration vs. instrumental functions for the
three experiments. It is seen here that all func-
tions except dominant wavelength, hue, and b
provide high correlations: dominant wave-
length and hue are analagous. The hue of a par-
tially degraded system is due to a combination
of effects brought about by the intact pigments
and the degradation products (browning com-
pounds). These two factors are not perfectly
correlated; thus, hue is not a good indicator of
either. The reason that b is a poor index might
be explained in the conversion equation:
b = 7.0(Y-0.847Z)/Y1/J
Acys absorb maximally near 535 nm and
would affect the Y function. The browning
compounds absorb at 400 nm and would affect
the Z function, but to a greater extent. Thus,
these two values change at different rates and
give a low correlation to b. The cell thickness
does not seem to markedly affect the results as
seen by a comparison of the Gardner (cited in
Johnson et al.)37 data to the Hunter data which
were done with different cell thicknesses. Cor-
relations that were high for functions from one
instrument were also high for the other instru-
ment. The same was true for the function giving
low correlations.
As well as these conventional scales, Johnson
et al.37 developed new scales based on the work
of Eagerman et al." These scales were then
evaluated as a predictor of both chemical con-
tent and visual judgment.
Data from the cranberry cocktail experi-
ments in the CIE system were put through the
SEARCH program" in order to find the con-
version formula which gave the highest corre-

TABLE11

Correlations of Anthocyanin Concentration vs. Instrumental

Tristimulus Data for Cranberry Juice Cocktail

Correlation coefficient
Correlation of anthocy-
anin cone with: I II III

General Electric X -0.990 -0.960 -0.993

Y -0.991 -0.974 -0.981
Z -0.994 -0.939 -0.993
(X J + Y 2 )" 2 -0.991 -0.967 -0.986
(X22 + 2z'y2 -0.994 -0.949 -0.989
(Y + Z )" -0.993 -0.965 -0.992
Purity 0.991 0.966 0.966
Dominant wavelength -0.947 -0.678 -0.642
X 0.994 0.959 0.992
y -0.974 -0.949 -0.839
Gardner L -0.985 -0.966 -0.961
a 0.983 0.958 0.915
b 0.982 0.059 0.928
Hue -0.941 0.863 0.272
Chroma 0.985 0.973 0.933
Hunter L -0.992 -0.972 -0.985
a 0.992 0.975 0.973
b 0.885 -0.736 0.658
Hue -0.490 0.976 0.235
Chroma 0.990 0.973 0.978

Note: I, II, and III refer to three separate experiments having

10, 12, and 16 samples, respectively.

Reprinted with permission from Johnson, L. E., Clydesdale,

F. M., and Francis, F. J., J. FoodSci.,41, 74, 1976. Copyright
© by Institute of Food Technologists.
288 CRC Critical Reviews in Food Science and Nutrition
lation of converted data against Acy concentra-
tion. After obtaining the converted L, a*, b*
data, correlations of converted a* values vs.
visual mean rank for the corresponding experi-
ment were calculated. Table 12 shows the cor-
relations of a* vs. Acy concentration and vs.
visual rank for the two sets of data. It can be
seen that high r values were obtained for both
types of correlation.
The conclusion is that the scales generated do
indeed indicate what the eye is seeing as well as
providing an excellent measure of pigment con-
tent (r = 0.970 and 0.997). The equation gen-
erated was:
a* = 170(1.02X -Y)/Y'-60+200
This equation would probably have to be ad-
justed for different types of foods depending
upon such factors as physical state, type of pig-
ment, scattering, etc. Table 12 shows the A E
values and hue ranges for the three experiments
using the Gardner and Hunter data. High cor-
relations using conventional color functions
could be due to the relatively large differences
between the samples. It can be seen that the
range of hue for the samples is relatively small,
which could be a factor in the hue and domi-
nant wavelength functions having low correla-
tions.
It should also be pointed out that the reason
for the high correlation using conventional
techniques was that the samples utilized were
not dark enough to carry the measurements
into the area of confusion, as described by Ea-
germanetal. 'A6
This methodology is suitable to produce
more functional color scales for many foods.
Such scales might add a more versatile dimen-
sion to current colorimetric practices and qual-
ity assessment.
ESTIMATION OF TINCTORIAL
STRENGTH
The literature is becoming more replete with
comparisons of the tinctorial strength of differ-
ent types of colorants, both natural and syn-
thetic. The reasons for this are obvious. Federal
regulatory actions are providing an atmosphere
for the possible use of new colorants, both nat-
ural and synthetic, and are limiting or eliminat-
ing the use of many of the other colorants.
Since many of the natural colorants possess
a great deal of variation, due to their biological
origin and limitations on the amount of pro-
cessing which they may undergo, it becomes ex-
tremely important to specify tinctorial strength
between batches and between different pig-
ments.
Unfortunately, in the good industry, the tinc-
torial strength is being too often specified in
terms of optical density (OD) or wavelength of
maximum absorbance, A. This problem has
been noted by Johnson et al." and, more re-
cently, by Clydesdale.' However, due to the im-
portance of the proper specification of tincto-
rial strength, this problem should be discussed
briefly in this review.
Tinctorial strength or coloring power of a co-
lorant is at times a very difficult parameter to
specify. Often, optical-density readings at the
wavelength of maximum absorption are used
for dyes of similar color. Unfortunately, it has
been noted that sometimes there is a poor cor-
relation between optical density and visual
rank. Certain dyes, on the other hand, give a
good correlation between visual ranking and
optical density or correlation.
Table 13 shows the relationship of optical
density at the wavelength of maximum absorp-
tion to concentration (5 to 30 mg/1 with 5 mg
intervals) for F, D, C Yellow No. 5 (430 nm)
and D and C Yellow No. 10 (410 nm). From
this table it would appear that D and C No. 10
has greater tinctorial strength than F, D, and C
No. 5 at every concentration.
However, when these samples were presented
to a panel, the panelists did not rank them in
TABLE 12
Correlation Coefficients for Anthocyanin Concentra-
tion vs. a* and Visual Mean Rank vs. a* for Two Ex-
periments
Correlation coefficient
Correlation with a* Experiment 1 Experiment 2
Anthocyanin con- 0.970 0.997
centration
Visual mean rank -0.994 -0.961
Reprinted with permission from Johnson, L. E., Cly-
desdale, F. M., and Francis. F. J., J. FoodSci., 41, 74,
1976. Copyright © by Institute of Food Technologists.
 December 1978 289

TABLE 13

Color Differences and Hue Ranges for Three Experiments Using Gardner
and Hunter Data

Gardner Hunter

Expt. E a/b arctan a/b E a/b arctan a/b

I 16.7 0.75 3.6 40.2 5.63 4.3
II 47.3 2.37 27.7 30.6 6.30 43.3
III 59.3 2.29 13.5 50.4 5.55 52.1

Reprinted with permission from Johnson, L. E., Clydesdale, F. M., and

Francis, F. J., /. Food Sci., 41, 74, 1976. Copyright by Institute of Food
Technologists.

this order. In fact, the panelists began to rank TABLE 14

F, D, and C No. 5 as being more intense than
D and C No. 10 at a concentration level of ap- Relationship of Optical Density at the Wavelength
of Maximum Absorption to Concentration for
proximately 15 to 20 mg/1. In order to confirm F,D,C Yellow No. 5 and D and C Yellow No. 10"
this crossover effect, panelists were presented
with a sample of each at every concentration Optical density"
and asked to note which of the two in each case
was more intense. Again, the crossover effect Concentration FD and C D and C No
(mg/1) No. 5 10
was found indicating that visual rank varied
with concentration. In order to resolve this 5 0.022 0.044
problem, all samples were measured by means 10 0.052 0.095
of the GERS and Gardner XL-10®. It was 15 0.070 0.132
hoped that such an optical characterization in- 20 0.096 0.182
25 0.120 0.219
volving both spectral curves and psychophysi- 30 0.146 0.267
cal functions might explain this phenomenon.
Table 14 shows the correlation of instrumen- Wavelength of maximum absorption: F,D,C No.
tal tristimulus data with concentration of the 5 = 430 nm; D and C No. 10 = 410 nm.
two dyes. It should be noted that, although
high correlations were achieved with all psycho- From these results, it is apparent that it is
physical functions (except purity, dominant necessary for a psychophysical function to
wavelength, L, and hue), such correlations do achieve the following three objectives in order
not give any indication of the crossover effect to adequately describe the optical phenomenon
which was observed by the visual panel. In or- under investigation:
der to observe if a crossover occurred, it was
necessary to calculate lines of best fit by regres- 1. A high correlation between the function
sion analysis for each psychophysical function and dye concentration is necessary to
vs. dye concentration. When the data were plot- achieve a straight line relationship.
ted in this manner, it was found the CIE Y, 2. The crossover effect must occur at the same
Gardner chroma, purity, and b showed a cross- concentration as that noted by the visual
over. However, the latter three functions panel.
showed this effect at concentrations above 3. The function must provide data which ad-
those noted by the visual panel and/or did not equately distinguishes between the two dyes
separate the dyes adequately over the entire over the entire concentration range of in-
concentration range. terest.
Figure 24 shows the results of such a plot for
CIE Y vs. dye concentration. In this case, not Further characterization of the dyes was nec-
only was a crossover observed, but an excellent essary to explain this phenomenon. This was
separation between dyes was achieved. achieved by running spectral curves of the dyes
290 CRC Critical Reviews in Food Science and Nutrition

over the concentration range being investi- D and C No. 10 show a greater OD than F, D,
gated. Figure 25 shows spectral curves of the and C No. 5 at the wavelength of maximum ab-
two dyes at selected concentrations (5, 15, and sorption at all concentrations. However, as
30 mg/1). From these curves, it may be seen that concentration increases from 5 to 30 mg/1, the
OD becomes less important than the amount of
energy transmitted, and band width of the spec-
tra becomes more important. Thus, up to 15
mg/1, OD is the major factor in energy trans-
92- mittance, which is analogous to visual percep-
tion; and, at approximately 15 mg/1, OD and
band width are equally important, with band
91- width becoming the major factor above 15 mg/
1. In other words, OD alone is not a good esti-
mate of tinctorial strength because it does not
90-
consider band width. It should be emphasized
that visual perception is analogous to the total
area encompassed by the curves, which is
roughly the product of OD x band width. This
89-
explains why OD is not an adequate function
but does not explain why CIE Y was the best
psychophysical function to describe this phe-
88- nomenon. In order to do this, it is necessary to
10 IS 20 2S 30
realize that CIE Y is the luminosity function in
CONCENTRATION (MG/l)
the CIE system; this function has a standard-
FIGURE 24. Line of best fit obtained by regression an- observer curve which covers most of the spec-
lysis for CIE Y vs. pigment concentration. F,D, and C No. trum to some degree. Therefore, it includes
5 (—); D and C No. 10 (—). (From Johnson, L. E., Cly- brightness, which is somewhat comparable to
desdale, F. M., and Francis, F. J-, Proc. 4th Int. Congr. OD, as well as providing a measure of band
Food Science and Technology, Vol. 2, lnstituto de Agro-
quimica y Technologia de Alimentos, Valencia, Spain, width due to the shape of the y standard-ob-
1974, 150. With permission.) server curve.

i.l, S a|/|
c.<; IS a|/l
I.I: 30 Hf/I

45 760
sio (00
w»mt»ST» (••>

FIGURE 25. Spectral curves of F, D, and C Yellow No. 5 (—) and D and C Yellow
No. 10 (—) at selected concentrations. (From Johnson L. E., Clydesdale, F. M.,
and Francis, F. J., Proc. 4th Int. Congr. Food Science and Technology, Vol. 2,
lnstituto de Agroquimica y Tecnologia de Alimentos, Valencia, Spain, 1974, 150.
With permission.)

It should be stressed that this study does not
prove the need for the elimination of A in every
case nor the inclusion of a psychophysical func-
tion. However, it does show the importance of
testing whatever measure of tinctorial strength
is used against visual perception of that meas-
ure at all concentrations.
CONTINUOUS COLORIMETRY
An area which is often overlooked in the lit-
erature is that of continuous colorimetry. Kra-
mer43 has recently published an overview of
color measurement in quality control to which
the reader is referred; however, the area of con-
tinuous colorimetry was not dealt with in detail.
There should not be any question as to the de-
sirability of continuous color monitoring or
measuring systems. Such systems would give an
infinite variety of speeds to any operation and
allow control of many variables in raw mate-
rials handling at very exacting high speeds.
There is obvious interest here in quality control
applications such as those used for the sorting
of food products. There is also interest in pro-
cess control applications, for example, in the
baking industry with the manufacturing of
cookies and cakes, etc. There could also be
widespread use of this in the monitoring of bev-
erage or liquid-type products. Currently, the
most highly developed area for continuous
color measurement is probably that used in the
sorting of food products by color to upgrade
the quality." Firms such as the Icore Division
of Acurex Corporation, Sunnyvale, California
and Mandrel Industries, Houston, Texas have
developed a wide range of machines to color-
sort such products as beans, peas, grains, seeds,
rice, peanuts, walnuts, pecans, coffee beans,
cherries, dehydrated vegetables of all types,
sugar, potato chips, french fries, diced vegeta-
bles and fruits, wet-peeled potatoes, carrots,
onions and apples, etc. Each application re-
quires an optical scanning head, a sampling-
handling system to introduce the product to the
optical head, and a sample or rejection system.
All of these systems have many aspects in
common. Before discussing these common as-
pects, it might be of value to describe several
particular sorting devices which are currently
being used and then review the generalized as-
pects of all the systems. It should be noted that
December 1978 291
the systems which are described are not in any
way an endorsement of a particular company
or product but are merely representations of the
technology which is now in use.
Icore Industries has developed a Gromax
Model® 511 which is a fully transistorized
compact selector for use on all types of shelled
nuts including whole, split, blanched and
roasted kernels. A diagrammatic representation
of this sorter is shown in Figure 26. In this fig-
ure, it can be seen that the product fed to the
hopper (1) of the selector is withdrawn by a vi-
bratory feeder (2) and discharged onto two
counterrotating rollers (3). The rollers singulate
and accelerate the product. At the end of the
rollers, the stream of product passes through
the scanning head (4) between photoelectric
cells. Directly opposite each photocell are
color-standard cards. In effect, the photocells
compare each particle with these predetermined
standards. If the particle is darker or lighter in
light reflectance, an electronic impulse actuates
an air valve (5) which diverts that particle into
the reject shoot (6). All other particles are al-
lowed to continue undisturbed into the accept
shoot (7). In this case, it may be seen that only
one parameter of the total color of the material
is being used. That parameter is, of course, the
lightness/darkness factor which was described
previously in the discussion of color solids. In
this instance, however, the lightness/darkness
factor is quite important because predeter-
mined standards have been developed and the
lightness or darkness of a sample is being com-
pared to the lightness or darkness of a particu-
lar standard color rather than just a lightness/
darkness scale without consideration of the
overall color of the product.
Figure 27 depicts another type of continuous
color or quality monitoring instrument devel-
oped by Mandrel Industries, Inc. Depending
upon the size of the product examined, the fer-
rule drum picks up to 60 objects per revolution
at hundreds of revolutions per minute. Speed-
ing through the viewing chamber each object is
scrutinized and accepted or rejected. The prod-
uct is fed from overhead bins into a self-regu-
lating feed shoot which empties into a rotation
bowl, positioning the product in rows at its
outer edge.
The ferrule drum, revolving counterclock-
wise, dips into the first row of objects, picks up
 292 CRC Critical Reviews in Food Science and Nutrition

SCANNING
HEAD 5
FEEOER| AIRVALVE

REJECT ACCEPT

FIGURE 26. A diagrammatic representation of a Gromax

Model® 5111 Compact Selector for use on all types of
shelled nuts. (Courtesy of Icore Industries.)

FILTERS
/
PHOTOELECTRIC^^ M1RRORS AMPLIFIERS
LIGHT SOURCE PRODUCT FEED HOPPER

PRESENTATION
- WHEEL
CLASSIFIER

REJECTED PRODUCT

' * * *

ACCEPTED PRODUCT AIR EJECTOR

HOLDING FERRULES

FIGURE 27. A diagrammatic representation of a continuous sorter. (Courtesy of

Geosource, Inc., Houston, Texas.)

by vacuum one object on each ferrule, and
moves it through the viewing chamber for indi-
vidual judgment from both sides. Judgment is
based on comparing the color of the objects
with that of the standard background, through
the use of color filters, precision optics, and
phototubes. An air blast ejects all specimen that
do not conform to the preset standards. The ac-
cepted product is released onto a transport belt.
In this system it may be seen that filters, instead
of predetermined standards, are being used to
set quality parameters, and this device is as
equally acceptable as the one described previ-
ously.
A sorting system that inspects and grades
large objects suspended in water has been de-
veloped by the Electric Sorting Machine Divi-
sion of Mandrel Industries, a division of Am-
pex Corporation, and applied to the processing
of potatoes by J. R. Simplot Company, Cald-
well, Ohio.2 As is shown in Figures 28 and 29,
potatoes to be sorted pass through a clear plas-
tic shoot located in the center of a viewing
chamber. The viewing chamber is surrounded
by nine viewing stations, each containing a pho-
toelectric cell. When the area is flooded with
light, the light is reflected from a color refer-
ence reflector disk into the photocell. This es-
tablishes the reference reflectance.
Perfectly peeled potatoes match the reflec-
tors in color. The interval between breaks in re-
flectance represents the length of each potato
and sensing circuits classify it according to
length. A blemished or discolored potato does
not meet the color standard, tending to absorb
rather than reflect the light, which triggers a de-
 December 1978 293

Via wing
One of nine cok)r
De*otering
tank - references

-Tronsparent chute

Start ^

Peeler

FIGURE 28. A diagramatic representation of a sorting system that inspects and

grades large objects in water.

line into the trimming area. The acceptable po-

Color
reference
Front view
Nine viewing stations
FIGURE 29. A front view of the nine viewing stations
shown in Figure 3.
lay circuit used for ejection. Based on the
length of the blemished potato, a memory cir-
cuit determines which two ejectors will be
aligned with the exact center of the potato when
it reaches the ejector chamber at the end of the
shoot. At the instant that the blemished prod-
uct passes in front of the chosen ejector, the
delay circuit triggers a valve and a burst of air
from the ejector blasts the selected item off the
tatoes are sent along for final processing. From
these examples, it may be seen that measuring
devices used in continuous colorimetry have
some aspects in common.
In general, the sample introduction system
can be a simple trough or belt, a free-flowing
fall mechanism, or even a wheel with ferrules
to pick up individual pieces by vacuum and
convey them individually through the scanning
head. All are capable of considerable through-
put, depending on the size of the samples. For
example, wet-peeled potatoes up to 3 in in size
can be sorted by free fall at rates up to 3 ton/
hr. Blanched almonds can be handled at rates
up to 850 Ib/hr; peanuts, 1215 Ib/hr; and green
coffee beans, 650 Ib/hr.
The rejection system for these machines is
usually a jet of compressed air on a delay-signal
mechanism. In the case of machines with indi-
vidual sample pickups, a mechanical displace-
ment system can be used to separate the unac-
ceptable samples.
The optical scanning head for these color
sorters can be very simple or rather compli-
cated, depending on the sample. With most
products, a simple measurement of lightness or
darkness will suffice. The reflectance of the
sample is measured by one or more optical cells
 294 CRC Critical Reviews in Food Science and Nutrition
and is compared with a standard. Samples reg-
istering too dark, such as peeled potatoes with
blemishes, would be rejected. A simple refine-
ment would be to have a rejection signal acti-
vated for samples that are either too dark or
too light so that the product stream would be
divided into two portions.
A more complicated refinement would be to
introduce a color signal based on a function of
Munsell hue as well as value. This would re-
quire a double signal but certainly could be ar-
ranged.
Such machines have been employed for green
coffee beans, shelled peanuts, peas, beans, de-
hydrated vegetables, etc. The Electric Sorting
Machine Company Division of Mandrel Indus-
tries makes equipment to sort whole tomatoes
and lemons into five color grades. The Pack-
aging Machinery Division of FMC Corpora-
tion, San Jose, California makes equipment to
sort apples into four color grades at the rate of
50,000 apples per hour.
As in batch-type tristimulus colorimetry, the
choice as to whether to use a single function or
a multiple function will depend on analysis of
a particular color problem. Nearly all food
products differ in all three parameters of color;
however, not all of these are usually required
for sorting purposes. Usually one or more pa-
rameters correlate reasonably well with each
other, and the color sorting process can be sim-
plified within an acceptable range of error.
It should be remembered that if a fairly com-
plicated signal is required, reasonably expensive
optical and electronic instrumentation is also
required. However, the development of minia-
turized solid-state electronics has enabled man-
ufacturers to reduce the prices considerably in
recent years, and therefore, researchers can
look forward to more efficient and possibly
even less expensive equipment in the future.
This particular aspect of price vs. complexity is
quite common in all areas of instrumentation
and, in particular, when setting color toler-
ances. The closer the tolerances desired for any
given usage will increase the price commensur-
ately. Therefore, it is up to the customer and
instrument manufacturer to decide upon the de-
gree of error which is acceptable in their partic-
ular case.
As pointed out previously, other areas of in-
terest in continuous color measurement concern
process control. In the baking industry, the
proper degree of color obtained depends on
several factors, but an important one is the
time-temperature conditions in the oven. The
speed of the conveyor belts and the temperature
of the oven can be controlled by color signals
from the product; this would require a closed-
loop feedback system and considerable engi-
neering of the whole baking line. Apparently,
there has not been any commercial installations
of this design as of yet but the concept is cer-
tainly quite feasible.
The control of peanut butter color, as influ-
enced by the roasting temperature, is a similar
problem. The control of color in the breakfast
cereal industry is another problem with similar
difficulties; however, in this case sampling is
extremely difficult. These examples would all
require an optical head suspended over the con-
veyor belt to measure lightness or darkness.
Other applications may also require an addi-
tional signal based on a function of hue. Such
equipment is manufactured by Hunter Associ-
ates Laboratory, and Instrument Development
Laboratories Division of Kollmorgan Color
Systems, Attleboro, Massachusetts.
The concept of using continuous color meas-
urement by transmission for sorting opaque
samples such as apples, potatoes, peaches, to-
matoes, etc. is being developed by Birth and
Norris7 at the USDA Laboratory in Beltsville,
Maryland. The systems described up to this
point have relied on reflectance colorimetry and
one does not normally think of using transmis-
sion colorimetry on opaque or paniculate sub-
stances, such as fruits and vegetables. These
products require optical signals up to 10 optical
density ranges (wide-range spectrophotometry),
but the concept is feasible. An experimental
machine for quality sorting of apples is being
developed by Neotec Instruments, Rockville,
Maryland according to USDA specifications.
This kind of transmission sorting of opaque
objects is usually referred to as "quality" sort-
ing rather than "color" sorting, since wave-
lengths outside the visible region may be used.
For example, the apple sorter employs wave-
lengths at the red end of the visible spectrum
and the near infrared to sort for water-core (760
and 810 nm), maturity (cholorphyl 690 and 740
nm), and internal browning (720 and 810 nm).
The automatic egg candler employs energy at

577.and 597 nm to detect blood spots in eggs.
Green rot in eggs can be detected by fluores-
cence at 490 and 510 nm. Cholorphyl content
as an index of maturity in peaches can be mea-
sured at 700 and 740 nm. Mold damage in yel-
low corn can be measured at 800 and 950 nm.
Carotenoid content as an index of maturity in
peanuts can be measured at 490 and 520 nm.
The actual position of the signal of the elec-
tromagnetic spectrum is not important for sort-
ing purposes. Almost any signal that will differ-
entiate products into acceptable and
nonacceptable catagories can be the basis for
equipment design.
ANALYSIS AND DESIGN OF A NEW
SYSTEM
Thus far our discussion of color has centered
on the theories and concepts of colorimetry as
applied to the establishment of continuous co-
lorimetry and also the current methodology
available in this area. However, it has not es-
tablished the type of experimental design and
considerations which must be taken into ac-
count when one is dealing with a new product
and attempting to establish a continuous moni-
toring system for this product. Therefore, it
might be worthwhile to look at a particular sys-
tem where continuous colorimetry does not ex-
ist and trace the development of such a system
from the concept through completion. The sys-
tem which is going to be discussed is that of the
design of a continuous monitoring system for a
dark-colored beverage, in particular cranberry
cocktail. Staples and Francis68 investigated the
possibility of using wide-range spectrophoto-
metry for the measurement of cranberry cock-
tail. Therefore, their experimental results will
be used in an attempt to describe the use of such
results in the establishment of a continuous
color measurement line for a dark-colored bev-
erage.
Cranberry cocktail owes much of its popular-
ity to its attractive red color. Quality control
procedures to insure optimum color involves
continuous sampling of the product both prior
and subsequent to processing. This is a slow
and rather laborious technique which has to
stop production for a short time before correc-
tions can be made and, therefore, would be an
ideal system in which to establish a continuous
December 1978 295
monitoring line which would allow a feedback
system to correct color deviations.
In order to begin such a project, one must
take into consideration the four fundamental
procedures in any colorimetric technique which
were described previously in the section on
Generalized Application of Colorimetry. That
is, one must consider the object to be measured,
the viewing technique, the analytical proce-
dures to be used, and the establishment of the
relationship between subjective viewing tech-
niques and analytical procedures.
At present, it is not possible to use conven-
tional tristimulus colorimetry for on-stream
color control of cranberry cocktail, unless a
very thin cell is used. It would have to be of the
order of 1 cm thick or less. To use a path width
of 3 cm to make installation in a line practical,
the color has to be measured by spectrophoto-
metry. The use of absorption indices reflecting
the pigments in a given product have been used
in several applications.
Therefore, in an initial investigation of the
product, it may be seen that there is a transpar-
ent material which is fairly dark in color and,
therefore, requiring a high input of energy with
an excellent sensing device in order to obtain a
read-out. On the plus side, the material in ques-
tion does not contain particulate matter; there-
fore, one does not have to worry about scatter-
ing and the subsequent homogeniety or random
aspects of particles suspended in the juice ma-
trix. Therefore, for the first piece of informa-
tion which is required, it may be decided that
spectrophotometry can be effectively used for
measuring this kind of material. A model sys-
tem then has to be built in order to investigate
the other aspects of colorimetry which must be
considered prior to the development of a con-
tinuous measuring system or even a batch-type
system, for that matter. This may be done by
utilizing cranberries or any other material un-
der investigation in order to obtain a wide range
of color. In this case, cranberries were obtained
and sorted visually into four color categories,
white (A), pink (B), red (C) and dark red (D),
and then combined in various proportions in
order to produce a series of juices covering a
wide color range. In all, there were 48 combi-
nations of berries used, and these produced
enough combinations to cover the entire color
gamut which might be found in industrial prac-
296 CRC Critical Reviews in Food Science and Nutrition
tice. The samples prepared were processed into
cranberry juice cocktail. The pigment content
of the berries, the pomace (press cake) and
cocktail samples were determined by usual lab-
oratory procedures. These data were important
to show that on-stream measurement of pig-
ment content could be interpreted as a reliable
measure of the visual color of the juice.
Three different techniques were used to ob-
tain color data on the juice samples:
1. A general-purpose Biological Spectropho-
tometer (Biospect® effective over eight op-
tical densities, was used to obtain an ab-
sorption index of the samples as two points
on the optical curve, optical density 515 to
420 nm. This provided a basic measure of
the pigment content of the juice on the as-
sumption that there is close correlation be-
tween pigment content and visual color.
2. A GE Recording Spectrophotometer was
chosen as the referee instrument and the
values obtained from it were designated as
standard data for objective measurement
of color. A transmittance curve and X, Y,
Z values from a Davidson and Hemmen-
dinger Tristimulus Integrator® were Ob-
tained for each sample. The transmittance
values were changed to OD, and OD (515
to 420 nm) values were calculated as an in-
dex color.
3. A Lovibond Schofield Tintometer® was
the instrument used for subjectively judg-
ing the color of the cranberry cocktail sam-
ples. This is an instrument based on visual
judgment, and thus, it satisfied another
one of our criteria.
Therefore, it may be seen that the experimen-
tal design provides information on the chemical
and physical nature of the sample, methodol-
ogy for testing the viewing conditions and the
visual assessment of the sample, and method-
ology for testing objective measurements of the
sample. These may then be correlated to insure
that the analytical read-out is providing the in-
formation on both the visual assessment of the
sample and the actual pgiment concentration.
In addition, there is Biospect, a wide-range
spectrophotometer which can be applied to an
on-line continuous colorimetric technique.
From the results obtained by Staples and
Francis," the plot of most interest and impor-
tance was the Biospect difference OD (515 -
420 nm) vs. the total pigment content of the
cranberry cocktail sample (Figure 30). The plot
showed concave downward effect, and a "tail-
ing off" or clustering effect at three distinct lo-
cations on the graph. This, in fact, was due to
the different cell sizes and dilutions necessary
to obtain proper cranberry cocktail spectra.
The group at the high end of the graph arose
as a result of obtaining spectra of very dark
samples through a 4-cm cell with a 1:2 water
dilution. With lighter samples the clustering of
points disappeared and the "tailing off" effect
was shown to be strictly an instrumental prob-
lem and could be remedied by standardizing the
cell size and obtaining a stronger light source -
filter tube combination. Also, the Biospect em-
ployed was a very early model, and the base line
was not completely flat. This has been remedied
in later models by the manufacturer.
Figure 31 showed that the Biospect difference
plotted against Tintometer R (red) gave a rela-
tively straight-line positive slope. This indicated
that the objective values of the biospect should
correlate quite highly with the subjective red-
ness values of the Tintometer.
Simple correlation coefficients of Biospect
data with data from other instruments and with
total pigment content were high, with the ex-
ception of the coefficient obtained by relating
Biospect difference with Tintometer Y (yellow).
The multiple correlations were only slightly
higher than the simple correlations. Higher val-
ues for both the simple and multiple correla-
tions would probably result if the clustering ef-
fect were eliminated.
90
• *
7.0
3
2 5.0
J-
3 3.0 y^ - •
D
O
1.0
O.O2 0.06 0.10 0.14 0.18
TOTAL PIGMENT CONTENT (nvj C.R./ml)
FIGURE 30. Total pigment content of cranberry juice
cocktail plotted against OD difference as obtained with the
Biospect.®
 December 1978 297

8 « .— objective measurement was giving a good indi-

cation of visual appearance. A plot of Tinto-
5 6 meter R vs. total pigment content of the cock-
a tail (Figure 33) resulted in a curve concave
j j
5 downward similar to that of Figure 30 without
5 the clustering effects. A function of chromas
a was calculated from Tintometer data and plot-
d 4
ted against the pigment content of the cocktail.
3 A curve similar to Figure 33 was obtained, in-
10 15 20 25 30 35 dicating that Tintometer Y had little effect on
the color.
A plot of total pigment content of the cock-
FIGURE 31. Redness of cranberry juice cocktail, as ob- tail against the OD differences (Figure 34)
tained with a Lovibond Tintometer® plotted against OD
difference obtained with a Biospect.® yielded a line slightly concave downward with
no clustering, since one cell size was used and
the GE instrument has excellent baseline line-
1.8 arity. The high correlation indicated that the
spectrum was reflecting the pigment content ac-
_ ' 4
curately.
ui From this data, it seems that we have an ob-
2 1.0 jective measurement which indeed does give an
acceptable and accurate measurement of both
06 pigment content and visual assessment of a
sample. Therefore, the GE readout might be
Q2 considered as a possibility to place into produc-
tion as a continuous on-line monitoring device.
10 15 20 25 30 35
The use of two points on an absorption curve
TINTOMETER " R "
of fruit juice, as an index of the color of the
FIGURE 32. Redness of cranberry juice cocktail, as ob- juice, may be described as abridged spectropho-
tained with a Lovibond Tintometer® plotted against OD tometry. Obviously, a true index of color can
difference as obtained with a General Electric Spectropho- only be obtained from a consideration of the
tometer.®
absorption curve through the whole visual spec-
trum. The appropriateness of using two points
36
instead of the whole curve depends on how ac-
curately a two-point index such as OD (515 to
0=26
420 nm) correlates with the visual appearance.
The two-point index correlates well with visual
color; a correlation coefficient of 0.987 was ob-
16
tained when a GE Spectrophotometer was used
for the two-point index and a Lovibond Tinto-
meter was used for the visual index. If there are
factors which can change the visual color with-
0.02 0.06 0.10 0.14 018
out effecting the OD (515 to 420 nm) index,
TOTAL PIGMENT CONTENT (mg C.R./ml)
they were not present in the 48 samples used in
FIGURE 33. Total pigment content of cranberry juice this work.
cocktail plotted against redness as obtained with a Lovi- If the concept of an OD (515 to 420 nm) in-
bond Tintometer.®
dex is accepted as a good index of color, the
problem of measuring it through a cell in a
pipeline becomes one of extending the spectro-
A plot of GE OD differences against Tinto- photometric capabilities. With the general pur-
meter R (red) (Figure 32), produced a straight pose instrument (Biospect) used in this work,
line with high correlations, indicating that the which is capable of obtaining the above index
298 CRC Critical Reviews in Food Science and Nutrition
16
1.2
t / ^
2 0.8 • >^
t • j / ^
s j/+ • m
a
b 0.4
002 0.06 0.10
TOTAL PIGMENT CONTENT (mg C.R./ml)
FIGURE 34. Total pigment content of cranberry juice
cocktail plotted against OD difference as obtained with a
General Electric Spectrophotometer.®
over a range of eight ODs, the correlations with
visual color were very high (R = 0.893) but less
than with the GE instruments (R = 0.987). The
lower correlations were probably due to the
limitations of the general-purpose instrument.
A specific-purpose instrument, employing two
interference filters instead of a grating, could
be built with greater range and more accuracy.
Applications of this theory to many food stuffs
other than fruit juices have been previously
mentioned.7
A schematic for such a system may be seen
in Figure 35. This system shows how on-stream
colorimetry and density control could be ap-
plied in cranberry juice processing. The on-line
spectrophotometer would have a feedback sys-
tem to a color concentrate which would then
increase or decrease the degree of color re-
quired. A refractometer control could be in-
stalled on-line such that the sugar concentration
could be increased or decreased as required.
Building such a special-purpose instrument
does not present serious technical problems.
Because of its relative simplicity, it would not
be considerably expensive. One instrument
builder estimates it could be built for approxi-
mately $5000.
It is hoped that this latter section has indi-
1. Adams, E.Q..X-Z planes in the 1931 ICI system of colorimetry, J. Opt. Soc. Am., 32, 168, 1942.
2. Anon., Devices and systems at work — electro optical sorter checks for surface defects, Control Eng., p. 77, January
FIGURE 35. A schematic representation of a design uti-
lizing on-line colorimetry.
0.14 0.18
cated that many considerations must be in-
volved prior to the design of on-line continuous
colormetric techniques. It cannot be overem-
phasized that one needs visual assessment, a
very good idea of the physical and chemical pa-
rameters of a product, and the proper instru-
mentation for analytical procedures so that all
three of these may be interrelated to produce a
reliable final design.
CONCLUSIONS
There is no question that the necessary tech-
nical knowledge and sophistication are avail-
able for the design and use of increasingly so-
phisticated colormetric techniques. I sincerely
believe that with the constant questioning of the
colorant systems being utilized in our food sup-
ply today, we will have to take advantage of
every system available, utilize more and more
pigments and more efficient extraction proce-
dures from raw materials, and look more
closely at the color monitoring systems that are
currently available. The time is long past when
color can be ignored since limitations are going
to be placed on the use of the coloring materials
that are employed. Therefore, the problem re-
solves itself, not only to one of finding new,
more acceptable coloring materials but also to
finding and developing more acceptable meth-
ods of color control. Exact tolerances may then
be set and met, and the industry may come up
with techniques which maximize color without
overuse or abuse of allowable colorants.
REFERENCES
 December 1978 299

3. Atkins, J. T. and Billmeyer, F. W., Jr., On the interaction of light with matter. Color Eng., 6(3), 40, 1968.
4. Billmeyer, F. W., Jr., Fundamental processes leading to the appearance of foods. Objective methods for food evalu-
ation — a symposium, U. S. Army Natick Research and Development Command, Committee on Food Stability and
Personnel Supplies, National Academy of Science/National Research Council Washington, D.C. 1976, 229.
5. Billmeyer F. W., Jr., Beasly, J. K., and Sheldon, J. A., Color-order system predicting constant hue, J. Opt. Soc. Am.,
51, 656, 1961.
6. Birth, G. S., The light scattering properties of foods, J. Food Sci., in press.
7. Birth, G. S. and Norris, K. H., The difference meter for measuring internal quality of foods and pigments in biological
tissues, U.S. Dept. Agric. Technical Bulletin, No. 1341, 1965.
8. Breckenridge, F. C. and Schaub, W. R., Rectangular uniform-chromaticity-scale coordinants, J. Opt. Soc. Am., 29,
370, 1939.
8a. CIE, International Comm. on Illumination, Proceedings of the 8th session, Cambridge, England, 1931; cited by Cly-
desdale, F. M. and Francis, F. J., Color Measurement of Foods. VIII. Development of the CIE system. Food Prod.
Dev., 2, 50, 1968.
9. Clydesdale, F. M., Color Measurement, paper presented at Proc. Symp. on Colorants in Food held at the Annual
American Chemical Society Meeting, New Orleans, La., March 1977.
10. Clydesdale, F. M., Instrumental techniques for color measurement of foods, Food Technol. (Chicago), 30(10), 52,
1976.
11. Clydesdale, F. M., Measuring the color of foods, Food Technol. (Chicago), 26,45, 1972.
12. Clydesdale, F. M. and Francis, F. J., Colorimetry of foods. I. Correlation of raw, transformed and reduced data with
visual rankings for spinach puree, J. Food Sci., 34, 349, 1969.
13. Clydesdale, F. M. and Podlesny, C. H., Jr., A computer program for the interconversion of color data, Color Eng.,
6, 3, 55, 1968.
14. Eagerman, B. A., Orange juice color measurement using general purpose tristimulus colorimeters, program of the
37th Annual Meeting of the Institute of Food Technologists Philadelphia, Pa., June 5 - 8 , 1977.
15. Eagerman, B. A., Clydesdale, F. M., and Francis, F. J., Comparison of color scales for dark colored beverages, J.
Food Sci., 38, 1051, 1973.
16. Eagerman, B. A., Clydesdale, F. M., and Francis, F. J., Development of new transmission color scales for dark
colored beverages, J. Food Sci., 38, 1056, 1973.
17. Edwards, G. F., Wenzel, F. W., Haggart, R. L., and Barron, R. W., Comparison of subjective and objective methods
for determining the color of reconstituted frozen concentrated orange juice, Proc. Fla. State Hortic. Soc., 79, 321,
1966.
18. Francis, F. J., The origin of tan a/b, J. Food Sci., 40, 412, 1975.
19. Francis, F. J., Continuous color measurement, Food Technol. (Chicago), 26(12), 48, 1972.
20. Francis, F. J., A method of measuring the skin color of apples, Proc. Am. Soc. Hortic. Sci., 60, 213, 1952.
21. Francis, F. J. and Clydesdale, F. M., Food Colorimetry: Theory and Applications, AVI Publishing, Westport, Conn.,
1975.
22. Francis, F. J. and Clydesdale, F. M., Color measurement of foods. XIV. Color scales. Food Prod. Dev., 3(6), 117,
1969.
23. Fuleki, T. and Francis, F. J., Quantitative methods for anthocyanins. I. Extraction and determination of total antho-
cyanins in cranberries, J. Food Sci., 33, 72, 1968.
24. Glasser, L. G., McKinney, A. H., Reilly, C. D., and Schnelle, P. D., Cube-root color coordinate system, J. Opt. Soc.
Am., 48, 736, 1958.
24a. Glasser, L. G. and Tory, D. J., A new high sensitivity differential colorimeter, J. Opt. Soc. Am., 42, 652, 1952.
25. Grassmann, H., Zur Theorie der Farbenmischung, Ann. Phys. Chem. 89, 69, 1853; cited in Billmeyer, F. W., Jr. and
Saltzman, M., Principles of Color Technology, Interscience, New York, 1966, 31.
26. Gullett, E. A., Francis, F. J., and Clydesdale, F. M., Colorimetry of foods: IV. Orange juice, J. Food Sci., 37, 389,
1972.
27. Gullett, E. A., Francis, F. J., and Clydesdale, F. M., Colorimetry of foods. 5. An orange beverage — Tang, Can.
Inst. Food Sci. Technol. J., 5(1), 32, 1972.
28. Huang, I-Lo, Francis, F. J., and Clydesdale, F. M., Colorimetry of Foods. II. Color measurement of squash using
the Kubelka-Munk concept, J. Food Sci., 35, 315, 1970.
29. Huang, I-Lo, Francis, F. J., and Clydesdale, F. M., Colorimetry of foods. III. Carrot puree, J. Food Sci., 35, 771,
1970.
30. Hunter, R. S., The Measurement of Appearance, John Wiley & Sons, New York, 1975.
31. Hunter, R. S., Color difference systems and concepts, Hunter Lab Reflections and Transmissions, Part 3, Hunter
Associates Lab Inc., Fairfax, Va., 1973.
32. Hunter, R. S., personal communication to the University of Massachusetts Food Science Color Laboratory, Amherst,
Mass., 1969.
33. Hunter, R. S., Development of the citrus colorimeter. Food Technol. (Chicago), 21, 906, 1967.
34. Hunter, R. S., In-plant memo regarding the development of the L, a', b' color scales, Hunter Associates Lab, Fairfax,
Va., 1967.
300 CRC Critical Reviews in Food Science and Nutrition

35. Hunter, R. S., Photoelectric tristimulus colorimetry with three filters, Nat. Bur. Stand. (U.S.) Circ. 429; cited in J.
Opt. Soc. Am., 32, 509, 1942.
35a. Hunter, R. S., Photoelectric color-difference meter, J. Opt. Soc. Am., 38, 661, 1948.
36. Hunter, R. S. and Yeatman, J. N., Direct reading tomato colorimeter, J. Opt. Soc. Am., 51, 552, 1961.
37. Johnson, L. E., Clydesdale, F. M., and Francis, F. J., Use of expanded color scales to predict chemical and visual
changes in solutions, J. Food Sci., 41, 74, 1976.
38. Johnson, L. E., Clydesdale, F. M., and Francis, F. J., The use of colorimetric data to predict chemical and visual
changes in solutions, Proc. 4th Int. Congr. Food Science and Technology, Vol. 2, Instituto de Agroquimica y Tecno-
logia de Alimentos, Valencia, Spain, 1974, 150. (With permission.)
39. Joslyn, M. A. and Little, A., Relation of types of concentration of phenolics to the color and stability of rose wines,
Am. J. Enol. Vitic, 18, 138, 1967.
40. Judd, D. B., Colorimetry, Nat. Bur. Stand. (U.S.) Circ. C 478, 1950.
41. Judd, D. B., A Maxwell triangle yeilding uniform chromaticity scales, J. Opt. Soc. Am., 25, 24, 1935.
42. Judd, D. B. and Wyszecki, G., Color in Business Science and Industry, John Wiley & Sons, New York, 1963.
43. Kramer, A., Use of color measurement in quality control of foods, Food Technol. (Chicago), 30(10), 62, 1976.
44. Law, S. E. and Norris, K. H., Kubelka-Munk light-scattering coefficients of model paniculate systems, Trans. Annu.
Meet. Soc. Agric. Eng., 16(5), 914, 1973.
45. Little, A. C., Physical measurements as predictors of visual appearance, Food Technol. (Chicago), 30(10), 74, 1976.
46. Little, A. C., OFF on a tangent, J. Food Sci., 40, 410, 1975.
47. Little, A. C., Color evaluation of foods — correlation of objective facts with subjective impressions, in Sensory
Evaluation of Appearance of Materials ASTMSTP 545, American Society of Testing and Materials, Washington,
D.C., 1973, 109.
48. Little, A. C., Color measurement of translucent food samples, J. Food Sci., 29, 782, 1964.
49. Little, A. C. and Liaw, M. W., Blending wines to color, Am. J. Enol. Vitic, 25, 79, 1974.
50. Little, A. C. and MacKinney, G., Colorimetry of food — the sample as a problem, Food Technol., 23, 25, 1969.
51. MacAdam, D. L., Specification of small chromaticity differences, J. Opt. Soc. Am., 33, 18, 1943.
52. MacAdam, D. L., Projective transformations of the I.C.I, color specifications, J. Opt. Soc. Am., 27, 294, 1937.
53. MacKinney, O., Little, A. C., and Brinner, L., Visual appearance of foods, Food Technol. (Chicago), 20, 300, 1966.
54. Moon, P. and Spencer, D. E., A metric for colorspace, J. Opt. Soc. Am., 33, 260, 1943.
55. Moon, P. and Spencer, D. E., A metric based on the composite color space, J. Opt.Soc. Am.,33,270, 1943.
55a. Munsell, A. H., A New Classification of Color. 1905. Transactions of the New England Cotton Manufacturers Assoc.,
No. 78, April 26, 1905.
56. Color in foods — a Symposium, Advisory Board on Quartermaster Chair of Research and Development National
Academy of Sciences, National Research Council, Washington, D.C., May 1954.
57. Newton, Isaac, Optiks, or a Treatise of the Reflections, Refractions, Inflections and Colours of Light, reprint based
on the 4th ed., London, 1730 by Dover, New York, 1952.
58. Pauli, H., Proposed extension of the CIE recommendation on Uniform color spaces, color difference equations, and
metric color terms, J. Opt. Soc. Am., 66, 866, 1976.
59. Petrus, D. R. and Dougherty, M. H., Spectrophotometric analyses of orange juices and corresponding orange pulp
washes, J. Food Sci., 38, 913, 1973.
60. Petrus, D. R., Huggart, R. L., and Dougherty, M. H., Correlation of carotenoid visible absorbance and numerical
color score of orange juice, J. Food Sci., 40, 922, 1975.
61. Ponting, J. D., Sahshuek, D. W., and Brekke, J. E., Color measurement and deterioration in grape and berry juices
and concentrate. Food Res., 25, 471, 1960.
62. A Catalog of Color Measuring Instruments and a Guide to their Selection, preliminary Report of ISCC Subcommittee
for Problem 24, Inter-Society Color Council, April 11, 1970.
63. Richter, M., The of ficial German standard color chart, J. Opt. Soc. Am., 45, 223, 1955.
64. Robinson, W. B., Bertino, J. J., and Whitcombe, J. E., Objective measurement and specification of color in red
wine, Am. J. Enol. Vitic., 17, 118, 1966.
65. Robinson, W. B., Ransford, J. R., and Hand, D. B., Measurement and control of color in (he canning of tomato
juice. Food Technol. (Chicago), 5, 314, 1951.
66. Robinson, W. B., Wichnetsky, T., Ransford, J. R., Clark, W. L., and Hand, D. B., A study for the measurement of
tomato juice color. Food Technol. (Chicago), 6, 269, 1952.
67. Saunderson, J. L. and Milner, B. I., Modified chromatic value color space, J. Opt. Soc. Am., 36, 36, 1946.
67a. Scofield, F., A method for determination of color differences, Natl. Paint. Varnish, Laquer Assoc. Sci. Ser. Circ.
664, 1943.
68. Staples, L. C. and Francis, F. J., Colorimetry of cranberry cocktail by wide range spectrophotometry. Food Technol.
(Chicago), 22(5), 77, 1968.
69. Swain, T. and Hillis, W. E., Phenolic constituents of Primus domestica. I. Quantitative analysis of phenolic constitu-
ents, J. Sci. Food Agric, 10, 63, 1959.
70. Technical Notes. CIE colorimetry committee — working program on color differences, J. Opt. Soc Am., 64, 896.
1974.
 December 1978 301

71. U. S. standards for grades of processed fruits and vegetables and certain other products, USDA/AMS, U. S. Depart-
ment of Agriculture, Fruit and Vegetable Division, Washington, D. C., 1973.
72. USDA instructions for inspection of canned tomatoes, USDA/AMS, U. S. Department of Agriculture, Fruit and
Vegetable Division, Washington, D. C., 1968.
73. Van Buren, J. P., Hrazdina, G. and Robinson, W. B., Color of anthocyanin solutions expressed in lightness and
chromaticity terms. Effect of pH and type of anthocyanin, J. Food Sci., 39, 325, 1974.
74. Watada, A. E. and Abbott, J. A., Objective method of estimating anthocyanin content for determining color grade
of grapes, J. Food Sci., 40, 1278, 1975.
75. Wyszecki, G., Proposal for a new color difference formula, J. Opt. Soc. Am., 53, 1318, 1963.
76. Yeatman, J. N., Food color standards in the federal government. Objective methods for food evaluation — a sympos-
ium, U. S. Army Natick Research and Development Command, Committee on Food Stability and Personnel Supplies,
National Academy of Science/National Research Council, Washington, D. C., 1976, 239.
77. Yeatman, J. N., Tomato color standards need revision, Food Technol. (Chicago), 30(11), 50, 1976.