Journal of the

OPTICAL SOCIETY
of AMERICA
VOLUME 32 MAY, 1942 NUMBER 5

Visual Sensitivities to Color Differences in Daylight*

DAVID L. MACADAM
Research Laboratories,Eastman Kodak Company, Rochester,New York
(Received January 7, 1942)

An apparatus is described which facilitates the presenta- straight lines in the chromaticity diagram, and also as
tion of pairs of variable colors without variation of lumi- functions of direction of departure from points representing
nance. With this instrument, various criteria of visual certain standard chromaticities. Such representations are
sensitivity to color difference have been investigated. The simpler than the traditional representations of wave-
standard deviation of color matching was finally adopted length thresholds and purity thresholds as functions of
as the most reproducible criterion. The test field was two wave-length, and the accuracy of the representations is
degrees in diameter, divided by a vertical biprism edge, improved by this simplicity. Chromaticity discrimination
and was viewed centrally with a surrounding field of forty- for non-spectral colors is represented simultaneously and
two degrees diameter uniformly illuminated so as to have on the same basis as for spectral colors. Small, equally
a chromaticity similar to that of the I.C.I. Standard noticeable chromaticity differences are represented for all
Illuminant C (average daylight). The luminance of the chromaticities and for all kinds of variations by the lengths
test field was maintained constant at 15 millilamberts, and of the radii of a family of ellipses drawn on the standard
the surrounding field was 7.5 millilamberts. These fields chromaticity diagram. These ellipsescannot be transformed
were viewed monocularly through an artificial pupil, 2.6 into equal-sized circles by any projective transformation
mm in diameter. Over twenty-five thousand trials at color of the standard chromaticity diagram. The consistency of
matching have been recorded for a single observer, and these data with the results of other investigators is ex-
the readings are analyzed in detail and compared with hibited in terms of the noticeabilities of wave-length
previously available data. The standard deviations of the differences in the spectrum and of the noticeabilities of
trials are represented in terms of distance in the standard purity differences from a neutral stimulus, as functions of
1931 I.C.I. chromaticity diagram. These increments of dominant wave-length.
distance are represented as functions of position along

THE measurement and specification of color specified in a manner which indicates immedi-
T is a well-established technique. Familiarity ately whether or not several samples having
with the fundamentals of this subject will be different spectral reflectance characteristics ap-
assumed throughout this report. 1 Color specifica- pear alike under certain conditions of illumina-
tions reveal whether or not two sources radiating tion. These specifications give precise meaning
different spectral distributions appear to have to color standards. The specifications of color
the same color, for the average human observer. standards can be recorded, and can be communi-
Similarly, the colors of reflecting materials are cated, in the form of quantitative values which
are reproducible at will in all adequately equipped
* Communication No. 840 from the Kodak Research laboratories. Color standards need no longer be
Laboratories. dependent upon the preservation of material
1 A. C. Hardy and collaborators, Handbook of Colorimetry
(Technology Press, Cambridge, Massachusetts, 1936). samples of questionable permanence.
247
248 DAVID L. MAcADAM

The standard specifications of color also the relations exhibited between locations in this
provide a useful system for the classification of diagram are limited in significance. In a certain
colors. The luminance 2 of a source of radiant limited sense, differences of specifications corre-
energy, or of an object reflecting radiant energy, spond in order of magnitude to the apparent
is the most familiar of the quantities in terms differences of colors. For example, if two ad-
of which colors may be specified and classified. jacent objects having the same dominant wave-
Alternatively, luminous reflectance may be used length and luminance differ in purity, the object
to specify the ratio between the luminance of an having the greater purity will appear more
object and the luminance of a perfect diffuse saturated. An estimate of the ratio of the ap-
reflector similarly illuminated. This ratio is parent saturations may not confirm the ratio
characteristic of the reflecting surface, and con- of the measured purities, but equality, or the
sequently luminous reflectance is a convenient fact of greater or less purity will be confirmed
specification when the color of an object is to be subjectively. Similarly, two adjacent objects
discussed without reference to the illuminance2 having the same luminance and purity but
incident upon it. differing in dominant wave-length will differ
The chromaticity, or quality of a color may in the subjectively appreciated quality called
be specified and classified in terms of dominant hue, in the same sense as they differ in dominant
wave-length and purity.' Alternatively, wave-length. A third object sharing the common the
chromaticity of a color may be specified by a values of luminance and purity but having a
point in a plane diagram. This diagram repre- dominant wave-length greater or less than both
sents the relative chromaticities of all colors in of the original colors will exhibit a hue more
much the same manner as a plane map repre- nearly resembling that of the object from which
sents the relative locations of various places on it differs less with respect to dominant wave-
the earth. An example of the representation of length. The visually estimated ratio of differences
chromaticities by points on the chromaticity in hue will not, in general, equal the ratio of the
diagram is given in Fig. 1. As in the case of a corresponding differences of dominant wave-
plane map of an imperfectly spherical world, length, but an object having a dominant wave-
length intermediate between the dominant
0.62
wave-lengths of two other objects will appear
intermediate in hue also. In the presence of
"O 4 3Oop
0.7!
differences of luminance or purity, or both,
0.71
06
45

5
equality of dominant wave-length does not
044 0
06
6 01~~~olS
61
3560 necessarily assure equality of hue. Similarly, in
0.6, o46 o 3
9114 the presence of differences in luminance or
0.5 030 o71 110 59 70
7
dominant wave-length, equality of purity rarely
0.5 2OO 0111 2 36 0
A 8360.85 094 0108 oll6 S
corresponds to equality of saturation. Finally,
OA
6786 o
o60 .69
68 56 most of the difficulties peculiar to heterochro-
04 4 054 O.0 9160-0

0.335
055
077
930
095
°89
860807
00
27
matic photometry arise from the fact that in
0.3 0490
08
098
~~23
0 096
.73 4 2
the presence of differences in dominant wave-
3090\\ 1120 .1O2 2 99 8 633 49

025 075 odC

0
O 0o 43 1
length or purity, or both, comparisons of
84e 97 o70 34.
93
0.20 64 04101 79 0105 3 brightness are difficult and unsatisfactory.
5 90 .74 o100 080 0106 65 Many investigations of the noticeability of
0.11
480 6o66 88 o 076 4j
color differences 3 5
have been undertaken for
011
IS 470 the purpose of interpreting measured differences
X.I 40 ~
o 005 0.10 05 0.20 O.Z 00 35 .40 OAS5 O50 0.55 60 0.5,C 0.5~
I D. B. Judd, "A Maxwell triangle yielding uniform
FIG. 1. Chromaticities of filters used in chromaticity scales," J. Opt. Soc. Am. 25, 24-35 (1935).
discrimination apparatus. 4 I. G. Priest and F. G. Brickwedde, "Minimum per-
ceptible colorimetric purity as a function of dominant
2 L. A. Jones, "Calorimetry: Preliminary draft of a wave-length," J. Opt. Soc. Am. 28, 133-139 (1938).
report on nomenclature and definitions," J. Opt. Soc. Am. 5 W. D. Wright, "The sensitivity of the eye to small
27, 207-213 (1937). colour differences," Proc. Phys. Soc. 53, 93-112 (1941).
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 249

FIG. 2. Vertical cross section of chromaticity discrimination apparatus.

FIG. 3. Horizontal cross section of chromaticity discrimination apparatus.

in terms comparable with subjective estimates. 1. A two-degree comparison field, divided

In the present report, the results of such an semicircularly.
investigation will be reported in terms of the 2. A surrounding field of twenty-one degrees
standard 1931 I.C.I. specifications of the colors. radius, of uniform color, which may be changed
No modification of this standard system of or extinguished, to provide any reasonable
specifying colors will be recommended. Any conditions of adaptation.
modification of the chromaticity diagram ade- 3. The observer is called upon to manipulate
quately representing these data on notice- only a single control in order to produce the
abilities of color differences would be so com- desired variations of color in each half of the
plicated as to be useless in practice. Such a comparison field.
sacrifice of the simplicity of the standard system 4. Any selected kind of variation of chroma-
is unnecessary, in view of the ease with which ticity from any initial chromaticity can be
the supplementary data concerning noticeability produced in either half of the field, with (a)
can be applied to color specifications expressed automatically constant luminance, if desired, or
in the standard system. (b) the luminance can be made to vary pro-
portionately in any desired ratio with any
A. APPARATUS selected kind of chromaticity variation, or (c) the
The apparatus employed in this investigation luminance alone of any selected chromaticity
was designed to provide the followingfeatures: can be varied.
250 DAVID L. MAcADAM

5. All visual stimuli are produced by additive The light from each point of the extended
mixtures of filtered beams from a single cali- source is collimated by the lens 1. The light pass-
brated light source. The resultant heterogeneous ing through the upper half of this lens is reflected
stimuli are more nearly similar to ordinary visual vertically upwards and then downwards at an
stimuli than are the mixtures of homogeneous angle of ten degrees from the horizontal by two
spectral stimuli frequently used for visual total internal reflections in the prism 2. The
research. The results therefore correspond more other half of the collimated beam is reflected
closely than usual to the most practically im- downwards and then upwards at an angle of
portant circumstances of vision. The stimuli ten degrees from the horizontal by two total
produced in this apparatus are specified ac- internal reflections in prism 3. The light in these
curately by standard colorimetric analyses of two separated, but converging beams is then
the filters employed, and errors which frequently focused by lenses 4 and 5 to produce two coin-
arise from unsuspected impurities of spectral cident images of the light source in the plane of
components are avoided. the diaphragm 6. Light filters can be placed in
6. When the two halves of the comparison the separated beams at 7 and 8. Each of the two
field are matched visually, their energy distribu- beams which intersect within the Wollaston
tions are identical, so that individual differences double-image polarizing prism 9 is split into two
of color-mixture characteristics have a minimum beams, each diverging ten degrees in the vertical
effect on the data for chromaticity sensibility. plane from the direction of the incident beam.
Errors of calibration of the color filters result in Thus, the beam from the upper filter 7 is split
only unimportant second-order errors in the color into two beams, one of which proceeds along the
differences calculated from those calibrations. horizontal axis of the instrument, and the other
These features will be made evident by the of which is deflected downwards at an angle of
description of the construction and functioning twenty degrees from the horizontal, being ulti-
of the apparatus and by the discussion of its mately intercepted and absorbed by the black-
calibration and the method of investigation. ened plate at 10. The portion of the light from
A vertical cross section of the apparatus is the upper incident beam which is directed along
shown in Fig. 2. A horizontal cross section is the axis by the Wollaston prism is polarized in
shown in Fig. 3. The vertical cross section the vertical plane. The beam from the lower
follows the axial rays of one of the two pairs of filter 8 is also split by the Wollaston prism into
light paths, such as the ray indicated by A-B-C-E two beams, one of which proceeds along the
in Fig. 3. The light source, which is situated horizontal axis, together with part of the light
to the right of the instrument as represented in from the upper incident beam. The other of the
the drawings, is not shown on the drawings. two beams into which the lower incident beam
It consists of a ground-glass plate illuminated is split is deflected upwards at an angle of twenty
by the light from a 6-volt projection lamp degrees to the axis of the instrument and is
(Mazda C-8, vertical coil filament, clear bulb, intercepted and absorbed by the blackened
T-10, medium prefocus base). This lamp is plate at 11. The portion of the light from the
operated at the current which has been found by lower incident beam which is directed along the
previous calibration to be necessary to maintain axis by the Wollaston prism is polarized in the
the desired 28480 K color temperature of the horizontal plane. Coincident images of the light
light entering the instrument from the illumi- filters at 7 and 8 are focused by the lens 12 onto
nated ground glass. A ribbon filament lamp the pair of Rochon prisms at 13 and 14.
could be used in place of the coil filament and The diaphragm 15 intercepts the light from
ground-glass combination if much higher field all portions of the upper filter 7 except two areas
illuminance were desired. The ground-glass which are focused by lens 12 onto the Rochon
source has been used exclusively in the experi- prisms 13 and 14. Similarly, the diaphragm 16
ments which are now being reported. All observa- intercepts the light from all portions of the
tions were made with 15 millilamberts field lower filter 8 except two areas which are focused
luminance. by lens 12 onto the Rochon prisms 13 and 14.
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 251

Each of the Rochon prisms is mounted so that them, then the luminous flux transmitted with-
it can be rotated at will within a range of 90 out deviation by each of the Rochon prisms will
degrees. Prism 13 is rigidly connected to the be constant, independently of 0 and conse-
pointer 17 which indicates the azimuth of the quently of the proportions in which the colors
prism on the scale 18 which is graduated in from the two filters 7 and 8 are combined after
degrees. A vernier on the pointer 17 makes transmission by the Rochon prisms. This
possible the recording of tenths of degrees. property of the apparatus is responsible for the
Prism 13 is rotated by the attached toothed feature (4a) of automatically constant field
segment 19 which meshes with the gear 20. luminance for the additive mixtures in all propor-
This gear 20 is activated by the chain drive 21 tions of different colors, such as the colors of U
which is geared to a control knob which may be and V. Similarly, if the luminous transmittances
manipulated by the observer. Prism 14 is of the filters at 7 and 8 are unequal, the variation
rotated by the gears 22 and 23 and chain 24, of field luminance will be proportional to the
coupling to a second control knob. The azimuth variation of chromaticity resulting from the
of prism 14 is indicated by a vernier on the variation of the proportions of the two beams
pointer 25 read against the divided scale 26. transmitted undeviated by the Rochon prisms.
The proportions of the beams from the filters In this manner the experimental condition (4b)
7 and 8 which are transmitted undeviated by can be secured. Finally, of course, the condition
each Rochon prism depend upon the azimuth (4c) can be realized by substituting an opaque
of the prism. The portions of the beams which plate for the filter at 8 thus converting the instru-
are deviated by the Rochon prisms are inter- ment into one type of polarization photometer.
cepted and absorbed by the blackened plate, as The beams transmitted without deviation by
the Rochon prisms are collimated by the lens 28.
indicated at 27. The scale readings are 0 degrees
The biprism 29 refracts these beams and causes
when the Rochon prisms transmit, without
their axial rays to intersect on the dihedral
deviation, light polarized in a vertical plane,
edge of the field biprism 31. The field lens 30
and 90 degrees when the undeviated beams are
focuses coincident images of the light source on
polarized in a horizontal plane. Consequently,
the pupil of the observer's eye E. The dihedral
if U represents the energy distribution of the edge of the biprism 31 is at the focal point of the
light from the upper filter 7 incident on one of the eye lens 32 so that the eye lens of the normal
Rochon prisms, and if V represents the energy
observer is completely relaxed. Other eye lenses
distribution of the light from the lower filter 8 can be substituted for this lens 32 to compensate
incident upon the same prism, then the energy for variations from normal refraction in the eyes
distribution of the light transmitted undeviated of various observers. The boundary of the test
by the prism is field is defined by the end of the cylindrical
W= T(U sin 2
0+ Vcos
2
) () field stop 33. The inner bore of this cylinder is
sufficiently conical so that none of the inner
where 0 is the azimuth of the prism, as indicated surface can be seen by the observer at E. The
by the scale reading, and T is the transmittance circular edge of this conical hole, nearest the
of the Rochon prism for light of such polariza- eye, serves as the field stop. The hidden interior
tion that it is transmitted without deviation. surface of this tube is coated with a dull black
If the values of any additive functions of the finish to absorb most of the light refracted from
distributions U and V are equal, then the value the axial beams by the field biprism. This useless
of the same function of the transmitted distribu- light is kept at a minimum by the diaphragms
tion is constant, independently of all variations 15 and 16 which are of such shape as to intercept
of 0 and consequently of the proportions in which all except those portions of the filtered beams
U and V are combined. In particular, if the which are necessary to illuminate the test field.
filters at 7 and 8 have equal luminous trans- The length of the tube 33 is sufficient to shield the
mittances for the I.C.I. Illuminant A (2848 0K), biprism 31 from all parts of the whitened in-
which is the quality of the light incident upon terior of the sphere 34. The surfaces of the bi-
252 DAVID L. MACADAM

prism 31 and the field lens 30 are coated with a lated for the test field from the azimuth of the
"non-reflecting" layer which further reduces the Rochon prisms and the clorimetric specifica-
possibility of reflected stray light in the test tions of the filters, placed at 7 and 8. The various
field. This treatment of the surfaces of the bi- adjusting and clamping devices which are pro-
prisms 29 and 31 and of the lenses 28 and 30 vided for the focusing and centering of the optical
also reduces to negligible magnitudes the varia- elements are indicated in Figs. 2 and 3. The
tion of reflection losses at these surfaces, result- appropriate procedure for placing such an instru-
ing from the oblique incidence of the polarized ment in adjustment can be devised by any
beams from the rotatable prisms 13 and 14. The person familiar with the maintenance of pre-
interiors of the eye tube 35 and of the lamp shield cision optical instruments. An artificial pupil
36 are finished dull black, to minimize the reflec- may be placed over the eye lens at 41. Such an
tion of light from the convex surface of the field artificial pupil, with a diameter of 2.6 milli-
lens 30. The biprism 31 is cemented to the plane meters, was used to fix the level of retinal
surface of the field lens 30 to eliminate the reflec- illumination in the experiments described in
tion of light from these plane surfaces. Similarly, this paper.
the biprism 29 is cemented to the plane surface This instrument is designed for monocular
of the lens 28. The external surfaces of the field observation but it could conceivably be con-
stop 33, the entire inner surface of the sphere verted into a binocular instrument. Equal
34, and the outer surface 37 of the lamp shield transmissions for both eyes, and constant
are painted white. An opaque layer of smoke from transmission for all planes of polarization are
burning magnesium ribbon was deposited simul- essential requirements of any binocular viewing
taneously on the cylindrical field stop 33 and device employed with this instrument. The
on the hemisphere visible from E, in order to Maxwellian field, and the large illuminated
obliterate completely all joints and irregularities. surrounding field impose further requirements
The illumination of the interior of the sphere is which eliminate practically all known binocular
so uniform and diffused that the outlines of the eyepieces. Color-mixture data, luminosity data,
cylindrical field stop 33 can be distinguished and sensibilities to color differences are occa-
only with effort. The radius of the illuminated sionally, although not commonly, distinctly
surrounding field is limited by the length of the different for the right and left eye of a single
eye tube 35. The limiting rays are shown by 38. observer. Such differences are not confined to
The angular extent of the illuminated surround- observers who exhibit definitely anomalous
ing field could be increased somewhat by shorten- color discrimination, but have also been detected
ing the eye tube 35 or by increasing its diameter, in the case of observers classed as "normal"
but the advantage would appear to be negligible by all of the usual tests. Such differences would
in comparison with the increased difficulties of result in erratic results of binocular observations.
construction, and discomfort of the observer, The results of binocular observations would not
who would be required to place his head very be expected to average consistently the different
close to the sphere lamps and their connections. characteristics of the two eyes. Rather, one eye
An annular ring 39 of filter film can be placed would be expected to dominate and govern the
between the lamp shield and the sphere, to results of the majority of the observations, with
introduce whatever chromaticity may be desired the subordinate eye gaining dominance oc-
in the surrounding field. The intensity of the casionally and contributing erratic observations,
surrounding field can be changed by altering the as the result of either conscious or unconscious
number or size of the lamps 40 employed, or by binocular rivalry.
varying the current passing through them. The A movable eye shield which permits the use of
chromaticity and the luminance of the surround- either eye has been employed throughout the
ing field are calibrated by establishing a color monocular observations which are the subject
match between the surrounding field and the of this report. This shield protects the idle eye
test field. The calorimetric specifications of the from all light, permitting the idle eye to remain
surrounding field are then equal to those calcu- open and relaxed.
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 253

TABLE I. Filters for chromaticity instrument (in Illuminant A).

Filter Constituents Filter Constituents

number Chromatic Neutral X Y Z m x y number Chromatic Neutral X Y Z m x y

34,38 0.07 0.01640 0.00280 0.08855 0.10775 0.152 0.026 69 78A 2.13 0.00257 0.00262 0.00149 0.00668 0.385 0.392
65,8 1.25 0.00059 0.00265 0.00066 0.00390 0.150 0.680 70 78,30 0.85 0.00478 0.00290 0.00457 0.01225 0.390 0.237
75 0.42 0.00071 0.00280 0.00575 0.00911 0.077 0.302 71 69 1.87 0.00110 0.00285 0.00123 0.00518 0.212 0.550
xi 0.24 0.00044 0.00276 0.00023 0.00343 0.128 0.805 72 66 2.37 0.00217 0.00285 0.00070 0.00572 0.380 0.498
3(63) 0.00027 0.00280 0.00076 0.00383 0.070 0.736 73 78AA,30 1.34 0.00460 0.00290 0.00217 0.00967 0.475 0.300
3(74) 0.10 0.00062 0.00270 0.00011 0.00353 0.180 0.785 74 78,78AA 0.65 0.00265 0.00259 0.01052 0.01576 0.168 0.164
2(73) 0.50 0.00256 0.00272 0.00000 0.00528 0.485 0.514 75 78,78B 1.15 0.00251 0.00275 0.00574 0.01100 0.228 0.250
70 0.50 0.00769 0.00280 0.00000 0.01049 0.737 0.263 76 78,78AA,30 0.25 0.00522 0.00260 0.01280 0.02032 0.253 0.126
78 1.17 0.00278 0.00294 0.00339 0.00911 0.305 0.323 77 44A 1.62 0.00110 0.00280 0.00354 0.00729 0.151 0.364
X2 0.10 0.00035 0.00297 0.00043 0.00375 0.092 0.790 78 44 1.70 0.00111 0.00250 0.00261 0.00652 0.170 0.430
35 0.51 0.00943 0.00295 0.00980 0.02218 0.425 0.133 79 78,78B,30 0.60 0.00503 0.00290 0.00656 0.01449 0.347 0.200
X3 0.65 0.01153 0.00311 0.02400 0.03864 0.299 0.086 80 78,78A,30 0.45 0.00570 0.00310 0.01020 0.01900 0.300 0.163
72 0.78 0.00549 0.00293 0.00000 0.00842 0.651 0.348 81 78AA,78C,30 1.23 0.00421 0.00265 0.00260 0.00946 0.445 0.280
4(65A) 0.00029 0.00286 0.00268 0.00583 0.050 0.490 82 78,78AA,X4 0.37 0.00550 0.00288 0.01780 0.02618 0.210 0.110
45A 0.55 0.00188 0.00298 0.01247 0.01733 0.108 0.172 83 78,78AA,X5 0.45 0.00438 0.00272 0.01633 0.02343 0.187 0.116
3(65) 0.08 0.00029 0.00278 0.00199 0.00506 0.057 0.550 84 78,78AA,X6 0.60 0.00307 0.00280 0.00630 0.01217 0.252 0.230
21,53 1.21 0.00216 0.00268 0.00000 0.00484 0.446 0.553 85 67 1.70 0.00149 0.00308 0.00231 0.00688 0.217 0.448
16,53 1.45 0.00181 0.00283 0.00002 0.00466 0.388 0.608 86 67,78C 1.60 0.00132 0.00300 0.00229 0.00661 0.200 0.454
2(16),2(53) 0.48 0.00151 0.00260 0.00002 0.00413 0.365 0.630 87 30 2.20 0.00445 0.00295 0.00104 0.00844 0.527 0.350
31,86C 1.84 0.00583 0.00270 0.00102 0.00955 0.610 0.283 88 X4 2.30 0.00416 0.00290 0.00105 0.00811 0.513 0.358
2(61),15 1.00 0.00097 0.00280 0.00009 0.00386 0.251 0.725 89 X7 2.44 0.00387 0.00307 0.00115 0.00809 0.478 0.380
2(61),12 1.12 0.00075 0.00280 0.00011 0.00366 0.205 0.764 90 X8 2.44 0.00330 0.00280 0.00090 0.00700 0.471 0.400
35,86C 0.68 0.00919 0.00305 0.00611 0.01835 0.500 0.166 91 X6 2.48 0.00327 0.00285 0.00099 0.00711 0.460 0.401
35,86B 0.66 0.00879 0.00302 0.00427 0.01608 0.546 0.188 92 X9,34 0.70 0.01055 0.00300 0.01803 0.03158 0.334 0.095
35,86A 0.70 0.00784 0.00294 0.00252 0.01330 0.590 0.221 93 38,78B 1.85 0.00194 0.00265 0.00229 0.00688 0.282 0.385
34,78A 0.30 0.01293 0.00275 0.04780 0.06348 0.203 0.045 94 51,38A 1.77 0.00175 0.00306 0.00166 0.00647 0.270 0.473
34,78B 0.45 0.01075 0.00285 0.03180 0.04540 0.237 0.063 95 X7,78A 1.77 0.00361 0.00300 0.00188 0.00849 0.425 0.354
35,78B 0.40 0.00987 0.00297 0.01304 0.02588 0.382 0.115 96 X5,78A 1.74 0.00378 0.00300 0.00223 0.00901 0.420 0.333
35,86A 0.60 0.00756 0.00290 0.00125 0.01171 0.646 0.248 97 X7,78 1.74 0.00387 0.00312 0.00687 0.01386 0.279 0.225
63,2(65) 0.00028 0.00284 0.00132 0.00444 0.057 0.656 98 38A,78B 1.50 0.00190 0.00280 0.00405 0.00875 0.217 0.320
61,12 1.67 0.00105 0.00280 0.00011 0.00396 0.266 0.706 99 X7,78AA 1.45 0.00380 0.00315 0.00413 0.01108 0.343 0.281
3(65),2(4) 0.00025 0.00278 0.00161 0.00464 0.054 0.603 100 X7,78,78A 0.84 0.00385 0.00281 0.00957 0.01623 0.237 0.173
49A 0.47 0.00603 0.00301 0.03600 0.04504 0.133 0.067 101 38A,78 0.80 0.00224 0.00280 0.00895 0.01399 0.160 0.200
25 2.00 0.00605 0.00280 0.00000 0.00885 0.684 0.316 102 78A,38A 1.20 0.00195 0.00285 0.00564 0.01044 0.187 0.273
29 1.71 0.00695 0.00280 0.00000 0.00975 0.712 0.288 103 31 1.87 0.00627 0.00295 0.00171 0.01093 0.574 0.270
46 0.95 0.00225 0.00283 0.01405 0.01913 0.118 0.148 104 31,78B 1.50 0.00669 0.00310 0.00330 0.01309 0.511 0.237
15 2.60 0.00355 0.00290 0.00001 0.00646 0.550 0.449 105 31,78A 1.16 0.00622 0.00280 0.00505 0.01407 0.442 0.199
21 2.45 0.00382 0.00278 0.00000 0.00660 0.580 0.420 106 31,78AA 0.73 0.00694 0.00290 0.00912 0.01896 0.366 0.153
22 2.35 0.00453 0.00280 0.00000 0.00733 0.617 0.382 107 X5,2(78) 0.10 0.00763 0.00270 0.03700 0.04733 0.161 0.057
75,65A,2(4) 0.00038 0.00290 0.00396 0.00724 0.053 0.400 108 66,78A 1.70 0.00201 0.00290 0.00126 0.00617 0.326 0 470
75,65A,4 0.00040 0.00275 0.00408 0.00723 0.055 0.380 109 59A 1.28 0.00242 0.00310 0.00038 0.00590 0.410 0.525
61,2(15) 1.46 0.00136 0.00280 0.00005 0.00421 0.322 0.666 110 59A,38 0.95 0.00182 0.00315 0.00060 0.00557 0.327 0.565
2(16) 2.50 0.00378 0.00285 0.00001 0.00664 0.570 0.429 111 59A,78 1.13 0.00175 0.00315 0.00138 0.00628 0.279 0.502
21,53 0.90 0.00209 0.00280 0.00001 0.00490 0.427 0.571 112 3(38A) 1.13 0.00141 0.00265 0.00542 0.00948 0.149 0.279
38,78B 1.92 0.00206 0.00270 0.00195 0.00671 0.307 OA02 113 65 1.28 0.00065 0.00260 0.00174 0.00499 0.131 0.521
59 2.15 0.00149 0.00280 0.00040 0.00470 0.318 0.595 114 65,86 0.88 0.00069 0.00270 0.00125 0.00464 0.148 0.583
53,22 0.93 0.00287 0.00280 0.00000 0.00567 0.506 0.494 115 60 1.84 0.00103 0.00280 0.00063 0.00446 0.230 0.629
34,78AA 0.01450 0.00275 0.06550 0.08275 0.175 0.033 116 2(51),78A 1.75 0.00213 0.00255 0.00092 0.00560 0.380 0.455
75,45 0.56 0.00177 0.00310 0.01296 0.01783 0.102 0.195 117 48A 0.32 0.00944 0.00280 0.05312 0.06536 0.144 0.042
35,2 0.49 0.00850 0.00270 0.00783 0.01903 0.446 0.1415 118 48A,2(2A) 0.24 0.00766 0.00285 0.04235 0.05276 0.145 0.054
2(46) 0.36 0.00320 0.00285 0.02083 0.02688 0.119 0.106 119 34,30 0.72 0.00928 0.00290 0.00976 0.02194 0.423 0.132
2(21),53 1.00 0.00230 0.00269 0.00000 0.00499 0.460 0.539 120 X7,78AA,78B 1.20 0.00369 0.00300 0.00561 0.01230 0.300 0.244
2.60 0.00296 0.00274 0.00078 0.00648 0.457 0.423

B. CALIBRATED COLOR FILTERS of these Laboratories, whose interest and assist-

ance is here gratefully acknowledged.
A set of about 100 color filters has been pre-
Table I indicates the constituents and colori-
pared for use with this instrument. The luminous
transmittances of these filters are all within 10 metric specifications of the assembled filters.
percent of their average, 0.00285. The luminous The chromaticities of these filters are represented
transmittances of these filters, and their chro- in Fig. 1. The serial numbers with which these
maticities, have been calculated from their filters are designated in Fig. 1 and column 1 of
spectral transmittances and for the standard Table I bear no relation to the Wratten filter
I.C.I. observer and Illuminant A. The filters numbers of the components. The serial numbers
consist of one or more pieces of gelatin filter correspond approximately to the order in which
film cemented between microscope slide glasses. the filters were assembled. The Wratten filter
Most of the filter components were selected from numbers of the constituents of each filter are
the commercial Wratten filters. A few special listed in the second column. Special filter com-
filter films were required in order to secure ponents are numbered serially, with the letter X
chromaticities unobtainable with combinations preceding each of these numbers to avoid con-
of the standard filters. These special filter films fusion with the Wratten filter numbers. In
were designed and made by Mr. Albert Smith, many cases, two or more pieces of one type of
254 DAVID L. MACADAM

filter film were used. Such cases are indicated by spectrophotometric data could not be secured by
parentheses around the Wratten filter number direct measurements of the finished filters. The
preceded by a numeral indicating the number of spectrophotometric data, such as those repre-
pieces used. In the fourth column is given the sented by the curves shown in Figs. 4 and 5,
density of the Wratten neutral filter film em- have been determined in various ways, but
mostly by spectrophotometric measurement of
1.4 the separate components of each filter, followed
1.3 35 by computation of the transmittance of the
cemented combinations. The effects of cementing
on total absorption and in the reduction of
reflection losses were carefully investigated and
460
41.0 500 505733 taken into account in these computations. As a
final check, the relative luminous transmittances
0.9
of filters of similar chromaticities were com-
pared visually. The instrument represented in
0.7- Figs. 2 and 3 was used for these comparisons.
This apparatus functioned as a polarization
photometer, when an opaque plate was placed
z0.5 at position 8. The two filters being compared were
placed at 7, each filter covering one of the open-
0L0.4 ings in the diaphragm 15. Filters represented by
WAVE LENG (ov LLM45QN neighboring points in Fig. 1 have sufficiently
0.2 4 similar chromaticities so that these hetero-
chromatic comparisons of luminous transmit-
0.1 3 5
tances could be made quite satisfactorily. The
A460 500 540 580 G2o G6o relative luminous transmittances of filters com-
WAIVEF-
LENGTH (MILLIMICRONS) pared in this manner are inversely proportional
to the ratio of the squares of the sines of the
FIG. 4. Spectrophotometric curves of five filters.
azimuths of the corresponding Rochon prisms.
This use of the sine square law is very favorable
ployed to adjust the various filters to the de- for the comparison of nearly equal transmit-
sired value of luminous transmittance. The I.C.I. tances, such as are involved in the comparisons
tristimulus values, X, Y, and Z are given in the of the set of filters employed here. Systematic
fifth, sixth, and seventh columns. The value of Y variations from the computed values of the
is the luminous transmittance of each filter. The luminous transmittances were observed and
sum m of the three tristimulus values is given correlated with the positions of the points repre-
in the eighth column. This value is the weight senting the filters in the chromaticity diagram.
which may be assigned to the color transmitted These systematic variations were confirmed by
by each filter in computations of the chromatici- repetitions of the comparisons by the same ob-
ties of mixtures according to the center of server on different days, but were different for
gravity principle.6' 7 The coordinates, x, y, of the different observers. Although these variations
points representing each filter in the chromaticity and correlations require further study in relation
diagram, Fig. 1, are given in columns 9 and 10 to the actual luminosity curves for the observers
of Table I. involved, before any positive conclusions can be
These filters have such low luminous trans- stated, these variations are tentatively attri-
mittances that in most cases sufficientlyaccurate buted to variations of the luminosity curves of
I D. L. MacAdam, "The theory of the maximum visual the observers from the standard I.C.I. data used
efficiency of colored materials," J. Opt. Soc. Am. 25, in the computations. In particular, all observers
249-252 (1935). seemed to have higher than standard luminosity
7D. L. MacAdam, "Projective transformations of I.C.I.
color specifications," J. Opt. Soc. Am. 27, 294-299 (1937). values for the extreme red end of the spectrum,
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 255

C. COLORIMETRIC CALIBRATION OF APPARATUS

and somewhat lower than standard luminosity
values for the extreme violet end of the spectrum. Colorimetric specifications of the mixtures
Such an effect might be expected and attributed obtained in the test field of the instrument
to the differences in the conditions of observa- represented in Figs. 2 and 3 can be computed
tion from those employed by Gibson and Tyn- by use of the tristimulus values of the filters in
9 In particular, the large illuminated
dall. equations similar to Eq. (1). Let the symbols
surrounding field, and the maintenance of a X., Y., Zu,represent the tristimulus values of the
fairly high constant retinal illuminance (200 field when 0= 90 degrees, that is, when the color
photons, or 8 lux) for all comparisons, including of the field is produced by the filter at position 7,
the extreme red, violet, and purples, may have alone. Let the symbols X, Y, Z, represent the
contributed appreciably to this effect. Although tristimulus values of the field when 0 = 0 degrees,
the high, central portions of the standard curve, that is, when the color of the field is produced by
which are usually most important in practice, the filter at position 8, alone. Then, for any
were determined at normal levels of retinal scale reading 0 the tristimulus values X, Y, Z
illuminance, 5 the relative luminosities of wave- of the mixture are given by the equations:
lengths in the extreme red and violet regions of
X=X, sin2 0+X, cos2 6, (2)
the spectrum were determined with considerably
lower retinal illuminances. Consequently, near Y= Yu sin2 0+ Y, cos2 0, (3)
the ends of the visible spectrum the rates of
change of the standard curve with wave-length Z =Z sin2 0+Z, cos2 0. (4)
may be more nearly representative of the scotopic
curve than of the photopic curve. Such an effect
would not cause appreciable errors in most
practical problems, but would be expected to
give rise to noticeable discrepancies for filters
occur largely W
whose luminous transmittances C) 1.1
at either or both of the extremes of the visible z
4 1.0
spectrum. No attempt was made to compensate
F- 0.9
for the systematic discrepancies between the
observed and computed luminous transmit- Uj 08
tances. Very few discrepancies, other than the 07
systematic variations, were detected between
The few 1- 0.6
the observed and computed values.
isolated errors were traced to spectrophotometric
or computational errors. Such errors were cor- LU0.4
rected and eliminated before Table I and Figs. 1,
4, and 5 were completed.
All of the filters were stored so as to minimize (0.2
changes in any of the components which may be 0.1
unstable. The use of unstable components was
avoided as far as possible, and those filters which 50 540 580 620 660 7(
contained constituents suspected of instability WAVE- LENGTX- (MILLIMICRONS)
were checked periodically, both spectrophoto- FIG. 5. Spectrophotometric curves of six filters.
metrically and visually.
Since the value of Y is the luminance of the field,
I K. S. Gibson and E. P. T. Tyndall, "Visibility of the luminance of any mixture is given by Eq. (3).
radiant energy," Sci. Papers Nat. Bur. Standards 19,
131-191 (1924); Trans. Illum. Eng. Soc. 19, 176-196 If the luminous transmittances of the filters at 7
(1924). and 8 are equal for the quality of light which is
'Proceedings of the Sixth Session, International Com-
mission on Illumination (Geneva, 1924), pp. 67 and 232. incident upon them (Illuminant A), then the
256 DAVID L. MACADAM

luminance Y of the field is constant for all values the curves shown in Fig. 6. A special slide rule
of 0. was designed and used for all of the computa-
The coordinates of the point representing the tions off involved in this investigation. This slide
mixture in the chromaticity diagram can be rule, illustrated in Fig. 7, avoids the difficulties
calculated from the values X, Y, Z given by and errors of interpolation encountered in the
Eqs. (2), (3), and (4). The center of gravity use of Fig. 6. The bottom scale A on the frame
principle provides a much more convenient of the rule represents values of the ratio r from
method of locating the point. Let the symbol 0.01 to 100.0. The distances in inches to the
mu represent the "mass" X.+ Y.+Z., assigned right from the center index to the divisions are
in Table I to the filter at position 7, which pro- given by the equation:
duces the color of the field when 0=90 degrees.
D = 5 logio r. (6)
Let the symbol m, represent the "mass" Xv+ Y,
+Z, assigned in Table I to the filter at position 8. The lower scale B on the slide represents recip-
Then the point representing the mixture for the rocals of the ratio r from 1/r = 0.004 to 1/r =40.0.
scale reading divides the straight line joining Scale B is divided in exactly the same manner
the points representing the filters in the ratio as is scale A, except that the index is displaced
mu sin2 to m. cos2 . Let the chromaticity of the two inches to the right of the center of the slide.
field when = 90 degrees be represented by the The upper scale T on the slide represents values
point U. Let the chromaticity of the field when of the angle from 9 degrees to 86.5 degrees.
0 = 0 degrees be represented by the point V.
The distances in inches to the right from the
Let the chromaticity of the field for the scale index to the divisions of this scale a.e given by
reading be represented by the point W. If the the equation:
ratio of the distances VW/ VU is represented by D = 10 loglo cot 0. (7)
the fraction f then
The 45-degree index is at exactly the same
f= 1/(r cot2
6+1), position on the slide as the unit index on scale B,
where r = mV/mU. that is, two inches to the right of the center of the
slide. Finally, the top scale F on the frame, repre-
The fractional distance of W from V toward U sents values of the fraction f from 0.01 to 0.99.
may be estimated for any values of r and from The distances in inches from the center index
to the divisions of this scale are given by the
equation:
D=5 logi0 (1-f)/f. (8)

In all of the scales, positive values of D are

measured to the right of the index and negative
values to the left. When the -45-degree index of
the slide is aligned with the value of r on the A
scale, the value of f can be found on the F scale
opposite the value of on the T scale. Alter-
natively, the value of the reciprocal of the ratio
1/r on the B scale can be aligned with the center
index of the frame. Since Eq. (5) reduces to
f=sin 2 6, when r= 1.0, the values of sin2 6
required in Eq. (3) for the calculation of lumi-
nance can be determined by setting the 45-degree
index of the slide at r= 1.0 and reading the values
of sin2 0 from the F scale. The values of cos2 0
o= 10 20' 30' 40_. 50' 60' 70 80' which are also required in Eq. (3) are easily
determined by subtracting the values of sin2 0
FIG. 6. Calibration curves of chromaticity
discrimination apparatus. from 1.0.
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 257

.99 .95 ~~~ ~~~ ~~.8

.6
.7 .5 * . .2. .0F

FIG. 7. Slide
.70 A0
..
rul
8 0
fo t clcuaton f cr
6 0 .... ynteszedindisrimnaionaparaus
.o0. W 5 2 0 70 60 A,

FIG. 7. Slide rule for calculation of chromaticities synthesized in discrimination apparatus.

The distance VW of the point representing the ences. The just noticeable difference was used
mixture from the point representing the chro- extensively with results consistent with those
maticity of the field when 0 = 0 degrees is equal which are given in this paper, but intercompari-
to the value f multiplied by the distance UV sons of data for different series of colors were
between the points representing the chromatici- unsatisfactory because of rather great fluctua-
ties of the field when 0=0 degrees and 0= 90 tions to which this criterion appeared to be
degrees. When the distance UV is measured in subject. A consistent system of data, including
the same units as the coordinate scales of the representative kinds of color difference through-
chromaticity diagram, it is represented by the out the entire gamut of colors attainable with
symbol S. the instrument, could not be determined satis-
S = (X - X)2 + (y. - yv)2]12 (9) factorily with a criterion subject to such fluctua-
tions. Adjustments of the data to compensate
where x. = Xu/mu, Yu= Y//mu, for such fluctuations were considered inad-
Xv= X/MV Yv= Y 9/m,. missible, since they might be influenced by
The distance VW measured in the same units preconceptions concerning the probable trend of
as the coordinate scales is represented by the the data and accumulated errors of such adjust-
symbol s. The distances S and s may be meas- ments might introduce spurious characteristics
ured with reasonable accuracy by use of a pair and obscure essential features of the results.
of dividers with a chromaticity diagram drawn Attempts were made to adjust color differences
accurately to a fairly large scale. The multi- to equal certain standard color differences, sev-
plication, eral times greater than just noticeable. Observa-
s=fS, (10) tions of the standard color differences were
alternated with the adjustments of the test
can be carried out with comparable accuracy differences. This criterion provided more con-
by using in the customary manner the A and B sistent results than the criterion of just notice-
scales of the slide rule represented in Fig. 7. ability, but was exceedingly tedious and re-
Differences of the chromaticities correspond-
quired extensive training of the observer. Such
ing to small changes, AO,of the position of either training cannot be given to subjects who are
Rochon prism 13 or 14 in Fig. 3, can be de-
available for only a short examination during
termined with great accuracy by calculating the which an adequate determination of color dis-
product crimination must be completed. These examina-
As =AO(ds/do). (11)
tions are one of the most important immediate
A sufficiently accurate value of the derivative applications of the instrument employed in the
can be found by subtracting the values of f present investigation, and a criterion was sought
indicated by the slide rule for neighboring values which would be applicable in both routine testing
of 0, dividing this difference by the difference and research. Such a criterion must be more reli-
of the values of 0, and multiplying this ratio by able than the just noticeable differenceand require
the value of S. Small differences of chromaticity less training and experiencethan either that or the
computed from Eq. (11) are considerably more judgment of equally noticeable color differences.
accurate than the chromaticities of the filters Kbnig and Dietericill employed the mean error
used to synthesize the colors.
10A. Konig and C. Dieterici, "Ober die Empfindlichkeit
des normalen Auges fr Wellenlangenunterschiede des
D. STANDARD DEVIATIONS OF COLOR MATCHING Lichtes," Wied. Ann. d. Physik und Chemie 22, 579-589
(1884); Graefes Archiv. 30 (2), 158 (1884); Gesammelte
Several criteria were investigated for the Abhandlungen zur Physiologischen Oplik (Leipzig, 1903),
measurement of the noticeability of color differ- p. 23.
258 DAVID L. MIAcADAM

TABLE II. Standard deviations of color matching.

00 900 00 AO, Dis- 00 90, 0 AO, Dis-

Fit- Fil- Aver- Stand- tance Fil- Fit- Aver- Stand- tance
ter ter age ard Average from 00 A~s ter ter age ard Average from00 A~s
num- num- set- devia- chromaticity point Standard num- num- set- devia- chromaticity point Standard
ber ber ting tion x 3P =s deviation AY/Ax ber ber ting tion x y =s deviation AY/Ax
4 23 83.8 0.239 0.286 0.286 0.294 0.0027 1 1.94 79 23 28.9 0.6 14 0.340 0.220 0.02 1 0.00087 -2.92
4 23 79.1 0.256 0.259 0.233 0.232 0.00370 1.94 79 23 44.0 0.408 0.332 0.245 0.048 0.00084 -2.92
4 23 79.1 0.230 0.259 0.233 0.232 0.00332 1.94 79 23 73.9 0.603 0.3 10 0.309 0.115 0.00 106 -2.92
4 23 74.1 0.247 0.230 0.177 0.171 0.00290 1.94
4 23 69.0 0.382 0.208 0.135 0.122 0.00312 1.94 41 23 30.2 0.241 0.080 0.017
0.477 0.00059 3.82
4 23 68.5 0.304 0.206 0.13 1 0.117 0.00242 1.94 41 23 0.427 0.248 0.106 0.045
44.9 0.00113 3.82
4 23 58.4 0.335 0.180 0.081 0.062 0.00150 1.94 41 23 45.1 0.249 0.107 0.046
0.421 0.00111 3.82
4 23 42.1 0.477 0.162 0.045 0.022 0.00067 1.94 41 23 0.296 0.263 0.163 0.103
60.3 0.00151 3.82
4 23 25.1 0.563 0.155 0.031 0.006 0.00028 1.94
76 23 29.6 0.706 0.260 0.142 0.026 0.0012 7 3.80
100 23 43.7 0.823 0.259 0.254 0.056 0.002 13 2.21 76 23 33.5 0.729 0.262 0.158 0.033 0.00155 3.80
76 23 44.4 0.637 0.268 0.163 0.061 0.00178 3.80
66 23 73.9 0.284 0.268 0.280 0.229 0.00163 1.16 76 23 60.8 0.563 0.288 0.24 1 0.121 0.0023 1 3.80
66 23 60.9 0.314 0.2 16 0.2 19 0.139 0.00184 1.16 76 23 74.9 0.555 0.298 0.295 0.176 0.00176 3.80
66 23 52.0 0.334 0.185 0.183 0.102 0.00156 1.16
66 23 47.2 0.430 0.169 0.165 0.080 0.00176 1.16 40 23 30.1 0.662 0.208 0.058 0.014 0.00069 2.73
66 23 45.1 0.423 0.166 0.161 0.072 0.00154 1.16 40 23 45.1 0.531 0.216 0.080 0.038 0.00122 2.73
66 23 32.8 0.507 0.142 0.133 0.035 0.001 16 1.16 40 23 60.0 0.388 0.234 0.128 0.090 0.00193 2.73
66 23 16.7 0.642 0.125 0.113 0.008 0.00064 1.16 40 23 74.7 0.293 0.270 0.228 0.195 0.00274 2.73
64 23 73.0 0.400 0.274 0.303 0.203 0.00157 0.63 63 23 15.2 1.503 0.176 0.035 0.003 0.00058 2.23
64 23 58.7 0.302 0.220 0.269 0.139 0.00139 0.63 63 23 28.7 1.090 0.179 0.043 0.010 0.00078 2.23
64 23 44.3 0.372 0.169 0.237 0.078 0.00139 0.63 63 23 33.6 1.345 0.181 0.046 0.013 0.0012 7 2.23
64 23 30.2 0.652 0.132 0.214 0.035 0.00156 0.63 63 23 40.0 1.010 0.184 0.054 0.02 1 0.00133 2.23
64 23 15.9 0.988 0.110 0.200 0.009 0.00122 0.63 63 23 41.6 1.082 0.185 0.056 0.026 0.00173 2.23
63 23 45.2 0.810 0.188 0.062 0.029 0.00153 2.23
8 23 74.1 0.531 0.288 0.322 0.213 0.00113 0.09 63 23 50.0 0.920 0.193 0.082 0.039 0.00226 2.23
8 23 59.6 0.376 0.247 0.318 0.172 0.00147 0.09 63 23 59.4 0.759 0.206 0.103 0.076 0.00384 2.23
8 23 45.2 0.504 0.192 0.313 0.116 0.00206 0.09 63 23 59.6 0.800 0.207 0.103 0.078 0.00405 2.23
8 23 30.9 0.591 0.137 0.308 0.061 0.002 15 0.09 63 23 69.3 0.443 0.232 0.160 0.140 0.00370 2.23
63 23 70.1 0.443 0.234 0.166 0.148 0.00370 2.23
23 55 21.7 0.469 0.27 7 0.329 0.029 0.00119 -0.23 63 23 74.7 0.354 0.252 0.205 0.190 0.00373 2.23
23 55 30.7 0.322 0.251 0.335 0.056 0.00112 -0'23 63 23 79.0 0.296 0.272 0.250 0.238 0.00350 2.23
23 55 31.1 0.326 0.249 0.336 0.057 0.00117 -0.23
23 55 45.3 0.296 0.193 0.349 0.116 0.00131 -0.23 23 24 21.9 0.422 0.292 0.352 0.032 0.00130 -2.19
23 55 46.0 0.430 0.19 1 0.349 0.118 0.00170 -0.23 23 24 31.2 0.458 0.277 0.384 0.068 0.002 14 -2.19
23 55 46.2 0.436 0.190 0.349 0.119 0.00172 -0.23 23 24 44.3 0.508 0.245 0.454 0.145 0.00378 -2.19
23 55 60.6 0.614 0.126 0.364 0.185 0.00265 -0.23 23 24 59.4 0.55 1 0.190 0.576 0.280 0.00575 -2.19
23 55 61.2 0.587 0.124 0.364 0.187 0.00245 -0.23 23 24 73.1 0.880 0.131 0.704 0.422 0.00857 -2.19
23 55 75.6 1.080 0.074 0.376 0.238 0.00278 -0.23
23 55 77.4 0.953 0.070 0.377 0.243 0.00211 -0.23 23 36 30.0 0.539 0.293 0.375 0.054 0.00206 -4.41
23 36 44.5 0.470 0.27 7 0.445 0.126 0.00307 -4.41
23 28 21.4 0.432 0.282 0.338 0.027 0.00111 -0.66 23 36 58.2 0.526 0.254 0.549 0.233 0.00470 -4.41
23 28 30.7 0.430 0.258 0.354 0.057 0.00142 -0.66 23 36 72.3 0.794 0.225 0.675 0.363 0.00697
23 28 45.2 0.332 0.205 0.389 0.120 0.00173 -0.66 -4.41
23 36 72.9 0.777 0.224 0.680 0.369 0.00679 -4.41
23 28 45.7 0.391 0.203 0.390 0.123 0.00203 -0.66
23 28 59.6 0.490 0.139 0.432 0.200 0.00275 -0.66 23 33 28.6 1.043 0.3 12 0.360 0.038 0.00284
23 28 74.0 0.778 0.079 0.471 0.271 0.00310 -0.66 5.03
23 33 44.2 0.777 0.323 0.415 0.094 0.00350 5.03
23 46 20.2 0.370 0.289 0.341 0.024 0.00094 -1.12
23 33 59.1 0.812 0.338 0.494 0.176 0.00496 5.03
23 33 74.4 1.385 0.356 0.585 0.268 0.00729 5.03
23 46 20.8 0.515 0.288 0.342 0.026 0.00125 -1.12
23 46 30.1 0.425 0.269 0.364 0.055 0.00162 -1.12 23 59 15.2 0.310 0.332 0.011 0.00219
1.585
23 46 30.4 0.298 0.278 0.365 0.056 0.00113 -1.12 2.04
23 59 18.7 0.312 0.338 0.016 0.00293
1.438 2.04
23 46 30.8 0.327 0.267 0.366 0.058 0.00123 -1.12 23 59 25.8 0.319 0.35 1 0.031 0.00246
1.000
23 46 44.6 0.461 0.222 0.416 0.126 0.00232 -1.12 2.04
23 59 41.6 0.341 0.397 0.084 0.00286
0.689 2.04
23 46 45.8 0.441 0.2 18 0.420 0.132 0.00259 -1.12 23 59 42.5 0.341 0.400 0.088 0.00286
0.681
23 46 45.5 0.410 0.219 0.419 0.130 0.00241 -1.12 2.04
23 46 58.4 0.423 0.162 0.483 0.218 0.00308 -1.12 23 59 58.0 0.774 0.376 0.466 0.162 0.00418 2.04
0.437 0.15 7 0.489 0.225 0.00320 23 59 69.6 1.037 0.402 0.520 0.223 0.00563 2.04
23 46 59.4 -1.12
23 46 59.9 0.536 0.153 0.491 0.228 0.00393 -1.12 23 17 22.8 0.866 0.322 0.341 0.024 0.00185
23 46 60.9 0.510 0.149 0.497 0.235 0.00374 -1.12 1.06
23 17 44.7 0.510 0.370 0.392 0.096 0.002 15 1.06
23 46 59.8 0.428 0.154 0.491 0.228 0.00311 -1.12 23 17 59.9 0.608 0.419 0.444 0.168 0.00299
23 46 70.8 0.690 0.102 0.549 0.305 0.00498 -1.12 1.06
0.928 0.088 0.564 23 17 75.9 1.028 0.468 0.496 0.239 0.00368 1.06
23 46 74.1 0.327 0.00527 -1.12
23 46 77.5 1.027 0.076 0.578 0.346 0.00507 -1.12 23 51 17.4 0.557 0.32 1 0.331
0.018 0.00116 0.51
38 23 31.8 0.572 0.503 0.212 0.050 0.00182 -0.56 23 51 29.1 0.360 0.349 0.346
0.051 0.00 122 0.51
38 23 46.0 0.376 0.456 0.248 0.104 0.00171 -0.56 23 51 43.8 0.397 0.402 0.373
0.110 0.00190 0.5 1
38 23 60.6 0.267 0.391 0.275 0.179 0.00141 -0.56 23 51 43.8 0.378 0.402 0.373
0.110 0.00181 0.51
38 23 75.8 0.232 0.329 0.309 0.250 0.00090 -0.56 23 51 59.1 0.437 0.468 0.407 0.186 0.002 19 0.51
23 51 72.4 0.835 0.520 0.434 0.244 0.00302 0.51
65 23 30.7 0.392 0.426 0.168 0.033 0.00089 -1.29
65 23 31.1 0.408 0.426 0.168 0.033 0.00093 -1.29 23 53 13.3 0.415 0.318 0.326 0.014 0.00087 0.19
65 23 45.3 0.250 0.400 0.201 0.075 0.00089 -1.29 23 53 28.9 0.246 0.365 0.334 0.062 0.00102 0.19
65 23 75.4 0.172 0.323 0.300 0.201 0.00059 -1.29 23 53 43.8 0.334 0.437 0.348 0.136 0.00179 0.19
65 23 75.4 0.187 0.323 0.300 0.201 0.00065 -1.29 23 53 58.4 0.452 0.5 18 0.363 0.2 18 0.00243 0.19
23 53 71.3 0.668 0.578 0.375 0.280 0.00268 0.19
25 23 31.9 0.460 0.408 0.159 0.03 1 0.00093 -1.58
25 23 45.4 0.363 0.390 0.189 0.066 0.00119 -1.58 34 23 75.6 0.271 0.325 0.320 0.29 1 0.00073 -0.13
25 23 60.9 0.234 0.357 0.240 0.127 0.00107 -1.58 34 23 60.5 0.2 18 0.381 0.3 13 0.233 0.00104 -0.13
25 23 76.2 0.187 0.320 0.298 0.196 0.00071 -1.58 34 23 46.5 0.267 0.453 0.304 0.160 0.00 147 -0.13
34 23 41.5 0.289 0.479 0.300 0.134 0.00154 -0.13
42 23 15.2 0.865 0.380 0.120 0.006 0.00065 -2.70 34 23 32.1 0.390 0.527 0.294 0.085 0.00 187 -0.13
42 23 21.3 0.584 0.378 0.126 0.011 0.00065 -2.70 34 23 23.3 0.462 0.564 0.289 0.046 0.00 176 -0.13
42 23 31.1 0.372 0.373 0.138 0.025 0.00065 -2.70
42 23 44.5 0.250 0.362 0.168 0.059 0.00077 -2.70 21 23 75.5 0.166 0.336 0.3 19 0.405 0.00070 -0.14
42 23 60.4 0.221 0.342 0.223 0.116 0.00102 -2.70 21 23 60.4 0.2 10 0.422 0.307 0.3 18 0.00150 -0.14
42 23 74.4 0.241 0.3 19 0.285 0.182 0.00099 -2.70 21 34 35.2 1.342 0.698 0.269 0.039 0.00262 -0.16
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 259

TABLE II.-Continued.

0° 900 50 AO0 Dis- 00 900 00 AO, Dis-

Fil- Fil- Aver- Stand- tance Fil- Fil- Aver- Stand- tance
ter ter age ard Average from 00 As ter ter age ard Average from 00 As
num- num- set- devia- chromaticity point Standard num- num- set- devia- chromaticity point Standard
ber ber ting tion x y =s deviation Ay/Ax ber ber ting tion x y =s deviation Ay/Ax
4 40 10.2 0.663 0.153 0.026 0.001 0.000135 0.37 17 16 30.4 0.370 0.436 0.558 0.078 0.00185 -0.89
4 40 20.3 0.379 0.156 0.027 0.004 0.000156 0.37 17 16 30.6 0.397 0.435 0.559 0.078 0.00198 -0.89
4 40 30.1 0.296 0.160 0.029 0.009 0.000179 0.37 17 16 30.8 0.419 0.434 0.560 0.079 0.00210 -0.89
4 40 39.8 0.266 0.167 0.032 0.016 0.000218 0.37 17 16 30.7 0.327 0.435 0.559 0.079 0.00164 -0.89
4 40 45.2 0.254 0.171 0.033 0.020 0.000224 0.37 17 16 40.4 0.353 0.392 0.596 0.135 0.00222 -0.89
4 40 50.1 0.274 0.175 0.035 0.025 0.000268 0.37 17 16 45.0 0.308 0.370 0.616 0.165 0.00214 -0.89
4 40 60.0 0.360 0.185 0.038 0.035 0.000372 0.37 17 16 45.4 0.348 0.368 0.618 0.167 0.00237 -0.89
4 40 69.7 0.532 0.194 0.042 0.044 0.000483 0.37 17 16 45.6 0.327 0.367 0.619 0.169 0.00222 -0.89
4 40 70.2 0.509 0.194 0.042 0.045 0.000463 0.37 17 16 50.1 0.357 0.343 0.641 0.201 0.00258 -0.89
17 16 59.2 0.352 0.291 0.687 0.268 0.00261 -0.89
4 66 21.0 1.733 0.151 0.03 10.003 0.00068 -2.43 17 16 59.7 0.418 0.283 0.694 0.272 0.00309 -0.89
4 66 29.5 1.185 0.150 0.032 0.006 0.00057 -2.43 17 16 60.1 0.486 0.281 0.696 0.275 0.00360 -0.89
4 66 29.9 1.297 0.150 0.032 0.006 0.00062 -2.43 17 16 60.5 0.456 0.278 0.698 0.278 0.00338 -0.89
4 66 45.1 1.100 0.145 0.042 0.017 0.00104 -2.43 17 16 73.1 0.815 0.217 0.752 0.360 0.00418 -0.89
4 66 68.5 0.860 0.132 0.075 0.053 0.00178 -2.43 17 16 74.9 0.738 0.210 0.758 0.370 0.00378 -0.89
4 66 69.0 1.060 0.131 0.077 0.054 0.00219 -2.43 17 16 75.3 0.890 0.208 0.760 0.372 0.00455 -0.89
66 55 34.2 1.280 0.112 0.136 0.031 0.00274 -4.28 21 33 21.6 0.397 0.723 0.276 0.019 0.00074 -0.99
66 55 44.1 0.715 0.106 0.161 0.057 0.00222 -4.28 21 33 31.2 0.358 0.706 0.293 0.043 0.00114 -0.99
66 55 60.1 0.470 0.090 0.229 0.127 0.00265 -4.28 21 33 46.2 0.216 0.659 0.340 0.109 0.00127 -0.99
66 55 60.2 0.536 0.090 0.230 0.127 0.00302 -4.28 21 33 50.7 0.188 0.638 0.360 0.138 0.00137 -0.99
66 55 70.1 0.422 0.076 0.291 0.190 0.00286 -4.28 21 33 60.5 0.136 0.576 0.422 0.225 0.00141 -0.99
66 55 70.2 0.400 0.076 0.292 0.190 0.00271 -4.28 21 33 70.4 0.112 0.492 0.505 0.343 0.00143 -0.99
66 55 74.1 0.560 0.070 0.317 0.217 0.00363 -4.28 21 33 75.3 0.143 0.447 0.549 0.406 0.00179 -0.99
66 55 80.1 0.770 0.062 0.352 0.253 0.00393 -4.28
66 55 81.6 1.000 0.060 0.360 0.261 0.00450 -4.28 48 33 21.6 0.661
0.521 0.338 0.032 0.00154 -0.98
48 33 45.6 0.576
0.161 0.422 0.153 0.00115 -0.98
55 15 29.3 1.053 0.057 0.431 0.052 0.00385 16.18 48 33 60.5 0.491
0.178 0.506 0.272 0.00155 -0.98
55 15 43.8 0.747 0.060 0.499 0.118 0.00419 16.18 48 33 70.3 0.430
0.179 0.566 0.358 0.00147 -0.98
55 15 59.3 0.595 0.064 0.594 0.215 0.00409 16.18
55 15 59.4 0.790 0.064 0.595 0.216 0.00542 16.18 40 21 29.3 0.354 0.229 0.056 0.028 0.00078 +0.41
55 15 59.5 0.806 0.064 0.595 0.217 0.00554 16.18 40 21 39.4 0.282 0.256 0.067 0.058 0.00107 0.41
55 15 73.7 0.985 0.068 0.687 0.308 0.00530 16.18 40 21 40.2 0.299 0.260 0.068 0.061 0.00114 0.41
55 15 74.3 1.046 0.068 0.690 0.312 0.00563 16.18 40 21 49.9 0.369 0.304 0.086 0.109 0.00240 0.41
55 15 74.4 1.012 0.068 0.691 0.312 0.00545 16.18 40 21 50.2 0.326 0.306 0.087 0.111 0.00212 0.41
40 21 60.1 0.366 0.381 0.118 0.192 0.00379 0.41
15 16 27.5 1.264 0.092 0.746 0.024 0.00205 0.45 40 21 69.9 0.287 0.499 0.166 0.320 0.00447 0.41
15 16 28.1 1.441 0.093 0.746 0.025 0.00234 0.45 40 21 74.5 0.268 0.566 0.193 0.392 0.00456 0.41
15 16 42.9 1.092 0.118 0.758 0.054 0.00225 0.45 40 21 75.0 0.268 0.575 0.197 0.402 0.00456 0.41
15 16 59.3 1.009 0.150 0.772 0.088 0.00196 0.45 40 21 79.6 0.308 0.646 0.226 0.478 0.00499 0.41
15 16 60.3 1.237 0.151 0.772 0.090 0.00240 0.45 40 21 79.8 0.324 0.648 0.227 0.479 0.00524 0.41
40 21 84.6 0.498 0.705 0.240 0.545 0.00502 0.41
17 16 20.9 0.445 0.465 0.532 0.036 0.00156 -0.89 40 21 84.7 0.514 0.710 0.241 0.548 0.00518 0.41
17 16 30.2 0.339 0.436 0.558 0.076 0.00170 -0.89

of adjustment in their investigation of the notice- ently of all other variations. In an oral report"
ability of wave-length differences in the spec- of the present investigation, the results were
trum. This method has been mentioned as a presented in terms of the probable errors of
possibility by several other investigators, but it color matching. In the discussion of that report,
has not been used to any considerable extent. the suggestion was made that the results be
In a very significant recent report 5 of an investi-
published in terms of the standard deviation,
gation similar to the present, Wright discussed the root mean square of the individual deviations
his reasons for adopting a criterion of definitely from the average setting. The significance of the
noticeable color differences. He concludes this concept of probable error is questionable when
discussion with the following paragraph: the distribution of errors does not fit the "nor-
"It is, of course, quite true that some varia- mal" distribution curve very accurately. In
tions in judgment are inevitable. No doubt the order to avoid implications not justified by the
ideal method would be to make a large number of observations, the results will be reported here in
matches at each point in the colour chart and terms of standard deviation rather than in
then to analyse the spread of the observations, terms of probable error. This change in the data
but in practice this would be an impossibly reported previously1 involved only the omission
lengthy process." of a constant factor, 0.6745.
The program which Wright describes as ideal The apparatus employed in this investigation
becomes practical with the apparatus used in is particularly suitable for the matching method
the present investigation, in which any one 11D. L. MacAdam, "Noticeability of color differences
desired variable can be investigated independ- in daylight," J. Opt. Soc. Am. 30, 657A (1940).
260 DAVID L. MACADAMI

TABLE III. Standard deviations of color matching.

0
00 9 ° 0° A5O As 00 900 00 A0, As
Fil- Fil- Aver- Stand- Stand- Fil- Fil- Aver- Stand- Stand-
ter ter age ard Average ard ter ter age ard Average ard
num- num- set- devia- cliromaticity devia- num- num- set- devia- chromaticity devia-
ber ber ting tion x y Ay/Ax tion ber ber ting tion x y Ay/Ax tion
-

80 16 63.1 0.281 0.258 0.450 -5.18 0.00388 72 46 50.9 0.430 0.212 0.550 -0.32 0.00266
84 35 57.1 0.445 0.258 0.450 -500 0.00424 30 31 38.8 0.341 0.212 0.550 0.01 0.00230
23 14 44.1 0.586 0.258 0.450 -2.72 0.00413 69 44 53.9 0.441 0.212 0.550 -0.80 0.00339
64 33 65.0 0.292 0.258 0.450 1.66 0.00323 54 57 55.3 0.408 0.212 0.550 0.97 0.00283
90 28 48.8 0.241 0.258 0.450 -0.47 0.00188 28 61 51.8 0.440 0.212 0.550 0.39 0.00212
90 28 49.1 0.240 0.258 0.450 -0.47 0.00186 28 61 51.0 0.453 0.212 0.550 0.39 0.00218
74 57 66.5 0.320 0.258 0.450 3.26 0.00398 75 16 64.1 0.480 0.212 0.550 -11.15 0.00555
23 14 43.8 0.563 0.258 0.450 -2.72 0.00397 23 6 60.7 0.631 0.212 0.550 -2.30 0.00491
60 44 32.4 0.548 0.258 0.450 -1.02 0.00237 23 6 61.9 0.554 0.212 0.550 -2.30 0.00431
54 17 48.0 0.304 0.258 0.450 0.26 0.00234 77 35 55.9 0.603 0.212 0.550 3.61 0.00448
54 17 48.1 0.317 0.258 0.450 0.26 0.00244 78 45 48.3 0.601 0.212 0.550 2.87 0.00308
80 36 64.0 0.315 0.258 0.450 -6.33 0.00430 78 45 47.2 0.610 0.212 0.550 2.87 0.00313
106 104 52.7 0.874 0.441 0.198 0.58 0.00270 34 23 41.7 0.304 0.475 0.300 -0.16 0.00152
119 73 50.7 0.400 0.441 0.198 3.24 0.00123 37 72 41.0 0.235 0.475 0.300 -2.76 0.00102
25 81 51.0 0.452 0.441 0.198 7.35 0.00108 84 27 54.5 0.330 0.475 0.300 0.30 0.00246
79 38 39.7 0.352 0.441 0.198 -0.06 0.00137 70 53 43.5 0.593 0.475 0.300 0.64 0.00276
79 38 40.0 0.306 0.441 0.198 -0.06 0.00119 106 88 67.3 0.373 0.475 0.300 1.39 0.00190
79 38 39.6 0.387 0.441 0.198 -0.06 0.00150
80 43 44.4 0.300 0.441 0.198 0.25 0.00190 86 14 59.3 1.481 0.150 0.680 -4.88 0.00918
42 52 53.5 0.350 0.441 0.198 1.54 0.00176 86 14 59.2 1.481 0.150 0.680 -4.88 0.00918
92 53 52.8 0.338 0.441 0.198 1.02 0.00241 85 24 57.9 1.233 0.150 0.680 -2.74 0.00757
65 70 50.0 0.401 0.44 1 0.198 -1.70 0.00095 30 36 61.0 0.626 0.150 0.680 1.45 0.00313
37 120 44.5 0.278 0.441 0.198 -0.39 0.00114 71 24 55.0 1.462 0.150 0.680 -2.00 0.00719
71 24 50.0 1.507 0.150 0.680 -2.00 0.00741
77 69 48.0 0.330 0.280 0.385 0.12 0.00138 78 14 61.6 1.012 0.150 0.680 -8.93 0.00720
70 44 50.0 0.218 0.280 0.385 -1.26 0.00185 78 14 63.8 0.948 0.150 0.680 -8.93 0.00674
50 59 64.8 0.265 0.280 0.385 1.37 0.00320 44 45 43.6 0.501 0.150 0.680 0.24 0.00200
75 33 50.2 0.596 0.280 0.385 2.77 0.00391 44 45 44.9 0.505 0.150 0.680 0.24 0.00202
84 61 53.5 0.560 0.280 0.385 5.54 0.00375 72 15 65.5 0.551 0.150 0.680 -0.77 0.00359
84 61 53.6 0.566 0.280 0.385 5.54 0.00379 72 15 64.9 0.539 0.150 0.680 -0.77 0.00351
80 35 60.5 0.296 0.280 0.385 -11.47 0.00326 54 16 67.8 0.593 0.150 0.680 3.03 0.00489
46 35 52.6 0.587 0.150 0.680 0.62 0.00250

of determining the noticeability of color differ- of such discrepancy before the regular observer
ences. Since the equality of luminance is auto- commenced his observations.
matically maintained by the instrument, only Each of the angles observed was then sub-
one adjustment need be made by the observer tracted from the average for the series, and
in order to establish the match. The analysis of these deviations were squared. The sum of the
the results of observations is simple and un- squares of the deviations was divided by the
ambiguous, since variations of only one quantity number of observations less one. The square root
need be considered. Fifty adjustments of the of this quantity is the standard deviation for a
instrument for color match were customarily single observation of the angle for color match.
made at one sitting. The average of the observed This deviation of prism angle can be converted
angles was taken as the correct reading for the to standard deviation of distance in the chro-
color match. This practice compensated auto- maticity diagram by the use of Eq. (11). This
matically for any otherwise undetectable instru- standard deviation of chromaticity match can
mental errors. The difference between the be assigned to the chromaticity corresponding to
average and the angle to which the other prism the average observed angle.
was set throughout the series was an indication The criterion of color match is so familiar,
of the perfection with which the instrument had and with the present instrument the adjustments
been prepared for the observations. In cases of required to establish a color match are so easy
appreciable discrepancies, the observations were that the method has been found to be most
repeated after the instrumental arrangements useful in the investigation of the noticeability of
had been improved. Such imperfect arrangements color differences. This method has proved
of the instrument arose principally from slight practical in extensive tests of all kinds of color
nonuniformities in the filters. The experimenter differences for a few observers, and in tests for a
customarily made a short series of observations few important kinds of color differences, using a
of his own, and usually detected any serious large group of untrained and previously un-
disagreement between the angles of the two experienced observers. Extensive tests have
prisms at color match, and corrected the cause indicated that the just noticeable differences of
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 261

TABLE III.-Continued.

00 900 00 AO0 As 00 900 0 AO0 As

Fil- Fil- Aver- Stand- Stand- Fil- Fil- Aver- Stand- Stand-
ter ter age ard Average ard ter ter age ard Average ard
num- num- set- devia- chromaticity devia- num- num- set- devia- chromaticity devia-
ber ber ting tion x y AY/Ax tion ber ber ting tion x y Ay/Ax tion
.

106 34 59.0 0.464 0.5 10 0.236 0.53 0.00234 63 96 64.6 0.292 0.253 0.125 1.22 0.00227
39 70 44.0 0.453 0.510 0.236 -0.08 0.00173 41 97 57.7 0.392 0.253 0.125 3.84 0.00121
25 53 52.9 0.400 0.510 0.236 1.30 0.00202 40 99 62.1 0.437 0.253 0.125 1.71 0.00221
37 91 45.1 0.356 0.510 0.236 -5.87 0.00114
38 81 50.9 0.452 0.510 0.236 -0.91 0.00120 117 100 43.8 0.502 0.160 0.057 1.41 0.00084
65 52 49.7 0.428 0.510 0.236 2.08 0.00185 117 100 45.3 0.458 0.160 0.057 1.41 0.00079
65 52 49.6 0.327 0.510 0.236 2.08 0.00141 118 106 30.3 0.265 0.160 0.057 0.44 0.00047
65 52 49.4 0.405 0.510 0.236 2.08 0.00173 4 83 56.4 0.493 0.160 0.057 2.56 0.00081
40 47 62.3 0.278 0.160 0.057 -0.32 0.00036
69 33 48.8 0.643 0.380 0.498 -11.90 0.00264 40 47 58.5 0.276 0.160 0.057 -0.32 0.00036
95 57 51.7 0.329 0.380 0.498 -3.02 0.00191 63 66 50.9 0.287 0.160 0.057 -1.30 0.00040
23 59 62.1 0.760 0.380 0.498 2.03 0.00373
23 59 60.8 0.939 0.380 0.498 2.03 0.00462 26 87 53.7 0.564 0.365 0.153 1.16 0.00294
99 32 63.1 0.560 0.380 0.498 7.19 0.00353 26 87 52.4 0.610 0.365 0.153 1.16 0.00318
68 110 54.9 0.388 0.380 0.498 -1.09 0.00132 26 87 53.0 0.577 0.365 0.153 1.16 0.00302
68 110 51.9 0.444 0.380 0.498 -1.09 0.00150 92 81 49.3 0.576 0.365 0.153 1.67 0.00189
108 31 40.4 0.655 0.380 0.498 0.52 0.00163 25 80 52.5 0.462 0.365 0.153 -0.24 0.00122
108 31 41.8 0.833 0.380 0.498 0.52 0.00208 42 79 52.6 0.474 0.365 0.153 -2.43 0.00078
60 109 62.8 0.848 0.380 0.498 1.19 0.00219 76 38 41.5 0.332 0.365 0.153 0.21 0.00175
60 109 62.4 0.894 0.380 0.498 1.19 0.00224 119 70 51.5 0.590 0.365 0.153 -3.18 0.00132
41 105 67.6 0.607 0.365 0.153 0.66 0.00336
64 79 30.4 0.277 0.160 0.200 0.02 0.00099 92 81 51.0 0.554 0.365 0.153 1.67 0.00182
29 75 45.7 0.390 0.160 0.200 0.65 0.00095 41 105 67.3 0.666 0.365 0.153 0.66 0.00369
82 8 55.2 0.376 0.160 0.200 -1.44 0.00160
47 85 62.2 0.227 0.160 0.200 4.53 0.00168 38 67 56.3 0.185 0.527 0.350 -4.08 0.00128
66 102 62.6 0.353 0.160 0.200 2.46 0.00128 106 52 70.1 0.284 0.527 0.350 1.25 0.00201
74 112 41.6 1.127 0.160 0.200 -6.06 0.00203 48 60 44.0 0.293 0.527 0.350 -0.23 0.00189
23 27 54.9 0.332 0.527 0.350 0.07 0.00194
84 104 47.5 0.270 0.390 0.237 0.03 0.00126 120 53 62.8 0.416 0.527 0.350 0.44 0.00252
80 87 48.6 0.459 0.390 0.237 0.82 0.00240
80 87 48.4 0.452 0.390 0.237 0.82 0.00236 79 71 54.5 0.185 0.305 0.323 -2.59 0.00101
25 93 44.7 0.248 0.390 0.237 -1.76 0.00098 75 17 43.4 0.361 0.305 0.323 1.03 0.00203
42 69 60.5 0.225 0.390 0.237 92.3 0.00123 74 67 56.8 0.247 0.305 0.323 1.28 0.00228
92 91 63.4 0.258 0.390 0.237 2.42 0.00178 84 116 49.9 0.471 0.305 0.323 1.76 0.00213
26 89 66.8 0.258 0.390 0.237 1.64 0.00216 112 95 50.9 0.191 0.305 0.323 0.27 0.00095
37 98 50.9 0.176 0.390 0.237 -0.54 0.00107 99 78 34.1 0.272 0.305 0.323 -0.84 0.00085
65 99 63.2 0.286 0.390 0.237 -1.38 0.00105 84 72 45.7 0.475 0.305 0.323 2.09 0.00215
100 27 48.2 0.274 0.390 0.237 0.42 0.00197
100 27 46.9 0.237 0.390 0.237 0.42 0.00170 79 48 47.1 0.389 0.596 0.283 0.34 0.00241
39 51 38.9 0.434 0.596 0.283 -5.70 0.00132
70 72 65.1 0.471 0.385 0.393 -26.10 0.00242 37 27 59.8 0.376 0.596 0.283 1.20 0.00152
79 109 66.8 0.440 0.385 0.393 5.15 0.00306 43 69 35.9 0.324 0.596 0.283 -0.55 0.00138
97 67 58.1 0.511 0.385 0.393 1.74 0.00361 21 73 50.2 0.440 0.596 0.283 -0.14 0.00203
97 67 59.0 0.506 0.385 0.393 1.74 0.00357 81 49 51.0 0.503 0.596 0.283 0.03 0.00236
99 31 48.1 0.771 0.385 0.393 2.61 0.00375
103 94 59.3 0.204 0.385 0.393 -0.67 0.00160 28 61 37.5 0.377 0.131 0.521 0.39 0.00212
98 51 44.9 0.277 0.385 0.393 0.39 0.002 10 94 30 55.3 0.560 0.131 0.521 -0.36 0.00222
23 62 44.5 0.552 0.385 0.393 0.85 0.00247 8 36 55.7 0.366 0.131 0.521 3.60 0.00326
77 24 48.8 0.512 0.131 0.521 -7.21 0.00358
70 23 52.0 0.412 0.344 0.284 -1.01 0.00089 77 24 47.3 0.591 0.131 0.521 -7.21 0.00415
102 73 47.3 0.197 0.344 0.284 0.09 0.00102 78 44 43.0 0.956 0.131 0.521 -2.00 0.00418
84 96 48.0 0.409 0.344 0.284 0.61 0.00142 98 44 57.8 0.643 0.131 0.521 -2.04 0.00474
97 95 45.7 0.557 0.344 0.284 0.88 0.00190 23 46 64.6 0.546 0.131 0.521 -1.12 0.00389
80 69 59.9 0.393 0.344 0.284 2.70 0.00198 23 46 65.0 0.587 0.131 0.521 -1.12 0.00416
80 69 60.4 0.437 0.344 0.284 2.70 0.00222
79 108 45.1 0.394 0.344 0.284 -12.85 0.00163 76 23 57.4 0.342 0.278 0.223 3.80 0.00137
76 23 57.5 0.360 0.278 0.223 3.80 0.00145
76 93 63.9 0.310 0.270 0.275 8.94 0.00178 100 99 49.2 0.494 0.278 0.223 1.05 0.00129
100 99 49.1 0.518 0.278 0.223 1.05 0.00136
74 116 47.8 0.427 0.228 0.250 1.38 0.00228 82 69 56.7 0.367 0.278 0.223 1.61 0.00224
76 98 63.0 0.336 0.228 0.250 -5.38 0.00142 79 75 51.8 0.258 0.278 0.223 -0.42 0.00046
83 93 61.5 0.507 0.228 0.250 2.84 0.00296 106 75 62.3 0.197 0.278 0.223 -0.70 0.00059
82 94 58.0 0.378 0.228 0.250 6.05 0.00272 82 69 55.1 0.411 0.278 0.223 1.61 0.00252
79 8 47.2 0.158 0.228 0.250 -0.38 0.00080 80 86 41.8 0.241 0.278 0.223 -2.91 0.00089
50 23 60.1 0.3 16 0.228 0.250 0.94 0.00160 74 120 68.4 0.407 0.278 0.223 0.60 0.00103
74 120 69.0 0.407 0.278 0.223 0.60 0.00103
100 30 60.2 0.397 0.152 0.365 -2.09 0.00347 80 98 47.2 0.263 0.278 0.223 -1.89 0.00068
83 113 70.0 0.353 0.152 0.365 -7.23 0.00360
84 28 56.0 0.514 0.152 0.365 -1.29 0.00324 76 79 52.7 0.798 0.300 0.163 0.79 0.00181
64 71 58.5 0.373 0.152 0.365 3.23 0.00265 26 120 60.6 0.237 0.300 0.163 158.0 0.00080
8 94 45.4 0.416 0.152 0.365 0.89 0.00175 26 120 60.4 0.237. 0.300 0.163 158.0 0.00080
23 55 54.2 0.526 0.152 0.365 -0.23 0.00232 83 105 49.7 0.255 0.300 0.163 0.33 0.00121
106 100 46.2 0.261 0.300 0.163 -0.16 0.00062
66 76 47.5 0.221 0.187 0.118 0.15 0.00054 41 96 61.1 0.459 0.300 0.163 1.48 0.00283
107 75 56.5 0.652 0.187 0.118 2.88 0.00207 41 96 61.3 0.413 0.300 0.163 1.48 0.00253
107 75 56.0 0.608 0.187 0.118 2.88 0.00193 92 97 59.3 0.258 0.300 0.163 -2.36 0.00076
107 75 52.9 0.700 0.187 0.118 2.88 0.00222
63 98 65.3 0.373 0.187 0.118 6.82 0.00197 70 62 62.6 0.520 0.472 0.399 2.22 0.00290
63 98 64.5 0.324 0.187 0.118 6.82 0.00172 38 109 64.3 0.204 0.472 0.399 -1.69 0.00158
82 50 35.6 0.393 0.187 0.118 -0.41 0.00057 103 72 56.1 0.278 0.472 0.399 -1.18 0.00155
47 100 59.6 0.310 0.187 0.118 1.02 0.00092 69 52 41.7 0.449 0.472 0.399 0.14 0.00152
40 101 63.3 0.236 0.187 0.118 -3.61 0.00082 73 67 49.9 0.431 0.472 0.399 -16.0 0.00176
23 58 57.3 0.422 0.472 0.399 0.40 0.00231
26 8 45.5 0.160 0.253 0.125 -0.97 0.00057 81 17 54.3 0.456 0.472 0.399 5.85 0.00201
26 101 52.2 0.158 0.253 0.125 -0.82 0.00047 99 51 60.6 0.546 0.472 0.399 0.80 0.00269
107 80 66.2 0.390 0.253 0.125 0.76 0.00153 87 116 45.8 0.352 0.472 0.399 -0.72 0.00111
82 106 37.6 0.277 0.253 0.125 0.28 0.00067 99 51 58.9 0.593 0.472 0.399 0.80 0.00292
262 DAVID L. MAcADAM
P= EXCITATIONPURITY PI EXCITATION PURITY
IC O16__
89 II %X
I ! T, 2S T 80x 60% 40% 20% 0 20% 40x 60x 80x 00%

4 55mm AS=
Q004
Q
AP= 0.005
14
490mM _' 597 mu ,
0.005

_
0.003 0. I. 0.004 .0a004

0002 0.003

0001
't2
'4
4 70mw
4747 0,002
3'-. <~~~2
.~~~~~ ~~~~~ 02?
a. 0.001 3.001
0 0.1 0.2 0.3
S- _ . 0_ _
061 02 03
FIG. 8. Standard deviations of purity matching for
dominant wave-length 455 m (PGN, solid curve and FIG. 11. Standard deviations of purity for dominant
circles); and 470 mru (DLM, broken curve). wave-lengths 490 and 597 mA (PGN, circles and solid
curve; DLM, broken curve).
PI EXCITATION PURITY
P- EXCITATION PURITY
looK 80K 60K 40 0. x ox 20 40 SO O- I

0005 . 00.
F ala ~~~700o t 40 3 A" , f

03 03

0002 "., ,°002

0001 1 oo
1,2( ,'3i4 ,923
,i2 , , 5t,
3- 0 31 0 0.1 0.2 0330 1 32 0.3

FIG. 12. Standard deviations of purity for dominant

wave-lengths 493 and 700 my (PGN, circles and solid
curve; DLM, broken curve).

FIG. 9. Standard deviations of purity matching for P- EXCITATION PURI TY

0 60x 80x
dominant wave-lengths 476 and 576 mu (PGN, solid curve R0x 60x 40% ;0% 20X 40%

and circles; DLM, broken curve). as-

0.004 4 ~~~~~499
u 0.004
Pi EXCITATION PURITY A-
AP-
=8 6px IPX 2x 29x 4P 1 60x 80% IjOX 0003 0003
AS= ash
0.002 0D02
0.006 0.006
496 2 -2x S86m ,
0005 ooos 0001 0.001
38 V23 1 2
0004 1004 S 0 0.1 0.2 0 0.1 0.2 0.3

0003 0.003 FIG. 13. Standard deviations of purity for dominant

wave-length 499 mu and complementary (PGN, circles and
0002 0002 solid curve; DLM, broken curve).
0.001 00OI Pa EXCITATION PURITY
2 64 23 'SI
8OX 60x 40x 20X OX 20X 40% 60x 80x
.U 0., 02 0 0.1 0.2 0.3

.008
FIG. 10. Standard deviations of purity for dominant
wave-lengths 496 and 586 m (PGN, circles and solid 000f 0007
curve; DLM, broken curve).
0.007
0,00
k k ~~~505 A l .006
color, for constant luminance of approximately 0.006 .005
15 millilamberts, are proportional to the corre- 004
sponding standard deviations of color matching
,003
by the same observer. Fluctuations of the cri- 000:
terion of just noticeability will, of course, alter .002
Q~oI
the constant of proportionality, but this factor .oo
1,6
(approximately three) is constant throughout
S 0 al 020 0.1 Q2 0.3
any series of observations of the kind specified
in which the criterion of just noticeability has FIG. 14. Standard deviations of purity for dominant
wave-length 505 mAzand complementary (PGN, circles and
been proved to be constant. Therefore, the solid curve; DLM, broken curve).
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 263

P- EXCITATION PURI TY P- EXCITATION PURITY

808 608 40 2 08 20% 408 60X go
100% 0% 60% 40% 20% 0% 209 40X 60x 80x 100%

0.010. D.010 AS-

56
0000 300 Q 7 2.7 mi / D007

006 5208.. S 2%- %

-2', D06

0.00 0.005
4-/
004 . I . I . I
Q004
000 IQ005 .
0004 000. o003
000 0103
Q00; t -3 .002
3.002
I02
00.0 01 3001 0.00 ll ,/ 0.001
/4 <3'w
*. '21

S 0 Ql 0a2 01 o.;a 03 0.4 05

S= 0 0.1 0.2 0 0.1 0.2 03
FIG. 15. Standard deviations of purity for dominant 18. Standard deviations of purity for dominant
FIG.
wave-length 520 mu and complementary (PGN,rcircles and wave-length 562.7 mAuand complementary (PGN, circles
solid curve; DLM, broken curve). and solid curve; DLM, broken curve).
P' EOCITATION PURI TY
SO 608 408 20 0 20*
- 40X 60X 806 P= EXCITATION PURITY
80x 60% 40% 20% OX 20X 40X 60X 80 100x
0007 3 ).007 4I0
AS=
537.5 "'1 < 006
0o007 ).007
Zx-
0005 X005 440 p 567.5 mp
2X6
*1
-1, 0006 .006
CID04- 004 -2% 2x/
0003 ).003 00051- ).005
18
0002 0002 OD04 0 /P .004
0001 00001 '63
' 23 '30, 000 .003
S=0 01 020 I 0.1 02 0.3 04 0.5
0oo
FIG. 16. Standard deviations of purity for dominant
wave-length 537.5 mu and complementary (PGN, circles 0.001 0.001
and solid curve; DLM, broken curve). ;63 .;2 3 * 59
uE | X

S= I 0.1 0.2 0.3 0 0.1 0.2

P= EXCITATION PURITY
80x.
v v - w
0xv -
40x
f
20X -
0 - - - - ' -
20%
- -
40% W -
60% 809 100 DX FIG. 19. Standard deviations of purity for dominant
as= I wave-lengths 440 and 567.5 mju (PGN, circles and solid
0.008 curve; DLM, broken curve).
559
0.007 I mp .0D07
AP=
)2x ' z
0.006 2%- 0.006

0.005
0.005 0.005
0004 .. ^""' 0

0.004 0.004
0003 ,,

0.003 W- 0003

0 - 0.002
0002 ,- , 0.002I
0 S- 01 02 3 04 05 6 0.7
0,001 - .001
QOI I g ,;23 1 I 3
I FIG. 20. Standard deviations of chromaticity along
4 76 1 0 .,I0 0.1 0.2 0.3 spectrum locus from 535 to 700 mgu (PGN, circles and
S= 0 0050 0.1 0.2
solid curve; DLM, broken curve).
FIG. 17. Standard deviations of purity for dominant
wave-length 559 myoand complementary (PGN, circles and
solid curve; DLM, broken curve). As.
8
OID05

OID04
standard deviation of color matching is con- -----------
0003 1.
sidered to be a satisfactory measure of the 01002
0 0
noticeability of color differences, equivalent in 0.001

significance to the just noticeable difference and '66

I
u uo u , u.l 9 0 1 0.2 0.3 0 Wl
more consistent in magnitude.
Tables II and III and the circled points and FIG. 21. Standard deviations of chromaticity along lines
near spectrum locus from 400 to 535 m1A(PGN, circles and
solid curves in Figs. 8 to 47 represent the results solid curves; DLM, broken curve).
264 DAVID L. MACADAM

derived from the observations of a single ob- l

server. Mr. Perley G. Nutting, Jr., whose interest Y=

and patience is acknowledged here with gratitude -0.120

and admiration, has made nearly 25,000 settings

for color match, and practically all of his results
are represented in the tables and curves. Figures
8 to 19 represent the standard deviations of -0.118

matching the purities of the indicated chro-

maticities, when purity alone was variable and
subject to error. The dominant wave-lengths are
indicated on the figures. Figure 20 represents the -0.1
standard deviations of chromaticity matching
when the chromaticity varied along a straight
line very close to the spectrum locus from 533
-0.117

-0.116

X= 0.196 0.197 0.188

FIGS. 24 TO 47. Standard deviations of chromaticity from

indicated standards, observer: PGN.

FIG. 22. Standard deviations of chromaticity along the I

,

I
straight line connecting points representing 440 and 700 muz
(PGN, circles and solid curve; DLM, broken curve). Note Y=
enlarged scale of portions of curve shown at left. -0.127
16

. | , . | . * * |
Y=
-0.0578 83

-0-126
-0.0576

-0.0574 6 101.
06
-0.125

82

0.0568 118
-0.124 107
.0566 63
41
.0564

II,
.0562 4
40
X=0.1592 0.1594 0.1596 01598 0.100 01602 01604 0.606 01608
_
X= 0.252 0.213 0,254

FIG. 23. Standard deviations of chromaticity from point

(x=0.160, y=0.057), observer: PGN. FIG. 25.
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 265

I, ,. mating that represented by the center of the

24 14 ellipse. The data represented in Figs. 23 to 47
24> \ \ i\ are summarized in Table III. In every case, a
filter was available which would alone produce
D.685 16 the chromaticity indicated by the center of the
ellipse. The angle to which the prisms of the
36
35
45 ,
.680 -0.526
44 ~~~~~~~~~~~~~~4
24
46 -0.525
72 46
3 -0.524

9675

26.\
FIG. 0.519
/ \\
myuto 700 m,7. Figure 21 represents the standard 0.522 3 6
deviations of matching the chromaticity along 8 3

straight lines close to the spectrum locus from -0.517 7

440 m to 533 m7m.Figure 22 represents the
standard deviations of matching the chro- 0.516

maticities along the straight line from the X=0,127 0.18 0.129 0.10 0,13I 0.132 0133 0.134 0.135

spectrum locus at 440 m,4 to the spectrum locus FIG. 27.

at 700 mg. This last curve corresponds to the
noticeability of complementary wave-length
differences for highly saturated purples. Y=
The broken curves in Figs. 8 to 22 represent -0.55 65

the results for similar observations by the 0.554 5

author. The discrepancies between the results
for these two observers are significantly large, 0.553

and appear to be correlated with position and 27

direction of the deviations in the chromaticity \ \ if
diagram. Both observers have normal color 0.551 4 61

vision according to all of the usual tests. The \\5/ 1

magnitude and character of the discrepancies 0.550 30 31

shown in Figs. 8 to 22 are comparable with -0,549 28 72

those reported by Wright.5 i f54

The ellipses shown in Figs. 23 to 47 represent 0.548 69

the noticeability of chromaticity variations in 0.547

all directions from the chromaticities indicated

at the centers of the ellipses. The observations -0.546 78

represented by each ellipse were obtained by 7745 23

selecting from five to eight pairs of filters, each 21

pair of which could be used in the instrument to I 0 021 0 0 (7T0.123 0.214 0.215 0.216

synthesize a chromaticity very closely approxi- FIG. 28.

266

-0.455

-0,454

-0.453
Y=
' 16
~~~36

57

63
DAVID L. MAcADAM

FY=
0.389
..
I

44
35
I

T 3
I *

-0.452 -0.386

-0.451 28 -0.385 9
77
17

-0.450 -0.3841

54 70
-a449 90 -0.383

-0.448 60 -0.382

-0.447 64 -0.381 75 84

X= 0.276 0.277 0.278 0.279 0.280 0.281 282 0.283 0.284

. I
-0.446 - ,

FIG. 3 1.
-0.445

;( .254 02F55 0256 0.257 0.258 02F590.260 0.261I 0.262

Y= ~~~
~~~32
5

,044
-0.502
FIG. 29.

. . . . . . . .
0.501 33

Y=
-0.369 113 -0.500

-0,368 287 -0-409 ~110 31

106S68~~~~~~~10

-0.367
94
-0.366 -0.497 108
55
-0.365 0.496
23 60
-0.364
-0.495 69
8
-0.363
-0.494 99
23
-0.362 04 84 X=0.376 377 0.378 0.379 0.380 0.381 0.382 0.383 Q384

-0.361 83 FIG. 32.

X= Q148 0.149 0.150 0151 0.W52 Q153 0954 0155 Q156
satisfactory match was obtained. The standard
FIG. 30. deviation of fifty observations was determined.
This deviation was represented by two opposite
instrument should be set in order to synthesize radii from the center point towards the points
this standard chromaticity with each pair of representing the chromaticities of the two filters
filters was both calculated and confirmed by used for the synthesis. Each pair of filters used
observation of the resultant mixture in compari- for the synthesis of the standard color resulted
son with the light transmitted by the single in the establishment of another diameter of the
filter. With one of the prisms set at this angle, ellipse. The complete ellipse is drawn through
so as to synthesize the desired chromaticity, the the points at the ends of these diameters. Several
other prism was rotated by the observer until a different coordinate scales are employed in
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 267

Figs. 23 to 47 in order to represent the ellipses the ellipse around the point representing any
with an accuracy comparable to that of the chromaticity ordinarily encountered in art or
observations. All of these ellipses are represented commerce. From such an ellipse the noticeability
to the same scale on the composite diagram of a color difference in any direction can be
in Fig. 48. In this figure, every ellipse is drawn deduced. For instance, the ordinates of the
ten times its correct size with relation to the curves shown in Figs. 8 to 22 could be deter-
coordinate scale of the chromaticity diagram. mined from a sufficiently complete series of
The centers of the ellipses are placed at their ellipses. Such a series of ellipses would therefore
proper locations in the chromaticity diagram. represent all the information contained in all
Sufficient ellipses are shown in Fig. 48 to make the curves for the noticeabilities of purity and
possible the estimation of the shape and size of dominant wave-length change. In addition, these
IF I I I I I

Y=
-0,204 -0.327

-0.203 -0.326

112
-0.202 85 -0.325 7267
7
71
-0.201 -0.324

95
-0.200 64 79 -0.323
112
29 99
-0.199 -0.322
75 7
-0.198 47 -0.321 74

-0.197 -0.320

-0.196 -0.319

X= 0Q156 0.157 0.158

I
0.159
I
0.110
I
0.161
~~~~~~~~~~~~~~~~~~~~~~~I
0.162 0.163 0164 X= 0.301 0302
I l X
303
I
X..
0.304
I
0305
l_
0.306 0.307
l
0.308
l
L309

FIG. 33. FIG. 35.

. I I I 1 lI I I I I I I I

Y= Y=
-0.254
-0.397 1931 7

-0.253 9 3 0.396

-0.252 8 6 -0.395 62

23
-0394 99
8
0.250 0.393
79

-0.24 9 .0392 98 03

50
-0248 76 -0.391 23

0.390 7
97 7
-0.24 6 -0.389 9

X= 0.224 0.225 0.226 0.227 0.2P8 0.229 0.230 0.231 0.2132 X=0.381 0.382 0.383 0.384 05 3P6 0.387 0.388 0.389

FIG. 3. FIG. 36.

268 DAVID L. MACADAM

ellipses represent the noticeabilities of all Undue emphasis should not be placed on this
conceivable combinations of purity and domi- curve, since it is based on extrapolations of
nant wave-length differences. observed data. This curve facilitates comparison
Figure 49 represents in terms of wave-length of the present results with previous results for
the standard deviations of color matching for wave-length discrimination. Such data represent
spectral colors, at constant luminance. This curve a limiting special case of color discrimination.
was derived from Figs. 20 and 21, together with The curve in Fig. 49 cannot be expected to be
the information contained in Fig. 48, which as reliable as the best results secured with
permits the estimation of noticeabilities of color apparatus and methods designed specifically
differences in other directions, and for other for the investigation of wave-length discrimina-
chromaticities than those actually observed. tion in the spectrum. Conversely, the data for

I
Y=
-0.304

-0.303

-0.401 -0.302 88 53
2
-0.400 -0.301
27
72
-0.399 -0.300.

-0.398 -0.299 84'

34
70
-0.397 -0.298 6

-0.396 -0.297

-0.395 -0296
X=O.4680.469 0.470 0.4*1 .J72 0.473 0.474 0.475 0.476 X=.471 O.A720.473 0.474 0.475 0.476 0477 0478 0.479
I , I I I , ,

FIG. 37. FIG. 39.

Y. Y=
-0.354 -0.240

-0.353 0.239

-0.352 52 -0.238 5 53
67
_0,351 3
34
-a237 81
60
-0350 23 2 0236 39
48
-0.349 120 -0.235 . 38

a348~~~~~~3
-Q348 106 -0.234 25 65

0.347 .233

.346 0.232

X=0.523 0.1240.525 5'26 0.527 a528 50 6

I
529 0.530 0.531
I~~~~~~~~~~.
X=0. 0.507 0508 0.509 0.510 0.511 0512 .0.513 0.514

FIG. 38. FIG. 40.

 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 269

discrimination of colors of less than spectral the curve from 550 mu to 700 myu,which shows
purity, represented in Figs. 8 to 47, inclusive, no evidence of a secondary minimum near 630
secured with an instrument designed specifically my frequently reported,'3 -' 6 is derived directly,
for their investigation, should be more reliable E - h

than similar results obtained with apparatus

primarily designed for investigations of color y=
-0.241
matching and discrimination in the pure spec-
trum. The curve in Fig. 49 is essentially in -0240
agreement with the results of Kbnig and Diet- 89
-0.239 26 97
erici8 and of Wright and Pitt.12 The portion of 93
-0.238 27
I I I 98
Y=
-0.287
-0.237 8430
37
-0.236 .65
-0.286
80 42
0.285 -0.235

-0.234

-0.233

X=0.386 0.387 0388 0389 0.390 0.391 0392 0.393 0.394

FIG. 43.
0.281
I I I I -3

y.
0.280 -0.202

0.279 -. 201

X-0.592 0.593 0.594 595 0Q596a597 0.598 .599 0.600

-0.200 53
FIG. 41. 8173104
70
1399
120 43

Y= 0,198 73
-0.288 80 7

0.197 65
-0.287
69
0.1 6 42
-0.280
95
0.195
-C6285 23 96

73 -0.194
284 102
X=0.437 0.438 0.439 0.440 0.441 0.442 0.443 0444 0A45
0.283 84 70
97 FIG. 44.
-0.282 130. Steindler, "Die Farbenempfindlichkeit des normalen
80 und farbenblinden Auges," Sitzungsber. Akad. Wiss. Wien,
.-281 Math.-Naturw. KI., Abt. 2a, 115, 39 (1906).
14L. A. Jones, "The fundamental scale for pure hue and
retinal sensibility to hue differences," J. Opt. Soc. Am. 1,
-0.280 63-77 (1917).
x= 0.340 0341 0.342 0.343 0.344 Q345 0346 0347 0.348 15 H. Laurens and W. F. Hamilton, "The sensibility of
I I I I I I the eye to differences in wave-length," Am. J. Physiol. 65,
547 (1923).
FIG. 42. 16 E. P. T. Tyndall, "Chromaticity sensibility to wave-
12 W. D. Wright and F. H. G. Pitt, "Hue discrimination length difference as a function of purity," J. Opt. Soc.
in normal colour-vision," Proc. Phys. Soc. 46,459-468 (1934). Am. 23, 15-24 (1933).
270 DAVID L. MACADAM
without extrapolation, from the observed data The standard deviations As encountered in
represented in Fig. 20. The portion of the curve matching a neutral color very similar to Illumi-
most dependent on extrapolation falls between nant C are indicated in Fig. 35, and are tabulated
430 mAtand 470 m,4. For this reason, very little in the second column of Table IV. The corre-
significance can be attached to the absence of the sponding values of excitation purity Pe were
secondary minimum near 440 m reported by computed by dividing the values of s by the
all previous investigators. Figure 50, representing distance from the neutral point to the spectrum
the standard deviation of determination of locus for each wave-length. Colorimetric purity
complementary wave-lengths for highly satur. p, was computed by multiplying the excitation
ated purples, has been derived directly from purity by the values of a given in Table I of
Fig. 22 and should be quite reliable. reference 17. The logarithms of these values
of colorimetric purity are compared in Table V
Y.
with similar data for eight other observers. In

a~~~~~~~
X9
-0.227 Table V, values of log [p,(570m)/p,(X)] are
given for these nine observers. The values of
- 0.226 IGP and FGB were computed from the values of
69 p, given in Table I of reference 4. The values for
-0.225 23
FLW, WJM, and LCM were computed from the
-0.224 6 values of PC for light-adapted eyes, given in
98 0
75 Table V of reference 18. The data for FHGP and
-0.223
79
WDW were computed by subtracting the values
046 for 570 m from the values of the logarithms
-0.222 80
100 given in reference 19. The values for JHN were
-0.221 7
82 I, , , I . *I
Y=
-0.220 -0.157

-0.219 -0.156
87
X= 0.274 0.275 0.276 02,77 0.278 0.279 0.2,80 0.281 0.282
-0.153 81 05

FIG. 45.
I I I I I I
Y= -0.154 7
-0.167
~
-0151 ~ 8

-0.166
96 -al51 4
0.165

-0.164 9
12 9 -0.150
26

105 -0.149
-163 100 106
X=0.361 362 0.363 364 0.365 0366 0.367 368 0.369
83
-0.162 9
FIG. 47.
76 2
-0.161
17 D. L. MacAdam, "Photometric relationships between
complementary colors," J. Opt. Soc. Am. 28, 103-111
(1938).
-0.160 4 L. C. Martin, F. L. Warburton and W. J. Morgan,
18
"Determination of the sensitiveness of the eye to differ-
-Q159
ences in the saturation of colours," Med. Research Council,
Special Report Series, No. 188 (H. M. Stationery Office,
X 0.296 0297 0.298 299 0.300 0.301 0.302 0.303 0304 London, 1933).
I I .___t . I . I I _ I' I 19W. D. Wright and F. H. G. Pitt, "The saturation
discrimination of two trichromats," Proc. Phys. Soc. 49,
FIG. 46. 329-331 (1937).
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 271

.85

.80

.75

.70

.65

.60

.55

.50

.45

.40

.35

.30

.25

.90

.15

.10

.05

0.

FIG. 48. Standard deviations of chromaticity from indicated standards, represented ten times actual
scale on I.C.I. 1931 standard chromaticity diagram, observer: PGN.

similarly computed from those given in ref- correspond to the geometric means of the thresh-
erence 20. olds of colorimetric purity. Table IV indicates
The values for PGN, computed from Table IV, that the tremendous variation of the threshold
are quite similar to those for the other eight values of colorimetric purity through the spec-
observers, and to the average values, which trum is to a great extent an artifact resulting
from the use of the scale of calorimetric purity.
20 H. Nelson, "The colour vision characteristics of a
J.
trichromat," Part 2, Proc. Phys. Soc. 49, 332-337 (1937). The ellipse in Fig. 35 is considered a more satis-
272 DAVID L. MAcADAM

very few data of this kind have been available.

The solution of the general problem of the
noticeability of color difference is represented
most clearly in this way. The noticeabilities of
color differences involving simultaneous lumi-
nance and chromaticity variations are expected
to define ellipsoids around the points represent-
ing every color in the color solid. Such ellipsoids
will be required for the complete representation
of color differences. The ellipses shown in Fig. 48
500 600
may be regarded as the cross sections of the
WAVE-LENGTH (MILLIMICRONSI A ellipsoids, corresponding to constant luminance.
FIG. 49. Standard deviations of wave-length matching The development of a more elegant manner of
in spectrum, derived by extrapolations from Figs. 20, specifying the noticeabilities of color differences
21, 48.
than the system of ellipses suggested by Fig. 48
requires an application of the principles of
differential geometry.2 3 Such an analysis of the
results represented in Tables II and III and
Figs. 8 to 48 is now in preparation. A preliminary
investigation24demonstrated that results of the
0.5
TABLE IV. First step from white.

Wave-length As P.(%) Pc(%) -log Pc

560 550 540 530 520 510 500
COMPLEMENTARY WAVE-LENGTH (MILLIMICRONS) Xc 400 0.00205 0.59 0.0089 4.05
440 0.00209 0.62 0.0211 3.68
FIG. 50. Standard deviations of complementary wave- 450 0.00218 0.65 0.0364 3.44
length matching for highly saturated non-spectral colors 460 0.00219 0.66 0.047 3.33
(purples). 470 0.00221 0.69 0.126 2.90
480 0.00182 0.64 0.269 2.57
490 0.00104 0.40 0.37 2.43
factory representation of the data than the 500 0.00086 0.234 0.40 2.40
510 0.00092 0.178 0.42 2.37
customary curves showing the variation of log p, 520 0.00096 0.171 0.45 2.34
as a function of wave-length. This is one of 530 0.00107 0.211 0.54 2.27
540 0.00116 0.265 0.63 2.20
many clear examples of the advantages secured 550 0.00135 0.365 0.80 2.10
560 0.00172 0.55 1.03 1.99
by the use of the chromaticity diagram as the 565 0.00200 0.70 1.30 1.88
basis for the representation of color discrimina- 570 0.00223 0.82 1.41 1.85
575 0.00208 0.79 1.30
tion.2 ' The probability of accidental misrepre- 580 0.00172 0.65 1.00
1.89
2.00
sentation or misinterpretation of data is less 585 0.00145 0.53 0.76 2.12
590 0.00126 0.435 0.58 2.24
when the results are represented on a basis 600 0.00107 0.328 0.39 2.41
resulting in simple, regular figures than when 610 0.00099 0.274 0.29 2.54
620 0.00096 0.248 0.242 2.62
they are represented by curves exhibiting marked 630 0.00095 0.233 0.215 2.67
changes of slope, and several inflections. 640 0.00094 0.225 0.200 2.70
650 0.00094 0.221 0.191 2.72
The representation of the noticeabilities of color 660 0.00093 0.218 0.186 2.73
differences by loci around each point in the 670 0.00093 0.216 0.183 2.74
680 0.00093 0.215 0.181 2.74
chromaticity diagram has been suggested by 700 0.00093 0.214 0.179 2.75
several authors, including Martin, Warburton,
and Morgan,2 0 Judd,2 2 and Silberstein,23 but colorimetric coordinate system," J. Opt. Soc. Am. 26
421-426 (1936).
21 T. Smith, "The colour triangle and colour discrimina- 23L. Silberstein, "Investigations on the intrinsic proper-
tion," Discussion on Vision (Physical Society, London, ties of the color domain," J. Opt. Soc. Am. 28, 63-85
1932), pp. 212-226. (1938).
22 D. B. Judd, "Estimation of chromaticity differences 24 D. L. MacAdam, "Projective transformations of color-
and nearest color temperature on the standard 1931 I.C.I. mixture diagrams," J. Opt. Soc. Am. 32, 2-6 (1942).
 VISUAL SENSITIVITIES TO COLOR DIFFERENCES 273

TABLE V. Comparisons of first steps from white for nine observers. log [p,(570 mY)/Pc(X)].

Wave- Observers Aver-

length IGP FGB FLW WJM LCM FHGP WDW JHN PGN age

400 nmu 1.75 1.54 2.20 1.83

440 1.66 1.16 1.93 1.81 1.83 1.67
450 1.57 1.59 1.58
460 1.40 1.01 1.49 1.10 1.61 1.48 1.34
470 1.14 0.99 1.22 0.90 1.05 1.06
480 0.97 0.97 0.93 0.49 0.95 1.00 0.74 0.65 0.72 0.82
490 0.84 0.89 0.79 0.55 0.96 0.85 0.64 0.60 0.58 0.75
500 0.71 0.76 0.71 0.40 0.71 0.57 0.56 0.55 0.62
510 0.60 0.63 0.61 0.37 0.98 0.59 0.51 0.52 0.52 0.59
520 0.49 0.52 0.56 0.38 0.81 0.50 0.46 0.48 0.49 0.52
530 0.39 0.47 0.54 0.22 0.65 0.43 0.40 0.44 0.42 0.44
540 0.31 0.46 0.43 0.20 0.37 0.38 0.34 0.37 0.35 0.35
550 0.28 0.42 0.31 0.25 0.30 0.25 0.30
560 0:18 0.13 0.18 0.14 0.17 0.14 0.16
565 0.06 0.01 0.13 -0.18* 0.05 0.03 0.06*
570 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
575 0.03 0.22 0.04 0.09
580 0.36 0.32 0.44 0.39 0.37 0.18 0.28 0.15 0.31
585 0.46 0.39 0.26 0.27 0.34
590 0.52 0.46 0.54 0.44 0.45 0.39 0.46
600 0.63 0.58 0.50 0.39 0.77 0.63 0.63 0.55 0.56 0.58
610 0.72 0.70 0.70 0.77 0.67 0.69 0.70
620 0.82 0.78 0.71 0.44 1.05 0.74 0.85 0.76 0.77 0.77
630 0.90 0.80 0.79 0.93 0.85 0.82 0.84
640 0.98 0.82 0.78 0.55 1.12 0.81 0.99 0.96 0.85 0.87
650 1.03 0.84 0.85 0.52 1.11 0.85 1.04 1.00 0.87 0.90
660 1.06 0.85 1.06 0.88 0.96
670 1.08 0.85 1.14 0.89 0.98
680 1.08 0.86 0.92 0.85 1.29 1.18 0.89 1.01
700 1.27 0.90 1.08

Mean devia-
tion from
average 0.06 0.10 0.08 0.22 0.20 0.05 0.07 0.07 0.07

* The value shownfor WJM at 565 ma, -0.18, is go greatly at variance with the values for all other observers that it has been omitted in
computing the tabulated average. Better agreementswith the averagescouldbe secured by adding 0.24 to all of the values tabulated for WJM,
and by subtracting 0.20 from all of the values tabulated for LCM. The data for these two observers would then appear to differseriouslyfrom
those of the other observers only at the wave-lengths570 mg and 580 my. If the adjustments describedwere made, the mean deviationsfrom the
averagevalueswould be reduced to 0.09for both WJM and LCM.

kinds represented in Figs. 8 to 22 cannot be I.C.I. chromaticity diagram seems preferable

represented adequately by equal distances in for practical use, even though the system of
any projective transformation of the standard ellipses shown in Fig. 48 is required as supple-
1931 I.C.I. diagram. This fact eliminates the mentary data. The standard chromaticity dia-
possibility that such a transformation of the gram appears as convenient as possible for the
standard chromaticity diagram can constitute a representation of color differences,2 ' in the same
uniform chromaticity-scale diagram' adequately manner that plane maps of large portions of the
representing the present data. The more general surface of the earth are most convenient, regard-
representations23 25'26 which remain as possibili- less of the distortions of distances introduced by
ties for the construction of a truly uniform such representations.
chromaticity-scale diagram have very compli-
cated geometrical characteristics. The standard ACKNOWLEDGMENTS

25 R. H. Sinden, "A further search for the ideal color The author gratefully acknowledges the in-
system," I. "A new mechanico-graphical method," J. Opt. terest and assistance of Dr. Loyd A. Jones and
Soc. Am. 27, 124-131 (1937); II. "A reconsideration of the
Helmholtz line element," J. Opt. Soc. Am. 28, 339-347 Mr. Earl M. Lowry of these Laboratories, who
(1938).
26 W. Peddie, "The general applicability of Fechner's suggested this investigation, made available
law in colour sensation," Nature 124, 791-792 (1929); the special equipment employed, and contributed
"Colour vision and chromaticity scales," Nature 146,
717-718 (1940). essential ideas and advice at all stages of the
274 DOMINA E. SPENCER
work. The highly satisfactory performance of the in the ideas involved in this research was much
instrument is due in large measure to Mr. greater and more helpful than is implied by the
Ralph E. McAdam who was responsible for the fact that he was the principal observer. Many
mechanical details of its design, and to Mr. persons have made more or less extensive series
Max Zill who constructed the instrument with of observations, all of which contributed to the
great care and skill. It is a pleasure to acknowl- development of the method finally adopted and
edge again the essential contributions of Mr. to the understanding of the complex problems
Perley G. Nutting, Jr., whose critical interest involved.

MAY, 1942 J. 0. S. A. VOLUME 32

Calculation of Illumination from Triangular Sources

DOMINA EBERLE SPENCER
Massachusetts Institute of Technology, Cambridge,Massachusetts
(Received January 4, 1942)

1. INTRODUCTION 2. CHOICE OF THE COORDINATE SYSTEM

ALTHOUGH
actual light sources with tri- The use of the flux-density vector D is con-
angular boundaries occur but infrequently venient in this work.' A Cartesian coordinate
in practice, an important class of problems can be system is selected with the point P (Fig. 1) as
solved by the methods introduced in this paper. origin. The flux densities on three mutually per-
A knowledge of formulae for the triangular source pendicular planes through P are called D, D,,
allows the computer to calculate the luminous and D, and are considered as components of the
flux density, at any point on any surface, pro- vector D.
duced by a source having rectilinear outlines. The flux density at P on a tilted surface having
Any such source can be decomposed into tri- the normal vector N is obtained by vector addi-
angles and the flux densities produced by the z
individual triangles can be computed.
A special case of frequent occurrence is the
rectangular window, only part of which is effect-
ive in illuminating a given surface. The plane of
a tilted surface such as a drafting board will
ordinarily cut the rectangular window into two
parts, only one of which is useful in illuminating
the board. Either the shape of this illuminating
source is itself triangular or it may be decom-
posed into two, three, or four triangles. The
problem is an important one in illuminating
engineering and it is strange that no adequate
solution has hitherto been available.
The paper presents a derivation of general
equations for the flux density from triangular
windows. A set of nomographs for the convenient FIG. 1. The coordinate system and the spherical triangle.
solution of these equations is then given and an I See, for example, A. Gershun, "The light field," trans-
lated by Moon and Timoshenko, J. Math. Phys. 18, 51
illustrative example is worked out. (1939).