VOLUME 53, NUMBER 4
JOURNAL OF THE OPTICAL SOCIETY OF AMERICA
APRIL 1963
Effect of Wavelength on the Relationship between Critical Flicker
Frequency and Intensity in Foveal Vision*
AMEDEO GIORGIt
Departmient of Psychology, Fordhamn University, Bronx 58, New York
(Received 18 June 1962)
The precise effect of wavelength on the judgment of critical flicker frequency (CFF) has been a controversial issue for a number of years. Some studies report no effect of wavelength; other studies indicate that
wavelength is a basic determiner of the CFF threshold. The purpose of this study was to attempt a resolution of the controversy by means of a systematic and thorough investigation of the problem. The apparatus
used was the Fordham calorimeter. By means of monochromators it was possible to deliver various spectral
colors with a constant passband of 10 mg to the subject. The intensity of the light source was controlled by
means of neutral tint filters and an optical wedge. Flicker was produced by intercepting the beam of light
with a sector disk driven by a Graham variable-speed motor. Eight wavelengths covering most of the
range of the visible spectrum were used, and CFF thresholds were obtained at seven different luminance
levels in eight experimental sessions for each wavelength. Three subjects with normal color vision were
employed.
The results showed that the wavelength of the stimulating light changes the slope of the curve relating
CFF to log I. The curves at the blue end of the visible spectrum are steeper than those at the red end.
Furthermore, the greater the separation between wavelengths, the greater is the probability of a significant
difference between slope values. The fact that the slopes of the curves differed with wavelength does not
mean that the Ferry-Porter law is invalid, but rather that one should adjust the "constants" of the equation for the particular wavelength being used. Consequently, it can be stated that CFF is a function of the
wavelength of the stimulating light. The most explicit interpretation that could be made was that this change
reflected some sort of change in the receptor system of the eye.
INTRODUCTION
T
HE Ferry-Porter law is an expression of the
T
relationship between critical flicker frequency
(CFF) and intensity. It states that CFF is directly
proportional to the logarithm of the luminance of the
stimulating light, or in mathematical terms, that
N=a logl+b, where N= CFF in cycles per second,
I=luminance of the stimulating light, and a and b are
constants. Previous studies have demonstrated that
the relationship is actually sigmoid, but it is linear for
the middle luminance levels, so that most experimenters
acknowledge the breakdown of the law at the extreme
luminance ranges, but affirm its validity for the intermediate ranges of luminance.'
In their independent and initial proclamations both
Ferry2 and Porter' asserted that the law was valid for
colored light as well as white light. Because of these
statements, the inference was drawn that CFF was
independent of the wavelength of the stimulating light.
Thus, flicker photometry became one of the standard
procedures whereby lights of different color could be
matched for brightness. The assumption underlying
this procedure was that since CFF was systematically
related to luminance level, and since CFF was independent of wavelength, the point at which two lights of
different color fused was also the point where they were
equal in luminance.
*Submitted in partial fulfillment for the degree of Doctor of
Philosophy in the Department of Psychology at Fordham
University.
t Present address: Psychology Department, Duquesne University, Pittsburgh 19, Pennsylvania.
I S. Hecht, Bull. N. Y. Acad. Med. 14, 21 (1938).
2 E. S. Ferry, Am. J. Sci. 44, 192 (1892).
3
T. C. Porter, Proc. Roy. Soc. (London), A63, 347 (1898).
The statement that CFF is independent of wavelength was first suspected by Ives. 4 His data showed
that the relationship between CFF and logI was indeed
a straight line for all wavelengths, but the slope of the
line differed with wavelength. Ives' interpretation was
strengthened by the results of Allen,5 who also obtained
straight lines that differed in slope as a function of
wavelength. Consequently, both men claimed that CFF
was dependent upon wavelength as well as luminance.
Some time later, Hecht, 6 aware of the different interpretations in the literature, conducted a comprehensive
experiment in which he obtained CFF values as a
function of luminance for seven different wavelengths.
His results were of such a nature that he could draw a
single line through all the experimentally obtained
points at the photoptic luminance levels and this led
him to the conclusion that CFF was independent of
wavelength.
Most recently, however, Landis 7 has reversed the
decision again. On the basis of a review of all the
pertinent data and modified plottings of some of the
standard data, his claim is that the bulk of the evidence
indicates that wavelength is a determiner of CFF in the
same manner as flash rate, or light-dark ratio, or a
number of other variables.
Briefly, then, the effect of the wavelength of a light
upon CFF has been investigated a number of times, but
the results of these investigations have not been consistent. Some experiments indicate that CFF is indeE. Ives, Phil. Mag. 24, 149 (1912).
5 F. Allen, Phil. Mag. 38, 81 (1919).
6 S. Hecht and C. D. Verrijp, Proc. Natl. Acad. Sci. U. S. 19,
522 (1933).
7 C. Landis, Physiol. Rev. 34, 259 (1954).
480
4 H.
April 1963
EFFECT
OF WAVELENGTH
pendent of wavelength (Ferry, Porter, Hecht), whereas
others seem to show that wavelength is one of the
determiners of the CFF threshold value (Ives, Allen,
Landis). The purpose of this investigation is to attempt
a resolution of the conflicting results by means of a
thorough and systematic investigation of the relationship between CFF and luminance as a function of
wavelength.
ON
CFF
481
EYE OR PM
TC
DARKROCMWALL
#ML
TC OR M
CALIBRATION
0
SP
| B
---
0
|B
SP
APPARATUS AND PROCEDURE
The Fordham four-beam colorimeter was the apparatus employed in this experiment. A complete and
detailed description of this apparatus is given by
Zegers.8 Only the two beams actually used in the experiment are described here. From a functional viewpoint,
the apparatus can be divided into two systems: the
optical system and the calibrating system.
l l
l
t
lSl
l
~~ ~~SD
MONOCIHRGMATOR
N.
-
SSD
B SP
Optical System
A diagram of the optical system is presented in
Fig. 1. The color-producing units in the system are
83 Perkin-Elmer universal monochromators. One
of the four monochromators served as the color source
for the test patch and another as the color source for
the surround. The light source for each monochromator
was a 6-V, 9-A, tungsten vertical ribbon-filament bulb,
and the power for the bulbs was delivered by a Sorenson
Nobatron, Model E-6-40A. The function of the Nobatron, which received its power from the ordinary
110-V ac house mains, was to transform the current
from 110-V ac to 6-V dc. In addition, a 6-V storage
battery was connected across the Nobatron in order to
eliminate any minor current variations not removed by
the voltage regulator.
The test-patch beam of light was delivered by
monochromator No. 1, and it proceeded through the
system as follows: The diverging beam of light exiting
from the monochromator was first collimated by an
achromatic lens which was positioned at its focal length
from the exit slits of the monochromator and then it
successively passed through a beam splitter formed by
a 3-in. glass cube, a diaphragm which yielded a centrally
located field of 140' on the retina, a sector disk, a
second beam splitter that deflected the light 900, an
achromatic double-convex lens which initiated the
convergence of the light, and finally a negative lens
which sharpened the convergence in such a manner
that it formed a real image of the light source at the
artificial pupil.
The surround beam of light was delivered by monochromator No. 2 and it proceeded as follows: Exiting
from the monochromator as a diverging beam, the light
first passed through a double-convex lens which
collimated the beam, then it successively passed
7Nlodel
8R. T. Zegers, Psychol. Mono. 73, 1 (1959).
I
1
",,#" VP
|
MONOCHROMAToR
No. 2
B
FIG.
1. Schematic diagram of apparatus.
through a beam splitter similar to the one described
above, a diaphragm which yielded a visual angle of 7,
and finally a second beam splitter, at which point the
surround beam joined the path of the test beam and
followed the path described above to the artificial pupil.
The surround beam always matched the test beam in
wavelength and luminance; the former was determined
by calibration, and the latter by the subject's equality
matching.
Flicker was produced by a sector disk that was
driven by a 220-V Graham variable-speed transmission.
A Weston tachometer generator, Alodel 44, Type A,
which generates dc voltage in direct proportion to the
speed of the disk, and a Weston voltmeter calibrated in
rpm were used to measure the speed of the disk. The
flux of the test beam was controlled by means of Kodak
neutral-tint filters and the flux of the surround beam by
means of filters and a neutral-tint optical wedge.
Calibration System
The basic equipment used for the energy calibration
of the stimulus beam was a Farrand photomultiplier
photometer, a Farrand thermocouple and low-frequency
amplifier, and a Ballantine electronic voltmeter, Model
302-B. The three basic steps in the calibration procedure
were: (1) determination of the number of MW/V response on the thermocouple, which was made possible
AMEDEO
482
by the use of a standard lamp calibrated for irradiance
in gXV/cm 2 ; (2) the determination of u,W/unit response
of the photometer, which is accomplished by establishing a relationship between the volt response (which is
converted into MpW) generated by the test beam and
the number of unit responses obtained on the photometer; and (3) the determination of the uW at the
artificial pupil under the conditions of experimentation
by setting the photometer at the eyepiece under exactly
the same conditions as during experimentation and
observing the number of unit responses and converting
them into yuqW. Each of these steps was performed for
each wavelength employed in the experiment.
In order to calibrate the transmittance of the filters
and wedge, a Farrand photomultiplier photometer was
employed. Exact transmittance values for each filter or
filter combination were obtained under conditions
identical to experimentation.
Procedure
The design of this experiment was such that foveal
CFF-vs-logI curves were to be obtained for eight
different wavelengths from three subjects. The eight
wavelengths employed were 450, 480, 510, 540, 570,
600, 630, and 660 u and the order of presentation of
these wavelengths differed for each subject and from
session to session. A constant passband of 10 mu was
used for both test and surround beams. The subjects
were tested for color blindness and weakness by means
of an anomaloscope and none was found to be deficient.
Each experimental session was preceded by a 10-min
dark-adaptation period. The psychophysical method of
limits was employed in which each subject was first
presented with a visual field that was obviously flickering, and he was to judge when the light fused; and then
he was presented a light that was obviously fused and
he was to indicate when it first began to flicker. Three
alternate series of judgments such as these were obtained for each luminance level. A single experimental
session consisted of six threshold determinations for
each of seven luminance levels for a single wavelength.
There were eight experimental sessions for each wavelength and since eight different wavelengths were
employed, it means that a total of 64 experimental
sessions were obtained from each subject. Three subjects were employed in the experiment.
The CFF values were plotted against the logarithm
of the energy (logAuqW), and the result of these plots
was a straight line. The least-square solution for determining the line of best fit was applied to the data, and
the slope values were calculated. All subsequent calculations were concerned with the slope values.
RESULTS
This study was conducted in order to determine the
effect of wavelength on the relationship between CFF
GIORGI
Vol. 53
and radiant flux entering the artificial pupil. According
to the design of the experiment, if the slope of the curve
relating CFF thresholds to the logarithm of the flux of
the stimulating light remains the same in spite of
changes in the wavelength of the light source, then it
can be said that CFF is a function of the luminance of
the light only. If, however, the slope described above
changes with wavelength, then wavelength must be
exercising some separate influence on the CFF
threshold.
Table I contains the slopes obtained for each experimental session by each subject, plus the mean slope for
each wavelength. Inspection of this table reveals a
general pattern of decreasing slope value as one progresses from 450 to 660 mrn for all three subjects.
Actually, the highest slope value for all three subjects
is the one at 450 mA, and the lowest slope value for two
of the three subjects is at 600 myu. For the remaining
subject, the slope value for 600 mg is his second lowest,
his lowest coming at 660 mu.
With respect to variability, Table I shows that
Subjects II and III have practically the same degree of
variability, and that both are less variable than Subject
I. In addition, it can be seen that the variability is
TABLE
I. The slope for each experimental session and the mean
slope for each wavelength for all three subjects.
Session 450
630
660
I
II
III
IV
V
VI
VII
VIII
Mean
18.38
12.84
11.96
14.39
15.56
17.79
15.48
14.66
15.13
0vM 2.06
18.87
12.31
13.21
12.36
12.56
14.29
15.07
13.76
14.05
2.04
13.70
11.31
13.72
12.36
12.31
12.09
12.12
15.40
13.00
1.35
14.38
13.39
10.19
12.92
11.12
12.83
13.27
15.63
12.97
1.60
13.90
12.17
10.50
11.60
12.18
11.72
14.49
14.92
12.69
1.46
13.20
12.56
12.62
10.60
12.54
10.90
13.61
15.66
12.71
1.48
16.49
13.12
12.75
10.70
11.73
11.57
13.94
13.49
12.97
1.68
I 13.43
II 13.38
III 13.08
IV 15.19
V 16.21
VI 15.49
VII 14.73
VIII 14.96
Mean 14.56
afl
1.05
14.15
13.22
12.96
14.32
13.82
14.14
13.38
14.61
13.83
0.46
Subject II
13.86 13.90 13.07
15.18 12.96 13.20
13.47 13.07 12.88
14.42 13.80 13.07
14.26 12.93 13.00
13.99 13.53 12.95
14.57 13.92 13.57
14.90 13.61 13.74
14.33 13.47 13.19
0.52 0.39 0.29
12.40
12.71
12.12
13.01
13.01
12.65
13.65
13.35
12.86
0.45
14.29
12.82
12.00
12.78
13.03
11.38
12.87
13.98
12.89
0.89
12.90
12.49
11.50
12.82
13.48
12.35
12.39
13.12
12.63
0.56
I
II
III
IV
V
VI
VII
VIII
Mean
osef
12.67
12.45
13.65
12.97
12.93
12.36
11.55
12.12
12.59
0.58
Subject III
14.24 11.59 11.23
12.11 13.98 12.20
13.79 11.74 14.55
13.71 12.61 11.82
13.40 14.20 13.02
12.53 12.20 12.14
13.46 12.35 11.89
12.65 11.58 12.28
13.24 12.53 12.39
0.68 0.97 0.94
8.9
10.37
9.64
10.57
11.39
10.30
9.63
10.08
10.12
0.69
11.40
12.38
12.13
11.38
11.76
12.41
11.06
12.08
11.83
0.47
12.62
12.33
11.24
11.18
12.74
12.42
11.60
12.19
12.04
0.56
13.23
13.14
15.19
12.98
14.34
11.54
14.00
13.89
13.54
1.02
480
Subject I
Wavelength (mu)
510
540
570
600
17.04
11.85
13.05
13.10
13.40
13.45
14.33
14.96
13.90
1.47
I
I
April 1963
EFFECT
OF
WAVELENGTH
considerably greater about the blue end of the spectrum
than at any other portion.
A two-way classification analysis of variance was
applied to the data of each subject. In each case the
main effects tested were wavelengths and sessions, and
the first-order interaction was used as the error term.
This analysis reveals that for all three subjects the
difference among wavelengths is significant.
Snedecor's modifications of Tukey's test of comparison among means was applied to the data in order
to discover which particular wavelengths were different
from each other. These data are presented in Table II.
It can be seen that for Subject I the slope value obtained
for 450 m is significantly different from the slope
values obtained for 600, 630, 660, 570, and 510 myt and
ON
483
CFF
600
450
5UBJEC I
ma.
'5l
ze
91
/0
d5
0
1.5
225
3'0
Ad W
2.'0
Log
3s
4.0
4.5
FIG. 2. The wavelengths with the greatest discrepancy in slope
for Subject I. The points represent the obtained mean values for
eight sessions and the curves are the lines of best fit as determined
by the method of least squares.
II. Results of Snedecor's modification of Tukey's test
of comparison among mean slope values for all three subjects.
TABLE
*50
660
5UBJECT 4
Subject I
X (mm)
450
480
540
510
570
660
630
600
15.13
14.05
13.90
13.00
12.97
12.97
12.71
12.69
XS-12.69
2.44a
1.36
1.21
0.31
0.28
0.28
0.02
i ?-12.71 X-12.97 X-12.97 JX-13.00 X-13.90 k-14.05
2.42'
2.16'
2.16'
2.13'
1.23
0.98
1.34
1.08
1 08
1.05
0.15
1.19
0.93
0.93
0.29
0.03
0.03
0.26
0.26
X (m,)
450
510
480
540
570
630
600
660
k X-12.63
1.93b
14.56
14.33
1.70b
13.83 1.20b
13.47 1.09b
13.19
0.56
12.89 0.26
12.86 0.23
12.63
X-12.86 JX-12.89 X-13.19 X-13.47 X-13.83 X-14.33
1.67b
1.70b
1.37b
1.09b
0.73
0.23
1.47b
1.44b
1.14b
0.86b
0.50
0.97b
0.94b
0.64
0.36
0.61
0.58
0.28
0.33
0.30
0.03
IC
Subject II
Subject III
X(ma)
450
510
480
540
570
660
630
600
a
b
)X
13.54
13.24
12.59
12.53
12.39
12.04
11.83
10.12
X-10.12
3.42'
3.12'
2.47°
2.41c
2.2 7*
1.92'
1.71°
X-1 1.83 X-12.04 X-12.39 X-12.53 X-12.59
1.71'
1.50'
1.50'
1.01
0.95
1.41'
1.20
0.85
0.71
0.65
0.76
0.55
0.20
0.06
0.70
0.49
0.14
0.56
0.35
0.21
X-13.24
0.30
Difference greater than 1.96 is significant at 0.05 level.
Difference greater than 0.89 is significant at 0.05 level.
Difference greater than 1.20 is significant at 0.05 level.
that no other slope values were significantly different
from each other. For Subject II, the slope value for
450 muI is significantly different from the slope values
for all of the other wavelengths used in the experiment
except for the slope value for 480 and 510 mu, but also
for Subject II these latter slope values were significantly
different from most of the slope values for the wavelengths at the red end of the spectrum. Finally, Subject
III has the slope value for 600 mu differing significantly
from every other wavelength employed, and the slope
value for 450 m differing from the slope value for
570 mInplus the rest of the slope values for the wavelengths in the red region of the spectrum.
I
G. W. Snedecor, StatisticalMethods (Iowa State College Press,
Ames, Iowa, 1956).
To
_
0
_,
05
,
/0
,
I5
.
.
,
20
Log
2.5
ppW
3.0
.
35
.
4S.0
-
.
45
FIG. 3. The wavelengths with the greatest discrepancy in slope
for Subject II. The points represent the obtained mean values
for eight sessions and the curves are the lines of best fit as determined by the method of least squares.
'0
600
450
,5UBJECT 2T
0_
A,
-
,
0.5
,
/0
.
/5
.
20
LU,
.
.
'25
Y1
3.0
.
3.5
._
4.0
45
W
FIG. 4. The wavelengths with the greatest discrepancy in slope
for Subject III. The points represent the obtained mean values for
eight sessions and the curves are the lines of best fit as determined
by the method of least squares.
Generally, most of the significant differences are
between the red and blue regions of the spectrum. In
only one instance was a wavelength significantly
different from an adjacent (as employed in this experiment) wavelength, and that was the difference between
600 and 630 m that occurred with Subject III. In
484
AMEDEO
TABLE III. The log energy values and the corresponding mean
CFF threshold for the two wavelengths with the greatest difference
in slope for each subject.
GIORGI
Vol. 53
the ends of the spectrum deviated from the general
statement of his law, but he explained that these
deviations were probably due to the uncertainty of
observations in faintly illuminated regions. It was seen
Subject I
450 mg
600 ma
that the biggest differences in the present study were
loguuW Mean CFF (cps)
logguN V Mean CFF (cps)
also at the ends of the spectrum.
2.3759
24.3
1.243E8
24.9
Further on in the article Ferry2 states that color, at
2.5360
27.6
1.881(0
32.1
most, is a slight factor in retinal persistence and that
2.7401
30.8
2.233
37.7
2.9610
34.3
2.552(5
42.4
the all-important function is intensity. The present
3.1081
36.9
46.3
2.915l 3
experiment also shows that in comparison to intensity,
3.3237
39.6
wavelength effects are indeed slight. If one considers
3.6609
43.2
Subject II
that Ferry was, for the first time, formulating the
450 mu
660 mg
regularity of the intensity variable, one can readily
log,uuW Mean CFF (cps)
logpuN V Mean CFF (cps)
understand how he would consider the wavelength
2.0411
19.4
2.141' 1
19.6
effects as being minor. The important point is, there2.3769
23.6
2.488' 2
22.9
2.5360
26.1
2.800(0
27.0
fore, that Ferry likewise found wavelength effects, but
2.945: 6
2.7401
29.6
29.5
that thev were so small he probably considered them
3.1081
34.5
3.363.3
34.6
to be within the range of experimental error.
3.3237
38.0
3.705'7
39.0
3.6609
42.2
4.011:1
42.5
Porter's10 experiment was performed under conditions
Subject III
very different from Ferry's. 2 He had a disk that was
450 mu
600 m,
painted half white and half black which he illuminated
logiu,aW Mean CFF (cps)
logpuN V Mean CFF (cps)
with white light and lights of the various spectral
2.0411
19.5
1.243< 3
22.3
colors. He discovered that CFF varied as a function of
24.3
2.3769
1.881( 0
29.5
2.5350
26.6
2.233. 2
34.6
the intensity of the light only, and not with wavelength.
2.7401
29.2
2.552i 5
39.1
However, there are many limitations in Porter's experi3.1081
33.5
2.915S 8
43.9
37.0
3.3237
ment, due to lack of adequate apparatus, which could
3.403'
49.5
42.2
3.6609
have obscured the presence of wavelength effects. First
of all, his light sources were limelight and sunlight,
which are difficult to keep at a constant intensity for
order to demonstrate the order of magnitude of the any length of time. Secondly, the rate of rotation of the
differences between slopes for a given subject, the plots disk was found by the pitch of a note obtained by blowfor the wavelengths with the greatest discrepancy in ing air through a known number of equidistant holes
slope are presented for Subjects I, II, and III, respec- pierced in the circumference of the disk, and comparing
with those of a set of standard forks. Even
tively, in Figs. 2, 3, and 4. The curves are the lines of these notes
10
so,
Porter
reported that his results held within the
best fit as determined by the method of least squares,
range
of
experimental
error. How much error he was
and the points represent the obtained mean values for
eight sessions. Table III contains the data from which willing to tolerate is a matter of conjecture, but at least
it allows for the fact that some differences were found.
the figures were plotted.
Finally, Hecht's"" 2 data may be questioned for two
reasons. The primary limitation in Hecht's experiment,
DISCUSSION OF RESULTS
at least for this purpose, is the use of broad passband
The main finding of this dissertation points to the color filters. The transmittance curves of the Wratten
fact that CFF is affected by wavelength. If this is so, filters he used differed in passband from about 40 mA to
it seems that two questions remain to be answered: almost 100 mj.. Since these passbands were so wide,
first, how to reconcile this finding with those experi- there was often an overlap in the wavelengths being
ments that reported the opposite conclusion; secondly, transmitted although the primary wavelength passed
to ascertain what this finding means.
by two different filters were as much as 30 or 40 mjx
Those who claim that CFF is independent of wave- apart. This would tend to obscure wavelength effects.
length base their conclusion either on the work of
Secondly, there is the fact of Landis'7 replot of
Ferry,2 Porter,' 0 or Hecht."," Ferry 2 was the first to Hecht's data, which indicated that there was a differenreport such a finding when he published the article in tial effect of wavelength, in spite of Hecht's claim of no
which he formulated his law. Closer scrutiny of Ferry's difference. Landis plotted CFF vs wavelength for a
original article, however, indicates that he did find given luminance level, and since Hecht had equated the
wavelength effects, but he apparently considered them various colors, the line should have been parallel to the
unimportant. For example, he found that the values at x axis, but it was not. Consequently, Landis concluded
that color was affecting the CFF value.
T. C. Porter, Proc. Roy. Soc. (London) A70, 313 (1902).
Hence it can be seen that for various reasons the
S. Hecht, Bull. N. Y. Acad. Med. 14, 21 (1938).
12 S. Hecht and S. Shlaer, J. Gen. Physiol. 19, 965 (1936).
major articles in the literature reporting the independ-
I
April 1963
EFFECT
OF
WAVELENGTH
ence of CFF from wavelength can be brought in line
with those experiments that report CFF as dependent
on wavelength. These latter, on the other hand, remain
unchanged with respect to the major conclusion.
The problem of interpretation of the results of the
present study is something else again. The fact that the
blue end of the spectrum has a higher slope than the
red end means that, other things being equal, in terms
of flicker, the blue end of the spectrum is more efficient.
Efficiency here refers to the eye's ability to perceive
discrete stimuli as discrete, and for some reason when
working with blue the eye can do this better than when
working with red. Or stated in another way, changing
from red light to blue light will raise the CFF threshold
in somewhat the same manner as going from one
luminance level to a higher luminance level will raise
CFF. Those characteristics of wavelength or the eye
which could possibly account for this will now be
considered.
The primary psychological effect of change in wavelength is the corresponding change in hue which takes
place. A secondary effect is an apparent change in
luminance level of the light. This apparent change in
luminance is different for different colors. However,
because of the radiometric calibration employed, the
effect of this differential increment in brightness that is
associated with hue is clear-cut. It merely has the effect
of displacing the curve along the energy axis in accordance with its spectral luminosity. Therefore, this cannot
be the cause of the differences in slope value.
Consequently, one must turn to some physiological
explanation. Here two distinct possibilities remain:
Either some of the ocular media of the eye are acting
as differential filters, and are therefore changing the
slope, or this change in slope is reflecting a difference in
function of the receptor system of the eye. The first
possibility can be eliminated. When the various transmittances of the ocular media as presented by Judd' 3
are applied to the data, the only result is a change of the
curve along the intensity axis, and the slope does not
change in any way. Hence, as far as can be determined
by the author, the change in slope is reflecting some
change in the receptor system of the eye. Whether this
change is in the end organs (cones) themselves, or in
the brain, or in some of the intermediate structures or
tracts cannot be determined by the design of this
experiment.
Some positive support for the hypothesis that CFF
changes with wavelength is offered by a study by
Crittenden and Taylor.' 4 They report an investigation
where an attempt was made to determine the best
procedure for calibrating filters. Six sets of filters containing seven different colors were sent to six different
laboratories to be calibrated by certain specified
13D. Judd, Handbook of Experimental Psychology, edited by
S. S. Stevens (John Wiley & Sons, Inc., New York, 1951).
14S. C. Crittenden and A. H. Taylor, Trans. Illum. Eng. Soc.
25, 89 (1941).
ON
CFF
485
procedures. One of these procedures was the method of
flicker photometry. It was found that the spectrophotometric methods (i.e., methods that do not involve
judgment on the part of the observer) will give results
of high precision. With the flicker method, the averages
did not differ greatly from those obtained with the
spectrophotometric method, but the different laboratories showed systematic differences in the transmittances assigned to the filters. The systematic differences remained even after the results were corrected to
the basis of a normal observer by the Ives-test-solution
method. Consequently, Crittenden and Taylor while
admitting that the differences were small for such
difficult measurements nevertheless concluded that the
differences were real. They also felt that the differences
were so regular that further investigations would eventually lead to an explanation of the discrepancies.
The present investigation could well be the experiment they intended to conduct to find an answer to
their difficulties. If their results had been small but
irregular, then they could have attributed the discrepancies to individual differences. But they were not;
so they found them difficult to explain. This research
shows that CFF is a function of wavelength. Therefore,
when working with CFF and color unless the data are
corrected one would expect systematic differences such
as Crittenden and Taylor report.
Another important factor that may have tended to
obscure the wavelength effects is the attitude of the
experimenter in interpreting his results. Thus, one
experimenter may have obtained results that approximated a straight line and concluded that the line was
straight and that departures indicated experimental
error." Another may have interpreted the same type of
results literally, and then reached the conclusion that
the line was not straight.' 5 This experiment avoids this
possible source of error by basing the decisions on
statistical tests rather than on the experimenter's
judgment. This is the first investigation to do so in the
area of flicker and color.
The fact that two subjects differed significantly with
sessions and one did not indicates that there is significant interindividual variability with CFF as well as
intraindividual variability. This, of course, is not an
unprecedented finding. Landis 7 reports wide interindividual variability among CFF observers, and Tice' 6
claims a difference of as great as 25% between the CFF
of two different observers at the same intensity level.
The results of Tukey's test show which wavelengths
are significantly different from each other. It can be
seen that no two subjects have the same total number
of wavelengths significantly different from each other.
This means that individual differences also enter into
the wavelength effects, even though all the subjects
are consistent in that they all have at least some wave15 F.
16 F.
Allen, J. Opt. Soc. Am. 13, 385 (1926).
G. Tice, Psychol. Bull. 38, 691 (1941).
486
AMEDEO
lengths differing from each other. Subject I has the
fewest significant differences but this is probably
because he has the greatest day-to-day variability.
Subject III, who has the smallest day-to-day variability, is the only one who has two adjacent (as employed in this experiment) wavelengths significantly
different from each other. Again, this is understandable
because his variability is small enough to allow the
wavelength effects to demonstrate themselves.
The best generalization that can be made with
respect to this finding is that the wavelengths which
are at the opposite ends of the spectrum are the ones
which show greater degree of difference. The wavelengths in the middle regions of the spectrum, as a rule,
are not significantly different from the extremes or from
other wavelengths in the same region. Apparently, the
greater separation between the wavelengths the greater
is the probability of significance.
One final application of this finding is the light it
sheds on the differences that have been reported in
heterochromatic photometry between the equalityof-brightness method and the method of flicker photom-
GIORGI
Vol. 53
etry. 4 17 It had long been thought that the differences
between the two methods were primarily due to the
difficulty involved in trying to make a brightness match
between lights of different hue. Ferree and Rand,
however, have demonstrated that the differences between the two methods could not be explained away
solely by differences in hue, but that some other factor
was involved. This experiment shows that at least one
of the variables is the fact that CFF differs with color.
This fact does not necessarily eliminate flicker photometry as a color-matching procedure, but it does indicate
that one has to apply a correction factor when using this
method for matching colors.
ACKNOWLEDGMENTS
The author would like to express his deepest gratitude
to Dr. Richard T. Zegers, S.J., whose constant inspiration and sound advice were essential for the fulfillment
of this project. In addition, much gratitude is due to
Dr. Edward Hogan, CSSP, William C. Obert-Thorn,
and Miss Natalie L. Wright.
17C. E. Ferree and G. Rand, Psychol. Rev. 22, 110 (1915).