Copyright 1994 by the American Psychological Association, Inc.
0735-7044/94/S3.00
Behavioral Neuroscience
1994, Vol. 108, No. 5, 993-1004
Spectral Sensitivity, Photopigments, and Color Vision
in the Guinea Pig (Cavia porcellus)
Gerald H. Jacobs and Jess F, Deegan II
Behavioral discrimination tests and electroretinogram (ERG) flicker photometry were used to
measure spectral sensitivity and to define the spectral mechanisms of the guinea pig (Cavia
porcellus). Results from these 2 approaches converge to indicate that guinea pig retinas contain
rods with peak sensitivity of about 494 nm and 2 classes of cone having peak sensitivities of about
429 nm and 529 nm. The presence of 2 classes of cones suggests a retinal basis for a color vision
capacity. Behavioral tests of color vision were conducted that verified this prediction: Guinea pigs
have dichromatic color vision with a spectral neutral point centered at about 480 nm. The cone
pigment complement of the guinea pig is different from that known to characterize other rodents.
Cavia porcellus has been so frequently used in scientific and
medical investigations that its common name, guinea pig, has
entered the English language as an informal descriptor for the
subject of any sort of experimentation. In this role as subject, in
addition to that as frequent household pet and occasional food
source (Bradt, 1988), the guinea pig is surely among the most
familiar of the domesticated rodents. Given that fact, it is
remarkable how little is known about several basic features of
the visual system and vision in the guinea pig. For instance,
with regard to the identity of the photoreceptor and photopigment types and those visual behaviors most often associated
with their variations (e.g., spectral sensitivity and color vision),
there is sparse and often contradictory literature.
As for most other mammals, the guinea pig has a retina that
is rich in rods. The rod photopigment is a typical mammalian
rhodopsin with a peak absorption (A.max) of about 497 nm
(Bridges, 1959). There has been considerable inconsistency of
opinion about the presence of cones in the guinea pig retina.
Granit, who conducted many of the early studies of the
electrophysiology of the guinea pig retina, routinely labeled
the guinea pig as having a "pure rod-eye" (e.g., Granit, 1944, p.
103). Anatomists have offered variant opinions. O'Day (1947)
reported that he was able to confirm the early observations of
Kolmer (1936), who had detected cones in the guinea eye and
concluded, "To the histologist, the retina of the guinea pig is
certainly not "pure rod" (O'Day, 1947, p. 648). In the course of
an analysis of the fine structure of the retina by electron
microscopy, Sjostrand (reviewed, 1965) conducted a detailed
examination of the synaptic terminals of guinea pig photoreceptors. He differentiated two types of photoreceptors in the
guinea pig eye according to features of their synaptic morphologies. Some of these synaptic differences were of the sort that
have in other cases been used to differentiate rods and cones
Gerald H. Jacobs and Jess F. Deegan II, Department of Psychology,
University of California, Santa Barbara.
This research was supported by Grant EY-02052 from the National
Eye Institute. We thank Jason Siegel, Kris Krogh, and John Fenwick.
Correspondence concerning this article should be addressed to
Gerald H. Jacobs, Department of Psychology, University of California,
Santa Barbara, California 93106. Electronic mail may be sent to
jacobs@psych.ucsb.edu.
993
(e.g., lateral extent of the synaptic surface), but Sjostrand
argued these differences alone were not sufficient for determining whether a photoreceptor should be classed as a rod or a
cone, and, thus, he too referred to the guinea pig as having a
pure-rod retina (Sjostrand, 1965). There are apparently no
other contemporary studies of the anatomy of guinea pig
photoreceptors. In recent years, however, the presence of
different cone types has been demonstrated through the use of
various functional markers, including cone-specific, antivisual
pigment antibodies. In particular, Szel and Rohlich and their
colleagues have used two monoclonal antibodies that are
visual-pigment specific to label "blue-sensitive" and "green
and red-sensitive cones" in a number of mammalian species. In
an article directed to another topic, Szel and Rohlich note that
these two classes of antibody also label different populations of
receptors in the guinea pig retina, thus providing evidence to
suggest that the guinea pig retina not only contains cones but
also may contain two different types of such receptors (Szel &
Rohlich, 1992).
Electrophysiological studies have not clarified the picture.
In recordings made from fibers of the optic nerve of the guinea
pig, Granit (1947) could find no evidence for a Purkinje shift, a
shift that he easily detected in parallel recordings made from
cats. This result was in accord with his idea that the guinea pig
has a pure-rod eye. However, he did find occasional elements
in the optic nerve of the guinea pig that had very narrow
spectral sensitivity curves. These elements, termed modulators
by Granit, were recorded under conditions of photopic illumination, and their presence led to the speculation that the
guinea pig had some photoreceptors that were functionally
intermediate between typical rods and cones (Granit, 1947). In
a later study, Dodt and Wirth (1953) examined the behavior of
a retinal gross potential, the electroretinogram (ERG), under
conditions of intermittent light stimulation. Variation in stimulation rate and light intensity allowed a determination of the
critical fusion frequency (CFF) of the ERG as a function of
stimulus intensity. The curve so derived shows clear evidence
for two limbs—one limb became asymptotic at stimulus frequencies at 10-25 flashes per second and a second limb showed a
further increase in fusion frequency up to about 45-50 flashes
per second. The appearance of two such limbs in the CFF
curve is classically taken as an indication of visual duplicity
994
GERALD H. JACOBS AND JESS F. DEEGAN II
(Hecht & Schlaer, 1936), and it accordingly suggests the
presence of both rods and cones in the guinea pig. Thiele and
Meissl (1987) have recorded light-evoked responses from
single units in the pineal stalk and pineal body of the guinea
pig. The average spectral sensitivity function from such units
measured during light adaptation had a peak appropriate for
rod input, at about 500 nm, but it also showed a curious
secondary peak at about 450 nm. The latter could be interpreted as reflecting the contributions from receptors other
than rods. That same conclusion may be drawn from an earlier
study involving the measurement of lights reflected from
guinea pig eyes (Weale, 1955). In that case, the appearance of
separable bleaching spectra suggested the presence of multiple
types of photopigment in the guinea pig eye.
There are few behavioral studies of vision in guinea pigs. In
seminatural and natural settings, guinea pigs and their caviid
relatives show crepuscular activity rhythms (King, 1956; Rood,
1972); whereas in the laboratory, they may display essentially
continuous activity under conditions of either constant illumination or constant darkness (Harper, 1976). These observations contain no strong inference about the receptor complement, but they do suggest that the guinea pig eye has a large
dynamic intensity range. That range is typically associated with
the joint presence of rods and cones. There appears to be only
a single behavioral study of vision in the guinea pig that bears
on the issues of concern here. Miles, Ratoosh, and Meyer
(1956) sought to train an operant discrimination where reinforcement and nonreinforcement for lever-press responses
were signaled by the presence of lights that varied in brightness, in color, or in both these dimensions. Although the
guinea pigs rather quickly learned a brightness discrimination,
they failed to show any discrimination between red and green
light or between green and blue light. The authors concluded
that guinea pigs lack color vision.
In light of the inconsistencies of results among these various
studies, we have undertaken an electrophysiological examination of spectral mechanisms in the guinea pig retina and a
parallel study of spectral sensitivity and color discrimination in
the guinea pig. Our investigation was impelled by the lack of
consistent information about basic aspects of vision in an
important mammal and by the belief that there are now better
methods for studying these issues than were available in the
past, and also because recent studies involving phylogenetic
analyses of amino acid sequence data have raised the intriguing possibility that the guinea pig may not be a rodent at all.
Rather, it has been argued that the guinea pig may deserve,
instead, a separate ordinal status (Graur, Hide, & Li, 1991; Li,
Hide, Zharkikh, Ma, & Graur, 1992). If that idea is correct,
then comparisons of the spectral mechanisms of the guinea pig
with those of typical rodents could prove useful with respect to
ideas about the evolution of receptors and photopigments.
Method
Subjects
Adult pignnented guinea pigs (Cavia porcellus) were obtained from a
local supplier. Both males and females were examined. The guinea
pigs were caged separately under standard colony conditions that
included a 14:10-hr Jight-dark cycle. The guinea pigs that served as
subjects in visual discrimination tasks were fed daily following testing
in an amount sufficient to hold them at a constant body weight. The
remaining guinea pigs were given ad-lib access to food.
Electrophysiological Experiments
Apparatus
The electrophysiological experiments involved an analysis of the
flicker ERG in a version we have termed ERG flicker photometry. The
basic features of the apparatus and the general procedures have been
described in earlier publications (Jacobs & Neitz, 1987; Neitz &
Jacobs, 1984). Only brief descriptions are given here.
The stimulus lights were produced with a three-channel, Maxwellianview optical system. Lights from the three channels were optically
combined to illuminate a 57° circular spot on the retina. The test light
of the flicker photometer constituted one channel; it came from a
tungsten halide lamp of a high intensity grating monochromator
(Bausch & Lomb, Rochester, NY; half-energy passband = 10 nm).
The other two channels also originated from tungsten halide lamps.
One of these channels provided the test light of the photometer, and
the other channel was used to produce an adaptation light. The
intensity of the test light was varied by adjusting the position of a
neutral-density wedge having an attenuation range of 3 log units. The
intensities and spectral characteristics of the reference and adaptation
lights were controlled by inserting neutral-density and color filters into
the respective beams. All three lamps were underrun at 11 V from a
regulated DC power supply, and each channel contained a high-speed
electromagnetic shutter (Uniblitz, Vincent Associates, Rochester,
NY). Light measurements were made with a silicon diode photodetector (PIN 10 DL, United Detector Technology, Culver City, CA).
Electrical signals were recorded with a bipolar, contact-lens electrode. A ground electrode was placed against the inside of the guinea
pig's cheek. In the flicker photometry procedure, ERGs were elicited
by temporally modulated, square-wave lights. Stimuli from the test and
reference beams were interleaved such that there was a no-stimulus
interval equal in duration to that of the test and reference lights
between each successive light pulse. The frequency of stimulation was
specified as the total number of light flashes per second. The ERGs
elicited by test and reference lights were passed through a series of
filter stages and electronically compared such that the signals from the
two sources were cancelled when the two lights were equally effective
at generating an ERG (Neitz & Jacobs, 1984).
Procedure
Guinea pigs were anesthetized with an intramuscular injection of a
mixture of ketamine hydrochloride (33.3 mg/kg) and zylazine hydrochloride (6.7 mg/kg). The eye to be tested was dilated by topical
application of atropine sulfate (0.04%) andphenylephredine hydrochloride. The guinea pig was positioned for recording in a modified
stereotaxic instrument. Body temperature was supported with the use
of a heating pad. Except for one experiment explicitly concerned with
scotopic sensitivity, recordings were made in a room illuminated by
overhead fluorescent lights yielding an illuminance of 296 Ix.
To determine an ERG photometric equation, we averaged the
responses to the lights for the last 50 cycles of a total presentation of 70
cycles. These equations were made by iteratively adjusting the neutraldensity wedge in the test light beam until the test and reference lights
were equally effective in producing an electrical response. The
dependent measure was the radiance of the test light required to
achieve this equation. The wedge values at the point of equation were
read to the nearest 0.01 log unit. All photometric equations were made
at least twice during the recording sessions, and the separate values
were subsequently averaged together. Stimulus flicker tate, test
GUINEA PIG SPECTRAL SENSITIVITY AND COLOR VISION
wavelength, characteristics of the reference light, and adaptation state
of the eye were varied. The particular combinations of these features
are specified later in conjunction with the results from each experiment.
Behavioral Experiments
Detailed descriptions of the apparatus and procedure used to
measure visual discriminations have been published (Jacobs 1983,
1984), so only a brief treatment is offered here.
Apparatus
The discrimination was set in the context of a three-alternative,
forced choice. The guinea pig viewed three circular panels that were
transilluminated from an optical system located outside the test
chamber. There were two light sources; one was a tungsten halide
lamp whose output was used to diffusely and equally illuminate each of
the three panels (background lights). The other source was a grating
monochromator (Model H-10, Instruments SA, Edison, NJ) having a
75-W xenon lamp. The output from this lamp was directed via a mirror
system so as to illuminate any one of the three panels; it constituted
the test light. Depending on the experiment, the test light was either
added to the background or replaced it. Both lamps were run from
regulated DC power supplies. Except for measurements of scotopic
spectral sensitivity, the interior of the test chamber was always
ambiently illuminated to 100 Ix.
The guinea pig was trained to discriminate the panel illuminated by
the test light from the other two panels. To indicate its choice, the
guinea pig touched a panel and, if the choice was correct, received a
drop of liquid (Ocean Spray Cranapple Juice Cocktail). The location
of the panel illuminated by the test light was varied randomly over
trials, and the magnitude of the illumination difference between the
positive and the two negative panels was titrated to determine
discrimination thresholds. All aspects of the stimulus presentation,
reinforcement delivery, and response monitoring were done automatically under program control of a computer.
995
corresponding to the upper 99% confidence level was taken as the
threshold value.
The several test conditions under which increment-threshold spectral sensitivity was measured were as follows: (a) Background—
luminance = 0.92 log cd/m2; color temperature = 5350 K; test
light—420-610 nm in steps of 20 nm. (b) Background—luminance =
1.73 log cd/m2; color temperature = 5350 K; test light—420-600 nm in
steps of 10 nm. (c) Background—luminance = 1.73 log cd/m2; color
temperature = 5350 K; test light—420-610 nm in steps of 10 nm. (c)
Background—long wavelength light produced by the use of a high-pass
filter having 50% transmission at 580 nm that yielded a luminance of
1.95 log cd/m2, test light—420-580 nm in steps of 10 nm. A second set
of spectral sensitivity functions was determined with these same
conditions except that the luminance of the background light was
increased to 2.91 log cd/m2. (d) Background—short-wavelength light
produced through the use of a 440 nm interference filter having a
half-energy passband of 10 nm that yielded a luminance of either 1.3
cd/m2 or 0.9 cd/m2; test light—550-630 nm in steps of 10 nm or
570-610 nm in steps of 5 nm. (e) Background—none and with the
chamber darkened; test light—440-600 nm in steps of 10 nm.
Color discrimination. Following the increment-threshold measurements, the 2 guinea pigs were tested to see if they could make color
discriminations. For this purpose, the test light was presented in the
substitution mode, that is, on each trial it replaced one of the three
background lights. The test light was varied in wavelength from 460 nm
to 490 nm in steps of 3 nm. The ordering of wavelengths was random.
At each test wavelength, the luminance of the positive stimulus was
randomly varied in steps of 0.1 log unit to span a total intensity range of
0.7 to 0.9 log units symmetrically centered around the point where the
test light and the background lights had been calculated to be equally
bright for the subject. The procedure used to establish these brightness
equations is described later. The experiment was conducted for two
different conditions of background lighting: (a) luminance = 2.43 log
cd/m2, color temperature = 5350 K; (b) luminance = 2.79 log cd/m2,
color temperature = 5350 K.
Results and Discussion
Procedure
Using conventional procedures, 2 guinea pigs (1 male and 1 female)
were first trained to select the uniquely illuminated panel when there
was a large (> 2 log units) brightness difference between the test and
background lights. After success at that task, they were tested in a
series of experiments that fell into two categories.
Increment thresholds. Increment-threshold spectral sensitivity was
measured under five different test conditions. In each case, monochromatic test lights (half-energy passband = 16 nm) were presented over
a span of intensities in step values of 0.3 log unit. The intensity range
was selected after initial training so that it yielded performance values
running from about 80%-90% correct to 30%-40% correct. This
typically required five or six intensity steps. Each of the intensitywavelength combinations was presented in a three-trial block. The
occurrence of the trial was signaled by a cuing tone; the trial
terminated after the guinea pig responded or after 10 s without a
response. A noncorrection procedure was used. Testing continued in
daily sessions until a total of 100 or more trials had been accumulated
at each wavelength-intensity combination. From these data, psychometric functions were drawn for each test wavelength by plotting mean
percentage correct as a function of test light intensity. These points
were fit to a logistic function having asymptotes of 100% and 33%
correct with the variance and mean as free parameters. The function
that provided the best least-squares fit to each data set was determined
and from that the test stimulus intensity required to yield performance
Electrophysiological Measurements
of Spectral Mechanisms
To establish that ERG flicker photometry yields reliable
data from guinea pig eyes and that the spectra measured with
this technique can provide valid estimates of the photopigments, we first measured spectral sensitivity under scotopic
test conditions for comparison with the earlier measurements
of the rod photopigment in the guinea pig. Guinea pigs were
adapted to the dark for 20 min and then tested in a fully
darkened room. A slow flicker rate (8 Hz) and a dim achromatic reference light (radiance of 0.48 \L W/cm2 at the cornea)
were used to enhance contributions from scotopic mechanisms. ERG flicker-photometric equations were obtained from
each of 3 guinea pigs for test wavelengths taken at 10 nm steps
from 430 nm to 590 nm. The resulting spectral sensitivity
functions are shown in Figure 1. The solid circles represent log
quanta! sensitivity values at the level of the cornea, and the
results for the 3 guinea pigs are arbitrarily spaced on the
sensitivity axis. The sensitivity values for these subjects were
greatly similar in that the range of sensitivity values for the 3
subjects at any of the test wavelengths did not exceed 0.1 log
unit.
996
GERALD H. JACOBS AND JESS F, DEEGAN II
3,5
3.0
en
a
tn
2.5
a
2.0
I
9
1.6
1.0
450
500
550
600
Wavelength (nm)
Figure 1. Scotopic spectral sensitivity functions for guinea pigs. Each
individual symbol shows sensitivity as determined by electroretinogram flicker photometry from a dark-adapted eye. Results are shown
for 3 guinea pigs; the data for each guinea pig are arbitrarily positioned
on the ordinate. The continuous lines represent visual-pigment absorption curves determined as described in the text. The Xmax values and
the goodness of fit (the least mean difference squared between the
theoretical absorption curve and the data points) are, from top to
bottom: 495 nm—1.61 x 10~~3 log unit; 492 nm—4.62 x Ifr 4 log unit;
495 nm—9.38 x Ifr 4 log unit.
To compare these scotopic spectral sensitivity functions with
earlier estimates of the rod photopigment, we best fit each
function with a photopigment absorption curve (continuous
line in Figure 1). The procedure to accomplish the best fit was
to shift a standard retinali photopigment absorption curve
(Ebrey & Honig, 1977) along a log wavenumber axis (Baylor,
Nunn, & Schnapf; 1987) in steps of 1 nm until the best
least-squares fit was obtained between the data array and the
pigment absorption curve. As can be seen in Figure 1, these fits
are quite good for each of the 3 guinea pigs. The best-fitting
curves for the scotopic functions of the 3 guinea pigs had X.max
values of 495 nm, 492 nm, and 495 nm. The average peak value
(494 nm) is in reasonable agreement with the estimate for peak
absorption of guinea pig rhodopsin (497 nm) reported by
Bridges (1959). This correspondence suggests that ERG flicker
photometry can provide sensitive and valid estimates of photopigments of the guinea pig.
Light adaptation and rapid flicker rates are among the tools
traditionally used to diminish rod contributions to the ERG
(Armington, 1974; Birch, 1989). We used both of these in an
attempt to enhance any photopic components in the guinea pig
ERG. As the results of Dodt and Wirth (1953) first suggested,
such components are easily seen in the guinea pig ERG.
Figure 2 shows spectral sensitivity functions obtained from 2
guinea pigs under conditions of light adaptation when the
stimulus rate was increased to 40 Hz, and the reference light
had an intensity of 74 u, W/cm2. Relative to the scotopic
spectral sensitivity functions, peak sensitivity was shifted to
longer wavelengths (520-550 nm), and there was often a region
of secondary elevation in the short wavelengths. Functions
obtained under these test conditions are not accounted for by
contributions from any single photopigment. Rather, the
shape of the functions suggests that multiple spectral mechanisms (perhaps both rods and one or more classes of cone)
underlie measured spectral sensitivity.
The spectral sensitivity functions of Figure 2 imply the
presence in the guinea pig eye of at least one spectral
mechanism that has a peak that is displaced toward the long
wavelengths relative to that of the rod peak. We used three
strategies to obtain an estimate of the spectral sensitivity of
this mechanism. Each involved the use of a stimulus arrangement that would be expected to enhance contributions from
this putative middle-wavelength sensitive (MWS) mechanism
and to minimize contributions from other mechanisms having
higher sensitivity to shorter wavelengths. These strategies
involved using higher flicker rates or long-wavelength reference lights, or combining these features. In addition, we
minimized the effects of contribution from short-wavelength
mechanisms by restricting measurement to longer test wavelengths. The results from these three experiments are summarized as the spectral sensitivity functions of Figure 3. A total of
9 guinea pigs were tested under these three conditions. The
left panel in Figure 3 shows spectral sensitivity functions
obtained using more rapid flicker (62.5 Hz); the reference light
was achromatic (intensity of 74 JA W/cm2). The middle panel
shows functions obtained from 4 guinea pigs for experiments
involving a 50-Hz stimulus and 590-nm reference light (50 u,
W/cm2); the 2 guinea pigs with data shown in the right panel
were tested using 40-Hz flicker in conjunction with a longwavelength reference light (produced by inserting a high-pass
filter having a 50% half-energy cutoff at 580 nm; intensity
measured over the spectral range of 550-650 nm = 3.2 x 10~5
W/cm2).
Under the assumption that the spectral sensitivity recorded
under these conditions might principally reflect contributions
*••
> 1.4
"3>
i
CO
<u
,2
..
1.0
1
«g °-8
O)
5 0.6
0.4
t
450
500
550
600
Wavelength (nm)
Figure 2. Spectral sensitivity measurements obtained by electroretinogram flicker photometry from light-adapted guinea pigs. Results
obtained for 2 guinea pigs are indicated by the separate symbols.
Stimulus rate = 40 Hz.
997
GUINEA PIG SPECTRAL SENSITIVITY AND COLOR VISION
525
3.5
527
|? 3.0
525
35
g 2.5
«
(0
2.0
1.5
1.0
0.5
450 500 550 600 650
450 500 5SO 600 650
450 500 550 600 650
Wavelength (nm)
Figure 3. Spectral sensitivity measurements of the middle-wavelength sensitive mechanism of the guinea
pig made with electroretinogram flicker photometry. Each data set shows the results obtained from 1
guinea pig. Results for individual guinea pigs have been arbitrarily positioned on the ordinate. Three
different combinations of stimulus rates and reference lights were used: left panel (62.5 Hz, achromatic
reference light); center panel (50 Hz, long-wavelength reference light); right panel (40 Hz, longwavelength reference light). Each data set has been best fit with a visual-pigment absorption curve. The
from a single photopigment, each of the spectral sensitivity
functions was best fit with a photopigment absorption curve
using the procedures described previously. These best-fit
curves are shown in Figure 3 with the appropriate X.max values
indicated. Over the restricted spectral range that was examined, these photopigment curves reasonably fit the data arrays.
The spectral peaks varied modestly for the 9 subjects tested
with three different procedures (mean Xmax = 527.5 nm;
SD = 2.3 nm). Any residual contribution from shorter spectral
mechanisms would have the net effect of causing the computed
peaks of the best-fit functions to be shifted toward shorter
wavelengths. It is thus reasonable to assume that these
experiments indicate that the guinea pig retina contains a type
of cone whose spectral peak is not shorter than about 528 nm.
Given the good fit between the pigment absorption functions
and the long wavelength slope of these spectral sensitivity
functions, the actual peak is probably also not much longer
than that value.
The spectral sensitivity functions of Figure 2 also suggest the
presence in the guinea pig eye of a spectral mechanism having
peak sensitivity that is shorter than that of the rods. A series of
experiments were conducted to see if a more persuasive
indication of such a mechanism might be obtained. The
principal tools used to accomplish this were concurrent longwavelength adaptation and short-wavelength reference lights.
Figure 4 shows the average spectral sensitivity function obtained from 3 guinea pigs for 25-Hz flicker when the eye was
concurrently adapted to an orange adaptation light (produced
as in the right panel of Figure 3; radiance = 1.6 x 10~3
W/cm2). Under such conditions of adaptation, there is a
relative enhancement of sensitivity to the short test wavelengths (compare Figure 4 and Figure 2). The differential
distortion of the shape of the spectral sensitivity function with
1.2
1.0
o.s
ja
0.6
2
01
° 0.4
0.2
400
450
500
550
Wavelength (nm)
Figure 4. Electroretinogram spectral sensitivity for guinea pigs obtained with 25-Hz flicker and concurrent long-wavelength adaptation.
The solid symbols represent mean values obtained from 3 guinea pigs
that were comparably tested. The vertical bar to the lower right shows
the average total range of sensitivity for the 3 subjects.
998
GERALD H. JACOBS AND JESS F. DEEGAN I!
suggests that one mechanism has peak sensitivity of about
420-430 nm and the other has peak sensitivity in the vicinity of
530 nm.
3.0
Behavioral Measurements of Spectral
Sensitivity and Color Vision
5 2.5
!
CO
a,
1
"S
CC
at
o
2.0
1.5
1.0
400
450
500
550
Wavelength (nm)
Figure 5. Results from electroretinogram experiments intended to
define the spectral sensitivity of the short-wavelength sensitive (SWS)
mechanism of the guinea pig. Each data set shows the results obtained
from 1 guinea pig with flicker photometry under conditions intended
to advantage contributions from the SWS mechanism. These conditions included a short-wavelength reference light and concurrent,
intense long-wavelength adaptation. The results for the 4 guinea pigs
are arbitrarily positioned along the ordinate. The continuous curves
represent the best-fitting summative combination of two pigment
absorption functions. One of these functions was set to a Xmax value of
529 nm, and the spectral position of the other function was sought by a
best-fit procedure. The \max and proportion values for the SWS
mechanism determined in this fashion were (from top to bottom): 426
nm (81%); 425 nm (82%); 432 nm (81%); and 426 nm (83%).
chromatic adaptation indicates that multiple spectral mechanisms exist in the guinea pig retina. More importantly, these
measurements suggest that there is indeed a short-wavelength
sensitive (SWS) mechanism.
For 4 additional guinea pigs, we recorded spectral sensitivity
functions at 25 Hz under yet more stringent conditions of
adaptation (reference light of 440 nm at an intensity of 50 u,
W/cm2 with concurrent orange light adaptation at a radiance
of 3,2 x 10~3 W/cm2). The spectral sensitivity functions are
given in Figure 5. Under these conditions, sensitivity to test
lights of greater than about 500 nm was reduced so as to make
these points unmeasurable. What remained were spectral
sensitivity functions that show peak sensitivity in the vicinity of
420-430 nm. The functions for the 3 guinea pigs are very
similar. The derivation and meaning of the fitted curve is
described below.
In summary, measurements made with ERG flicker photometry suggest the presence of three spectral mechanisms in the
guinea pig retina. One is a scotopic mechanism having a peak
of about 494 nm. In addition, there are (at least) two spectral
mechanisms operative under photopic test conditions. It proved
comparatively difficult to completely isolate these mechanisms,
but utilization of a range of different adaptation conditions
Guinea pigs proved to be apt subjects for these tests of visual
discrimination. Once trained, 2 guinea pigs produced highly
reliable data throughout the 15-month period that was required to complete these tests. Results from our first attempt
to measure a photopic spectral sensitivity function are shown
in Figure 6. The solid symbols in Figure 6 represent the mean
threshold values obtained from the 2 subjects. There are two
notable features of these spectral sensitivity functions: (a)
They have two locations of peak sensitivity, one in the short
wavelengths in the vicinity of 420 nm and one in the middle
wavelengths at 520-540 nm; (b) there is a dramatic decrease in
sensitivity at a location intermediate to these peaks centered at
about 480 nm.
The spectral sensitivity function of Figure 6 has several
features characteristic of increment-threshold measurements
obtained under photopic test conditions from many other
mammalian subjects (Jacobs, 1993). In particular, the shape of
the function suggests the presence of inhibitory interaction
between two spectral mechanisms. Such inhibitory interaction
can frequently be enhanced by increasing the intensity of the
achromatic background light (Kalloniatis & Ffarwerth, 1990;
King-Smith & Garden, 1976; Sperling & Harwerth, 1971). That
was the purpose of the second experiment in which the
1.5
I
5
°3>
c
«
(A
1-0
§>
0.5 450
500
550
600
Wavelength (nm)
Figure 6. Increment-threshold spectral sensitivity functions for the
guinea pig. The points are mean values obtained from 2 guinea pigs
with an achromatic background light (0.92 log cd/m2). The average
range of sensitivity values for all test wavelengths was 0.18 log unit. The
largest individual variation in threshold occurred at 420 nm; the
magnitude of this variation is indicated by the vertical line drawn
through the data point. The curve represents the best-fit, subtractive
combination of two pigment absorption curves having the following
\max values and relative proportions: 529 nm (44%) and 438 nm (56%).
999
GUINEA PIG SPECTRAL SENSITIVITY AND COLOR VISION
background light was substantially increased (by ca. 0.8 log
unit) in intensity. These spectral sensitivity functions are
plotted separately for the 2 subjects in Figure 7. As predicted,
these functions show even more clearly the contributions to
behavioral discrimination from two spectral mechanisms.
Sampled at interval steps of 10 nm, the locations of peak
sensitivity for the 2 subjects were similar—430 nm and 440 nm
for the short mechanism and 530 nm for the other mechanism.
The intermediate location of lowered sensitivity is even more
apparent for this higher level of light adaptation and is again
centered at about 480 nm. Although the shapes of the curves
were similar for the 2 subjects, the curves differed in the
relative heights of the two spectral peaks with one having
higher relative sensitivity to the short wavelengths.
Full spectral sensitivity functions were measured for 1
guinea pig when the achromatic background light was replaced
with an orange light. The intent of this experiment was to
determine whether the two spectral mechanisms contributing
to increment-threshold detection responded differentially to
chromatic adaptation and to try to get a better estimate of the
spectral properties of the SWS mechanism. The two spectral
sensitivity functions that were obtained are shown in Figure 8.
Note that the long-wavelength adaptation light caused a much
greater loss of sensitivity to the long wavelengths than it did to
the short wavelengths (compare Figure 7 and Figure 8).
Increasing the intensity of this background light (bottom
function of Figure 8) had a further differential adaptation
effect in that the additional loss of sensitivity in the long
wavelengths was much greater than it was in the short test
wavelengths. The more intense orange light caused an additional threshold elevation of about 0.8 log unit at 560 nm,
3.5
3.0
|
2.5
a
to
a>
2.0
1
1.5
CD
0.5
450
500
550
600
650
Wavelength (nm)
Figure 7. Increment-threshold spectral sensitivity functions for 2
guinea pigs. The data were obtained in the presence of an achromatic
background light (1.73 log cd/m2). The results for the 2 guinea pigs are
arbitrarily positioned on the ordinate. The curves are best-fit, subtractive combinations of two photopigment absorption curves. The Xmax
values and relative proportions of these best-fit curves for the guinea
pig at the top are 529 nm (21%) and 425 nm (79%); comparable values
for the guinea pig at the bottom are 529 nm (45%) and 439 nm (55%).
3.0
f2.5
•5
I 2'°
o>
"I
1-5
1
a 1.0
0.5
450
500
550
600
Wavelength (nm)
Figure 8. Increment-threshold spectral sensitivity functions obtained
from 1 guinea pig under two conditions of chromatic adaptation. In
each case, the background was an orange light. For the results shown
at the bottom, the background light was about 0.3 log unit more
intense than the background used for the data shown at the top. The
functions were best fit by additive combinations of two photopigment
absorption curves; these were 529 nm (18%) and 427 nm (82%) for the
top data set, and 529 nm (11%) and 425 nm (89%) for the bottom data
set.
whereas the same light elevated threshold for a 430 nm light by
only an additional 0.2 log unit. The experiment makes clear
that the two components of the increment-threshold spectral
sensitivity function of the guinea pig may be differentially
adapted. We return below to the question of the spectral
specification of the two underlying mechanisms.
Two additional experiments were conducted to better estimate the spectral properties of the MWS mechanism. In
essence, the idea was to carry out behavioral measurements
analogous to the electrophysiological measurements that were
summarized in Figure 3. Spectral sensitivity measurements
were made when the background light was of short wavelength
(440 nm). Our intent was to diminish as much as possible any
contribution from the SWS mechanism or the rods. Data were
obtained at two different intensity levels; the test wavelengths
were chosen to fall on the long wavelength slope of the MWS
component. Results from the dimmer of these wavelengths are
shown in the left panel of Figure 9; the results on the right were
obtained after the background light had been slightly increased in intensity. In each case, the sensitivity data have been
best fit to photopigment absorption curves using the procedures described previously. For the dimmer adaptation, the
Xma>j of the best-fit function was 525 nm; with the brighter
adaptation and more detailed sampling (in steps of 5 nm) of
this long-wavelength slope, the best-fit curve has KmaK of
529 nm.
A final spectral sensitivity measurement was made with
background and test-chamber light extinguished. The single
subject was adapted to the dark for a 20-min period before the
initiation of each test session. The spectral sensitivity function
obtained is shown in Figure 10. The threshold measurements
1000
GERALD H. JACOBS AND JESS F. DEEGAN II
1.5
to
» 1.0
CO
0)
V3
.2
0)
oc
O>
0.5
o
0.0 -
I
I
500
550
I
600
650
500
550
600
650
Wavelength (nm)
Figure 9. Attempts to define the spectral positioning of the middle-wavelength sensitive (MWS)
photopigment of the guinea pig from increment-threshold spectral sensitivity functions. Both panels show
sensitivity measurements obtained with stimulus conditions intended to advantage contributions from the
MWS mechanism. The background light in each case was 440 nm; the intensity of this light was higher for
the data presented in the right panel (see text). The absorption curves for visual pigments drawn through
the data sets was obtained by a best-fit procedure. The Xmax values for the curves are 525 nm (left) and 529
nm (right).
were best-fitted in the usual manner. The function that best fit
this data array had a Xmax value of 498 nm (solid line in Figure 10).
The photopic spectral sensitivity functions for guinea pigs
8.0
>*••
7.5
I
4**
1 7.0
§>
o
6.5
6.0
450
500
550
600
Wavelength (nm)
Figure 10. Scotopic spectral sensitivity function obtained in a behavioral discrimination test from a dark-adapted guinea pig. The curve
represents the visual pigment absorption function obtained by a
best-fit procedure. The Xmax of this curve is 498 nm (fitting error = 8.86 x 10~3 log unit).
(see Figures 6-9) strongly suggest the presence of two cone
photopigments, and they accordingly predict the species to
have dichromatic color vision. A color discrimination experiment was conducted to evaluate that possibility. In this
experiment, the test lights were monochromatic and variable in
wavelength over the spectral range from 460 nm to 490 nm.
The negative lights were achromatic. The stimuli were presented in the substitution mode, that is, on each trial the test
light replaced one of the three background lights. Successful
discrimination of a monochromatic light from an achromatic
light under conditions where there is no consistent brightness
difference between the two is a color discrimination. The
major obstacle to be overcome in tests of this sort is to ensure
that subjects cannot use consistent brightness cues, that they
are in fact forced to make a color discrimination. To equate the
monochromatic and achromatic lights so that they were
equally bright, we used a procedure used earlier (Jacobs,
1984). The procedure exploits the fact that in all of the
previous experiments the guinea pigs had been trained to pick
the brightest of the three lights. To establish brightness
matches, we arranged the situation such that the monochromatic test light was presented as a substitute for one of the
three background lights. As long as this light was perceived as
brighter than the achromatic lights, the guinea pig selected it
consistently, just as it had in the increment-threshold experiments. The test light was then made progressively less intense
(in steps of 0.2 log unit). At some point, the two negative lights
become relatively brighter than the test light and the subjects'
behavior reflected this by shifting from consistent preference
GUINEA PIG SPECTRAL SENSITIVITY AND COLOR VISION
of the test light to consistent avoidance. The intensity of the
test light leading to this reversal was taken as indicating the
location of the brightness match. Such matches were made for
test wavelengths drawn from each end of the spectral range.
Once these brightness matches were determined, the subject
was trained in the color discrimination.
The left panel of Figure 11 shows asymptotic discrimination
performance obtained across the spectral range tested for each
subject. The plotted points are average discrimination recorded over the last 50 test trials for those test light intensities
at the point of computed equal brightness. These functions are
not materially changed either by averaging the results obtained
at the equation value and for values ±0.1 log unit from this
value or by simply plotting the poorest performance recorded
at each test wavelength. These facts suggest that the discrimination results do reflect those from a test where there were no
consistent brightness cues. The 2 guinea pigs performed
similarly in this test. Both guinea pigs discriminated the
spectral lights almost perfectly at either end of the test range;
both guinea pigs also showed a loss of consistent discrimination in the vicinity of 480 nm. The horizontal dashed line of
Figure 11 indicates the 95% confidence level. As judged by this
performance criterion, the spectral range over which the
guinea pigs failed to make a color discrimination extended
from 470 nm to 484 nm for 1 subject, and from 477 nm to 482
nm for the other subject. The midpoints of this range are
similar for the 2 subjects—477 nm and 479.5 nm, respectively.
The entire color discrimination experiment was run a second
time after the achromatic lights had been increased in intensity
by about 0.3 log unit. These results are shown in the right panel
of Figure 11. The conventions were the same, and so were the
results. In this case, discrimination failed for 1 subject for a
spectral range of 471-482 nm (midpoint 476.5 nm) and failed
for the other subject in a narrower range—from 477 nm to 483
nm (midpoint of 480 nm). These regions of discrimination
failure define the presence of a spectral neutral point and
verify the prediction that guinea pigs have dichromatic color
vision.
General Discussion
Guinea Pig Photopigments and Spectral Mechanisms
The results reported here suggest that the eye of the guinea
pig contains three classes of photopigment—one is a rod
pigment and the other two are cone photopigments. Both
electrophysiological and behavioral measurements allow estimates of the absorption spectra for these pigments. We noted
that the spectral positioning of the MWS photopigment can be
most reliably characterized by fits of absorption spectra to the
long-wavelength portion of the curve. The ERG measurements (see Figure 3) and behavioral measurements (see
Figure 9) are in good agreement in this regard. The former
yielded an average Xmax value of about 528 nm, whereas the
behavioral measurement, which was most likely to have indexed activity of the MWS pigment alone (see Figure 9, right
curve), gave a corresponding value of 529 nm. Because any
residual contributions from other pigments having shorter Xmax
values would tend to shift the peak toward the short wavelengths, the longer of these two wavelengths is assumed as the
best current estimate of the spectral positioning of the guinea
pig MWS cone pigment.
By virtue of the considerable spectral overlap of the MWS
and SWS photopigments in the short wavelengths, the spectral
100190
o
E
O
u
1001
80
70
60
(B 50
a»
40
30
20
460 470 480 490 500 510 520
460 470 480 490
Wavelength (nm)
Figure I I . Color discrimination results for 2 guinea pigs. Each data point indicates the asymptotic levels
of performance achieved in a test involving the discrimination of various monochromatic lights (abscissa
values) from equally bright achromatic lights. The two panels show results from two experiments that were
identical except for the achromatic lights, which were about 0.3 log more intense for the experiment shown
to the right. The horizontal dashed lines indicate the 95% confidence level.
1002
GERALD H. JACOBS AND JESS F. DEEGAN II
positioning of the SWS photopigment may be best estimated
from those experiments in which contributions from MWS
receptors were reduced through the use of chromatic adaptation. These are the ERG measurements shown in Figure 5 and
the behavioral measurements in Figure 8. In these experiments, even with concurrent long-wavelength adaptation, there
likely remains some residual contributions from the MWS
pigment. Accordingly, to estimate the spectral positioning of
the SWS pigment we assumed that the ERG and behavioral
results of Figures 5 and 8 represent summative contributions
from MWS and SWS mechanisms. To determine the spectral
positioning of the SWS mechanism, the array of sensitivity
values were best fit using summative combinations of a
photopigment having \max value of 529 nm and an SWS
photopigment. The spectral position of the putative SWS
photopigment was varied in steps of 1 nm (over the spectral
range of 420-440 nm) and in relative proportions of 1% until
the best-fitting combination of the two pigments was obtained.
These best-fit curves are the continuous lines drawn through
the data in Figures 5 and 8. This strategy accounts quite well
for the shape of the spectral sensitivity functions. The estimated peak values for the SWS mechanism show relatively
small variability. The 4 subjects whose ERG results are shown
in Figure 5 have an average SWS Xmax value of 427.3 nm
(SD = 3.1); the subject of the behavioral experiments shown in
Figure 8 yielded corresponding values of 425 nm and 427 nm
for the two experiments.
The increment-threshold measurements of Figures 6 and 7
also show clear evidence of inputs from both cone types. The
combination of these inputs in this situation, as for other
increment-threshold measurements on achromatic backgrounds (e.g., Kalloniatis & Harwerth, 1990; Sperling &
Harwerth, 1971), involves inhibitory interaction between the
two mechanisms. These data were best-fitted by using a
strategy earlier used to account for such functions obtained
from other dichromatic subjects (Jacobs, 1990; Neitz, Geist, &
Jacobs, 1989), that is, by searching for the best subtractive
combination of a 529-nm MWS pigment and an SWS photopigment. The best-fit functions so obtained are given by the curves
of Figures 6 and 7. This strategy provides a reasonable account
of the discrimination data. Taken together, the summative and
subtractive fits to the electrophysiological and behavioral data
yielded a total of nine individual estimates of the spectral
positioning of the SWS cone pigment of the guinea pig. The
range of such estimates was 13 nm; the mean estimate was 429
nm.
The behavioral and electrophysiological measurements of
scotopic spectral sensitivity (see Figures 1 and 10) are in
reasonable agreement with each other and with the earlier
spectrophotometric measurements of rod pigment in the
guinea pig (Bridges, 1959). All these measurements suggest
that the rods of the guinea pig contain a pigment whose peak
sensitivity is in the range from about 494 nm to 498 nm.
Two other issues of concern in deriving estimates of photopigment spectra from spectral sensitivity measurements deserve
mention. One is that in producing these estimates no corrections have been made for possible preretinal filtering of the
spectral input. The principal source of such filtering in the eyes
of mammals is lens pigmentation (Muntz, 1972). Cooper and
Robson (1969) measured density spectra for the intact lens of
the guinea pig, and their measurements show little spectrally
selective light losses in the guinea pig lens over the portion of
the spectrum examined in the present experiments. Accordingly, we concluded that no corrections need be made for
preretinal filtering. A second source of general concern in
comparing pigment absorption spectra and spectral sensitivity
measurements are possible effects of pigment self-screening. It
has long been recognized that the shape of the absorption
spectrum of a photopigment depends on pigment concentration (Dartnall, 1957) and that comparisons of action spectra
(such as the spectral sensitivity curves of the present experiment) with standard photopigment absorption curves may be
inaccurate if there is substantial pigment density. There are no
direct measurements, nor indeed any other information, to
allow an estimate of the optical density of guinea pig photoreceptors, but several aspects of the results suggest that not
much improvement on the spectral estimates could be obtained by consideration of optical density. In the case of cones,
this may be true because all of the ERG measurements and
most of the behavioral measurements were obtained from
highly light-adapted eyes. This would tend to greatly reduce
pigment density and a priori would seem to obviate the need
for concern about the effects of this variable. On the other
hand, one might expect that the scotopic measurements would
reflect the operation of rods whose pigment density was near
maximal. The fact that the photopigment absorption curves
(see Figures 1 and 10) provide a good account of the shape of
the spectral sensitivity curves suggests that any effects of
self-screening must be small. Why this should be so is unclear,
but it may be noteworthy that similar uncertainty surrounds
comparisons of human rod vision and rhodopsin (Alpern,
Fulton, & Baker, 1987; Bowmaker & Dartnall, 1980).
We noted previously that the earlier studies of the spectral
mechanisms of the guinea pig were inconsistent in outcome.
However, one can perhaps find in them previews of the present
results. For instance, the narrow-band "modulator" elements
detected in early studies of the discharges of optic-nerve fibers
by Granit (1947) have peaks of about 450 nm and 530 nm.
Spectral peaks of about 460 nm and 530 nm also appear in
bleaching spectra obtained from retinal densitometric measurements (Weale, 1955). The spectra determined from visually
evoked responses of the guinea pig pineal gland show a
secondary peak at about 450 nm (Thiele & Meissl, 1987). In
each of these cases, the reported spectral peak values are close
enough to the current measurements of the guinea pig cone
pigments to suggest these earlier results might have reflected
signals of cone origin in the guinea pig visual system.
Color Vision
Given the historical arguments about whether the retina of
the guinea pig contained any cones at all and the emphatic
denial of color vision in the guinea pig in the only study
directed to that end (Miles et al., 1956), it was somewhat
surprising to find that cone signals can be easily detected both
electrophysiologically and behaviorally and that color vision
can be readily demonstrated. Like many other mammals, the
guinea pig is a dichromat. That diagnosis rests on the fact that
GUINEA PIG SPECTRAL SENSITIVITY AND COLOR VISION
both subjects were able to discriminate monochromatic lights
from equally bright achromatic lights except when the monochromatic light was set to a narrow wavelength range (see
Figure 11). The wavelength of the monochromatic light that
matched in appearance an achromatic light for the guinea pig
was located in the vicinity of 480 nm. This location, the neutral
point, represents a color match the details of which depend on
the spectral positioning of the two cone pigments, as well as
the effective spectral characteristics of the achromatic light.
For dichromatic observers, the neutral point effectively splits
the spectrum such that subjects can make color discriminations
between lights whose spectral energy distributions fall principally to either side of the neutral point. Dichromatic observers
also can discriminate most stimuli having narrow-band spectral
energy from stimuli having a continuous spectrum (see Figure
11). In the previous experiment on guinea pig color vision, the
discriminations tested were between a red and a green light
and between the same green light and a blue light (Miles et al.,
1956). These stimuli were produced by passing light through
broad-band filters. It seems unlikely that the first discrimination would have been possible for the dichromatic guinea pig,
because the effective energies of both lights were probably
restricted to the same side of the neutral point. On other hand,
one might expect that the guinea pigs would have been able to
discriminate a blue light from a green light. We can only
speculate why they did not; but whatever the reason, the
conclusion of the earlier study seems to be in error. Guinea
pigs do have color vision. Furthermore, the ease of demonstrating color vision in a laboratory test suggests that it may provide
a useful additional source of information about the environment of the guinea pig.
Phytogeny of Photopigments
There is only limited information about the distribution of
cone photopigments among the mammals. Even so, it may be
useful to briefly consider the cone pigment complement of the
guinea pig from a comparative perspective. What has so far
been learned about the cone pigments of rodents has been
recently summarized (Jacobs, 1993). In brief, several different
cone pigment arrangements have been documented for rodents. Species from the family Sciuridae (the squirrels) typically have two classes of cone pigment. All squirrels (with the
possible exception of the nocturnal flying squirrel) have an
SWS cone pigment with a peak somewhere in the range from
about 435 nm to 445 nm. In addition, there is a class of MWS
cone whose peak location varies—among the ground-dwelling
squirrels this pigment has a peak of about 520 nm, whereas for
the arboreal squirrels the corresponding peak is about 543 nm.
Cone pigments have also been measured for a number of
species from the family Muridae (mice, rats, etc.). Many of
these species also have two classes of cone photopigment.
However, one of these pigments has a very short Xmax value—in
the ultraviolet at about 360 nm. The other (MWS) pigment
varies among species with peak values that extend from about
493 nm to about 512 nm. In addition, there is evidence to
suggest that some murid rodents may have only a single cone
pigment with peak in the vicinity of about 500 nm. Beyond
these two large families, few other rodents have been exam-
1003
ined. The rodents that have been examined seem to fall into
one or another of the patterns just described (Jacobs, 1993). In
comparison to these rodents, the guinea pig cone pigments
thus appear unique. Like the sciurids (but unlike the murids),
the guinea pig has an SWS cone of about 430 nm. However, the
spectral location of the MWS pigment of the guinea pig does
not match the locations so far identified for any other rodent.
In fact, a cone pigment with a peak of 530 nm has not been
claimed for any mammals other than Old World primates. In
Old World primates, a 530-nm cone pigment is common and
maybe the rule (Bowmaker, 1991; Jacobs, 1993).
The factors that determine the details of photopigment
complement result from events of evolutionary history that can
at present only be dimly discerned (Crescitelli, 1990; Goldsmith, 1990; Jacobs, 1993). It was noted earlier that recent
molecular analyses suggest the guinea pig may deserve an
ordinal status separate from that of Rodentia (Graur et al.,
1991; Li et al., 1992). This new order may have arisen from one
of the earliest divergence events in the history of eutherian
mammals—at a point before the divergence of carnivores,
artiodactyls, lagomorphs, chiropterans, and primates (Li et al.,
1992). It is uncertain what other species ought to be included
with guinea pigs and other caviomorphs in this new order, but
Li et al. (1992) suggest that the hystricomorphs (porcupinelike rodents) may represent reasonable candidates. At present,
nothing appears to be known about the cone pigments in any of
these species, but the uniqueness of the guinea pig cone
complement strongly recommends the examination of the
retinas of other species in this new grouping.
References
Alpern, M., Fulton, B., & Baker, B. N. (1987). "Self-screening" of
rhodopsin in rod outer segments. Vision Research, 27, 1459-1470.
Armington, J. C. (1974). The electroretinogram. San Diego, CA:
Academic Press.
Baylor, D. A., Nunn, B. J., & Schnapf, J. D. (1987). Spectral sensitivity
of cones of the monkey Macaco fascicularis. Journal of Physiology,
390, 145-160.
Birch, D. G. (1989). Clinical electroretinography. Ophthalmology
Clinics of North America, 2, 469-497.
Bowmaker, J. K. (1991). Visual pigments, oil droplets and photoreceptors. In P. Gouras (Ed.), The perception of colour (pp. 108-127). Boca
Raton, FL: CRC Press.
Bowmaker, J. K., & Dartnall, H. J. A. (1980). Visual pigments of rods
and cones in a human retina. Journal of Physiology, 298, 501-511.
Bradt, H. (1988). Tasty Andean tidbits. Americas, 40, 64-65.
Bridges, C. D. B. (1959). The visual pigments of some common
laboratory mammals. Nature, 184, 1727-1728.
Cooper, G. F., & Robson, J. G. (1969). The yellow colour of the lens of
the grey squirrel (Sciurus carolinensis leucotis). Journal of Physiology,
203, 403^110.
Crescitelli, F. (1990). The natural history of visual pigments. Progress in
Retinal Research, 11, 1-32.
Dartnall, H. J. A. (1957). The visual pigments. London: Metheun.
Dodt, E., & Wirth, A. (1953). Differentiation between rods and cones
by flicker eiectroretinography in pigeon and guinea pig. Acta
Physiologica Scandinavica, 30, 80-89.
Ebrey, T. G., & Honig, B. (1977). New wavelength-dependent visual
pigment nomograms. Vision Research, 27, 147-151.
Goldsmith, T. H. (1990). Optimization, constraint, and history in the
evolution of eyes. Quarterly Review of Biology, 65, 281-322.
1004
GERALD H. JACOBS AND JESS F. DEEGAN II
Granit, R. (1944). Stimulus intensity in relation to excitation and preand post-excitatory inhibition in isolated elements of mammalian
retinae. Journal of Physiology, 103, 103-118.
Granit, R. (1947). Sensory mechanisms of the retina. London: Oxford
University Press.
Graur, D., Hide, W. A., & Li, W.-H. (1991). Is the guinea pig a rodent?
Nature, 351, 649-652.
Harper, L. B. (1976). Behavior. In J. E. Wagner & P. J. Manning
(Eds.), The biology of the guinea pig (pp. 31-51). San Diego, CA:
Academic Press.
Hecht, S., & Shlaer, S. (1936). Intermittent stimulation by light. V. The
relation between intensity and critical frequency for different parts
of the spectrum. Journal of General Physiology, 19, 965-979.
Jacobs, G. H. (1983). Within-species variations in visual capacity
among squirrel monkeys: Sensitivity differences. Vision Research, 23,
239-248.
Jacobs, G. H. (1984). Within-species variations in visual capacity
among squirrel monkeys: Color vision. Vision Research, 24, 12671277.
Jacobs, G. H. (1990). Duplicity theory and ground squirrels: Linkages
between photoreceptors and visual function. Visual Neuroscience, 5,
311-318.
Jacobs, G. H. (1993). The distribution and nature of colour vision
among the mammals. Biological Reviews, 68, 413-471.
Jacobs, G. H., & Neitz, J. (1987). Inheritance of color vision in a New
World monkey (Saimiri sciureus). Proceedings of the National Academy of Sciences, USA, 84, 2545-2549.
Kalloniatis, M., & Harwerth, R. S. (1990). Spectral sensitivity and
adaptation characteristics of cone mechanisms under white-light
adaptation. Journal of the Optical Society of America A, 7, 1912-1928.
King, J. A. (1956). Social relations of the guinea pig under seminatural conditions. Ecology, 37, 221-228.
King-Smith, P. E., & Garden, D. (1976). Luminance and opponentcolor contributions to visual detection and to temporal and spatial
integration. Journal of the Optical Society of America, 66, 709-717.
Kolmer, W. (1936). Vergleichende Anatomie der Netzhaut [Comparative anatomy of the retina]. In W. von Mallendorf (Ed.), Handbuch
der Mikroskopishcen Anatomie des Menschen (Vol. 3, Pt. 2, pp.
321-326 and 420-421). Berlin: Juluis Springer.
Li, W.-H., Hide, W. A., Zharkikh, A., Ma, D.-P., & Graur, D. (1992).
The molecular taxonomy and evolution of the guinea pig. Journal of
Heredity, 83, 174-181.
Miles, R. C., Ratoosh, P., & Meyer, D. R. (1956). Absence of color
vision in the guinea pig. Journal of Neurophysiology, 19, 254-258.
Muntz, W. R. A. (1972). Inert absorbing and reflecting pigments. In H.
J. A. Dartnall (Ed.), Handbook of sensory physiology: Photochemistry
of vision (Vol. 7, Pt. 1, pp. 529-565). Berlin: Springer-Verlag.
Neitz, J., Geist, T., & Jacobs, G. H. (1989). Color vision in the dog.
Visual Neuroscience, 3, 119-125.
Neitz, J., & Jacobs, G. H. (1984). Electroretinogram measurements of
cone spectral sensitivity in dichromatic monkeys. Journal of the
Optical Society of America A, 1, 1175-1180.
O'Day, K. (1947). Visual cells of the guinea pig. Nature, 160, 648.
Rood, J. P. (1972), Ecological and behavioral comparisons of three
genera of Argentine cmies. Animal Behaviour Monographs, 5, 1-83.
Sjostrand, F. S. (1965). The synaptology of the retina. In A. V. S. De
Reuck & J. Knight (Eds.), Colour vision: Physiology and experimental
psychology (pp. 110-151). Boston: Little, Brown.
Sperling, H. G., & Harwerth, R. S. (1971). Red-green cone interactions in increment-threshold spectral sensitivity of primates. Science,
172, 180-184.
Szel, A., & Rohlich, P. (1992). Two cone types of rat retina detected by
anti-visual pigment antibodies. Experimental Eye Research, 55, 4752.
Thiele, G., & Meissl, H. (1987). Action spectra of the lateral eyes
recorded from mammalian pineal glands. Brain Research, 424,
10-16.
Weale, R. A. (1955). Bleaching experiments on eyes of living guinea
pigs. Journal of Physiology, 127, 572-586.
Received January 19,1994
Revision received April 12,1994
Accepted April 12,1994 •