Behavioral measures and theoretical analysis of spectral sensitivity and spectral saturation in the goldfish,Carassius auratus

BEHAVIORAL MEASURES AND THEORETICAL ANALYSIS OF SPECTRAL SENSITIVITY AND SPECTRAL SATURATION IN THE GOLDFISH, CARASSIUS AURATUS’ DEAN

YAGER

Psychology Department, Brown University, Providence, Rhode Istand, U.S.A. (Received I2 December 1946) INTRODUCTION THE PRESENT experiments were designed to determine the spectral brightness sensitivity of

the goldfish, measured behaviorally, and to obtain a measure of the saturation of spectral stimuli, under carefully specified conditions of adaptation and viewing. The study of fish visual psychophysical functions is important theoretically because a great deal of information about the fish visual system has been obtained by physiological methods. This information has come from every level of the visual system: microspectrophometry on individual cones; recordings of graded potentials from the amacrine and bipolar cell regions; and recordings of individu~ ganglion cell action potentials from the retina, optic nerve, and optic tectum. However, very little quantitative info~ation has been obtained about the visual discriminating capacities of the same species on which the physiological experiments have been performed. This is especially true about discriminations which depend on differences in spectral composition among stimuli. Because of a lack of behavioral information, the precise interrelations between the physiology and psychophysics of color vision are unknown. PREVIOUS

EXPERIMENTAL

WORK

ON FISH

BEHAVIORAL

There has long been a considerable amount of interest in fish color vision. WARNER (I 93 1) reviews most of the types of experiments done up to that date, and a chapter on fish vision appears in Walls’ 1942 book. (Reprinted in 1963.) The available material has been reviewed more recently by BRETT (1957), AUTRUM (1960), VEAUD (1960), and by BERNSTEIN (1961). The most extensive survey of fish vision is in a book by KONRAD HERTER, Die Fischdressuren und ihre sinnesphysiologischen Grundlagen (1953). Because these excellent literature reviews are readily available, an exhaustive survey of experimental work on fish vision will not be presented here; only a few of the more significant later experiments will be described. There have been several behavioral studies on the goldfish (Carassius aurutus), the species used in the present experiment. Behavioral work carried out so far has done little more than demonstrate that goldfish are not monochromats. There have been no systematic investigations of their color discriminating capacities. MCCLEARY and BERNSTEIN (I 959) 1 Research on which this paper is based was supported by grant +GB 618 from the National Science Foundation. 707

708

DEANYAGER

used a conditioned autonomic response to measure discrimination. Fish were trained to discriminate between a red and a green stimulus that had been selected to be of equal brightness to humans. These animals were then tested for generalization with bright and dim red and bright and dim green stimuli. (Previous brightness discrimination studies had shown that in the range of luminances used in this study, the stimuli overlapped in brightness for the fish.) Under these circumstances, the goldfish generalized perfectly according to hue. YAGER (1962), using an operant conditioning technique, demonstrated that goldfish could discriminate lights whose spectral peaks were at 553 and 610 nm. Brightness cues were controlled by a procedure similar to that used in the McCleary and Bernstein experiment. MUN~Z and CRONLY-DILLON(1966) showed that goldfish could discriminate between pairs of panels painted with different colors, when brightness cues were eliminated. The same investigators (CRONLY-DILLONand MUNTZ, 1965) have also obtained a spectral sensitivity function for the goldfish. YARCZOWERand BITTERMAN (1965) have obtained ;I wavelength discrimination function for the goldfish which suggests that there is finite discrimination throughout the spectral region investigated (450-630 nm). They failed to control brightness cues, however, and it is premature, therefore, to make rigorous deductions from their data about the color vision systems of the goldfish. PHYSIOLOGICAL Several lines of physiological evidence indicate that fish have the appropriate nervous mechanisms for color vision. For example, SVAETICHIN (1956), working on several teleost marine fishes, discovered slow potential changes in response to flashes of light, in the region of second-order retinal neurons with which the light receptors synapse. An L (luminosity) response appeared abruptly when the electrode reached a certain depth in the retina, with the appearance of a negative resting potential of 25-35 mV and large negative-going (hyperpolarizing) electrical changes during illumination. A second type of response. the chromatic or C response, was obtained from a slightly more proximal region of the retina. When the retina was illuminated by flashes of light from the short-wave end ofthe spectrum, the responses were negative like the L type but with different amplitude variations as a function of wavelength. As progressively longer wavelengths were used, the response diminished in amplitude, until a transition point was reached at which only a transient response could be recorded. At still longer wavelengths, the response became positive. MACNICHOL,WOLBARSHT and WAGNER(1961) also found the L and C responses in the goldfish retina. They studied the retinal ganglion cells as well, and found here that most of the units gave an ‘on-off’ type of discharge pattern; that is, they gave a short burst of action potentials both at the beginning of a stimulus, and at the cessation of the same stimulus. In addition, some units were found that gave a prolonged discharge during illumination. and a few which were spontaneously active in the dark and were inhibited by the light. Also, units were found that gave an ‘on’ response at short wavelengths and an ‘off’ response at long wavelengths, and vice versa. In some units, the response changed from pure ‘on’ to pure ‘off’ for a wavelength change as little as 10 nm. It appears that there may be ;I translation of the hyper- or de-polarization effects into patterns of inhibition and excitation in the ganglion cells. The ganglion cells project to the optic tectum, the primary visual projection area of the fish brain. JACOBSON(1964) and JACOBSONand GAZE (1964) have recorded from single units in optic tectum of the goldfish. Following 30 min of dark adaptation, individual ‘on’ and ‘off’ cells showed spectral sensitivity functions that peaked around 533 nm, and that

709

Spectral Sensitivity and Spectral Saturation in the Goldfish

were similar, but not identical to, a 533 nm photopigment absorption function. After adaptation to what the investigators described as white light, several types of spectral sensitivity curves were recorded. They found three different kinds of ‘on’ or ‘off’ cells, with transition points at about 570, 535, and 505 nm. They also found several pure ‘on’ or pure ‘off’ cells, with response sensitivity maxima at around 620 nm. BURKNARDT(1966) has obtained both dark- and light-adapted spectral sensitivity curves for the goldfish electroretinogram in response to flickering stimuli. These curves show a Purkinje shift with the photopic function peaking at about 575 nm and the scotopic at about 500 nm with a shallow sub-maximum around 560 nm. In addition to the electrophysiological information, a great deal is now known about the visual pigments in fish. This literature has been reviewed by CRESXELLI (1960). The rod pigments, which have been investigated extensively in solution, fall into two general categories: rhodopsins, in marine fish, most of which absorb maximally between 486 and 511 nm, and the porphyropsins, in fresh-water fish, that have maxima1 absorptions between 520 and 535 nm (DARTNALLand LYTHGOE, 1965). It has been proven very difficult to extract cone pigments, but recently, two groups of investigators have succeeded in measuring difference spectra from individual cones in the goldfish retina (LIEBMANand ENTINE. 1964; MARKS, 1966). They found three classes of cones, with modal difference spectra peaking at 453, 535, and 620 nm. There is clearly a substantial amount of knowledge about isolated stages in the goldfish color vision mechanism in photochemistry, neurophysiology, and behavior. The present experiments were designed to determine systematic~ly some of the functional relations in goldfish color vision and to relate the quantitative visual discr~~nation data to the measured physiological responses. This was done within the framework of a quantitative theory of color vision, a theory which contains a coherent set of mathematical relations which are invariant under different sets of adaptation and viewing parameters. The predictions of the theory are testable experimentally.

SUBJECTS AND LIVING CONDITIONS Subjects for this experiment were common goldfish, Carassius auratus, five to eight inches long. They were housed in individual 3+ gallon tanks, with constantly filtered water. The temperature of the water was maintained at 78”-80” F. with thermostatic electric heaters. The home tanks were in a room illuminated with fluorescent light. The illumination was provided by an overhead installation of Electric ‘Cool-White’ Fluorescent tubes, 3 feet long, #F4OOW. At tank level the illuminance 36 foot-candles; this level of illumination was maintained so that the fishes’ eyes would be in light-adapted state.

constantly 12 General was about a constant

APPARATUS The general plan of the apparatus is diagrammed in Fig. 1. A complete description is given in YAGER (1966).

Ftc. 1. General plan of apparatus: A, stimulus box; B, ex~ri~ntal tank; C, translucent patches: D, transparent targets; E, electromechanical feeder; F, black target.

710

DEAN YAGER

At the front of the experimental tank, two transparent plexiglass targets were suspended in the water; they were attached to mechanical fish operant conditioning levers designed by HOGAN and ROZIN ( 1962). Directly behind each of these targets there was a rectangular stimulus patch, 9 mm wide and 11 mm high, visible through a 1 x 7 in. window in the near-white paper which lined the tank. The stimulus patches were made with translucent tracing paper (Dietzgen Tracing Vellum, i198M). A third target, of black bakelite. also attached to an operant conditioning lever, was suspended in the water at the back of the tank. An electromechanical food delivery magazine was fixed above the center of the tank; food reinforcement was provided by 20 mg lab rat food pellets from the P. J. Noyes Company, Lancaster, N.H. Visual stimuli were provided by illuminating the translucent rectangular patches with light from nsrrowband interference filters (Balzers, Liechtenstein; half-band width: I l-12 nm) inserted into a collimated beam of incandescent light. The source was a 150 W tungsten coil filament bulb with a clear envelope, run at 110 V a.c. The color temperature of this light as transmitted by the translucent patch was about 2750’ K. Two identical optical pathways were used, one for each stimulus patch. Figure 2 gives a detailed schematic diagram of the left optical system.

FIG. 2. Diagram of left optical system: A, incandescent primary source; B, collimating lens; C, neutral density filters; D, interference filters and shutter; E, neutral density wedge; F, balancing wedge; G, slide cover glass; H, neutral density screens; I, incandescent secondary source; J, opaque screen; K, translucent patch.

The intensity of light on each stimulus patch was controlled with calibrated Kodak Wrntten neutral density filters, a neutral density wedge, and a balancing wedge. A secondary light beam was used to provide a diluting light for the saturation experiment. The source was a Lampette (Model E6, Koch Creations, Lynbrook, N.Y.) run at a primary coil a.c. voltage of 110. The light was reflected onto the stimulus patch by a cover glass set at a 45” angle in the primary optical path. The color temperature of this illumination as transmitted by the translucent patch was 2850°K. The intensity of this illumination was controlled by wire screens which are essentially non-selective in their absorption, and thus the color temperature of the dilution light was kept nearly constant as its intensity was varied. A nearly light-tight cover was fitted to the tank, and the illuminance inside the tank with the cover in position and with no lamps switched on inside the tank was less than 0.019 ft-c. A 40 W incandescent interior-frosted bulb (G.E. #25 TIO/IF, 5% in. long) was attached to the underside of the cover. At a maintained voltage of 110 this lamp provided an illuminance of about 100 ft-c. at the center of the floor of the tank (color temperature 2400” K). The illuminance was measured with a Macbeth Illuminometer with no water in the tank. The luminance of the tank’s sides was about 6.2 ft-L and of its back, about 4.2 ft-L. The spectral distribution of the light reflected from the sides of the tank under this illumination had a color temperature of 2400” K. Also attached to the underside of the cover was a small 28 V incandescent lamp (color temperature 2500” K) which provided an illuminance of about 6.2 ft-c. at the center of the floor of the tank (about l/17 the level of the large adapting illuminant). The bulbs attached to the cover of the tank were approximately one inch above the surface of the water. A small exhaust fan dissipated the heat produced by the bulbs inside the tank. Reinforcement contingencies and the switching of stimulus lights were controlled by conventional relay circuitry. Wavelength and light intensity were controlled by changing the filters and wedge settings by hand between blocks of trials. Calibrations of stimuli were made in situ with a ‘Photovolt’ photomultiplier instrument which had been calibrated against a thermopile. The outputs of the stimulus lights were checked regularly with a Macbeth Illuminometer, and the spectral distribution of the sources were measured with an ISCO Model SR spectroradiometer.

Spectral Sensitivity and Spectral Saturation

in the Gotdfish

711

Subjects were trained to press the transparent targets for food reinforcement by a conventional operant ~onditjoning technique. The full details of the training procedure are given in YAGER(1964). The basic experimental method was to use a two-lever choice situation, in discrete trials, with a response on a third lever required to initiate a trial. The black target at the back of the tank was used for this observing response, which simultaneously turned on a stimulus light behind one of the transparent targets and extinguished the adapting lights. The introduction of this observing response accomplished two things. (I) The stimulus lights came on only when the fish was ready, in some sense, to ‘attend’ to them, and (2) the fish was forced to be at a maximum distance from the stimulus patches when they first were illuminated, and had to swim the length of the tank while oriented toward them before he made his choice response. This presumably would also increase the degree of ‘attention’ to the stimuli. ~XP~~rMENT

NO. 1

The apparatus was programmed to maintain the fish in a state of light-adaptation. The overhead bulbs stayed on for at least 30 see prior to each trial. After at least 30 see, a press on the black target in the rear of the tank simultaneously turned off the adapting lights and exposed the stimulus light behind one of the transparent targets. If the fish pressed the illuminated target within 15 set after pressing the black target, a correct response was recorded, the feeder operated, the stimulus light went out, and the small overhead lamp came on so that at no time during the experiment was the fish in total darkness. Fifteen see after the iIluminated target was pressed, the large adapting lamp came back on and once again stayed on for at least 30 sec. If the fish pressed the non-illuminated transparent target rather than the illuminated one within 15 set after pressing the black target, an error was recorded, the stimulus light went out, the small overhead lamp came on, and 15 set later the large overhead lamp came on. On the average, the fish tended to respond about eight set after the stimulus light onset. If the fish pressed neither transparent target within 15 set after pressing the black target, the overhead lights came back on simultaneously, and the trial was not counted. This schedule sought to insure that the fish’s eyes would be light-adapted at the time when a discrimination was made, and made it possible to present the stimuli with no surround illumination, In order to find the stimulus energy levei of a given wavelength required for a 75 per cent choice of the illuminated target, a modified method of constant stimuli was used. For each wavelength. the stimulus level used in the first block of trials was set at the m~imum intensity available in the apparatus. Repeated trials were continued at this level until-the fish made at least 20 correct responses in a block of 25. More often than not this criterion uas reached in the first block of trials at each wavelength. A few exploratory trials were then run to find the approximate range of energies for which choices would range from about 50 per cent to about 85 per cent correct. The stimuli in this range were subsequently presented in blocks of 25 trials, first in a descending, and then an ascending series, for a total of 50 trials at each energy level. Energy was varied between blocks of trials in steps of approximately O-25 log unit, and from four to six different energy levels were required to cover the total response range. In most cases both the descending and ascending series for a single wavelength were presented to the fish in a single experimental day. Occasionally it was necessary to run the two series on different days. Wavelen~hs were presented in sequential order in the spectrum, and the order was balanced among subjects. For each wavelength employed, a graphical representation of log energy vs. per cent correct responses was made. A smti,oih curve was fitted by eye to these points, and the energy for 75 per cent

712

DEAN YAGER

correct responses was determined by interpolation between empirically obtained data points. This type of energy determination was made at 12 points in the spectrum: 401, 453. 484, 510, 535, 554, 582, 598, 625, 653, 690, and 755 nm. Density settings for 75 per cent correct responding to broad-band tungsten light were also determined. Five subjects were used in this experiment. Results This procedure yielded data that were quite orderly, and in most cases it was possible to determine the energy level required for 75 per cent correct performance with a relatively high degree of accuracy. Figure 3 shows examples of these psychometric functions for a

I v’ -2.0

l -1.5 ‘I

.

554NM

1 535 NM . 1

* L.

-20

-1.5

LOG REL E FIG. 3. Psychometric

functions for photopic spectral sensitivity for one fish.

single fish. It is clear from these data that the fish’s choice behavior was under strong stimulus control. ET5perCentwas selected as the criterion choice response level because, in general, this was where the function was rising most steeply, and consequently there was less error in estimating the critical energy point. The uncertainty as to the correct ET5percent is probably not greater than f0.1 log unit, and, in most instances, less than that. Log ll% percentfor each wavelength is plotted in Fig. 4 for each of the five fish used in this experiment. These curves are in energy units at the cornea, no correction having been made for pre-retinal absorption or quanta1 energies. All fish show maximal sensitivity near 450 nm, there is a rapid decrease in sensitivity beyond 650 nm. and there are suggestions of secondary maxima in the regions of 535 or 554 and 625 or 653 nm. Further analysis of these data will be left to the discussion section. The point at 401 run for G4 requires explanation. Ci4 was the first fish used in this experiment. The experimenter was expecting the fish’s sensitivity to drop at short wavelengths. When G4 failed in the first block of trials to produce over 80 per cent correct responding to the highest intensity of this wavelength, the

713

Spectral Sensitivity and Spectral Saturation in the Goldfish

WAVELENGTH fnm)

1 at cornea. El5 percant displaced 0.5 log unit for clarity.

Fm. 4. Photopic spectral sensitivity for five fish: Log

Functions

experimenter nevertheless began to decrease the intensity in subsequent blocks of trials. The other four fish responded at over 80 per cent correct in the first Mock of trials at the highest intensity of 401 nm, and their performance did not reach chance level for several downwardsteps in intensity. In view of the performance of the other four fish, it is probable that G4 would also have performed better at this wavelength if it bad been m-run at the highest available intensity. G4 was not run with the 755 nm filter because it was unavailable at that time.

Comparisons are often made between the discriminative capacities of animals and those of humans. However, the conditions, the stimulus parameters and experimental procedures under which these discriminations are measured are, as a rule, different in important respects. Moreover, human spectral sensitivity experiments typically require that the subject’s head and eyes be carefully positioned, and a fixation point is used to guarantee that the stimulus image always falls on the same, precisely specified part of the retina. This is obviously not true for a fish that is swimming freely in the tank when the light stimuli are exposed. It was therefore decided to obtain comparable functions with a human observer in the same apparatus and with the same general procedure that was used for the fish. Since humans can verbalize it was hoped that some further insights might be gained from the reports of the human subject. In addition, there is at least a hundred years of visual experimentation on human subjects with which comparisons may be made. The covering was removed from the back of the experimental tank and the subject was seated behind the tank under a tent made of black cloth, which excluded most of the room illumination. The subject’s position was such that her eyes were appro~mately at the level of the stimulus patches, and about three or four inches from the end of the tank. The stimulus patches subtended an angle of about 1” 19’ in the horizontal dimension and lo 30’ vertically. The angular separation of the two patches was about 13”. No bite board or chin rest was used.

DEAN YAWR

714

The procedure was similar to that used in the light-adapted fish experiment. although the subject gave verbal responses rather than lever presses. She was also given verbal feedback rather than food reward. Stimuli were presented in discrete trials folIoGng a period of 30 see light adaptation. The stimulus lights stayed on for eight set and at the end of this period, the subject indicated which one of the two patches she thought was illuminated. The eight set stimulus duration was used, rather than 15 set, because that was the mean time that the fish waited before responding on one of the transparent targets, During the period of adaptation, the subject was instructed to fixate the area of the lining of the tank immedjately below the stimulus patches. During the period of stimulus presentation (with the adapting light off), she was instructed to move her eyes slowly back and forth between the stimulus patches in an effort to approximate the visual effect of the fish’s swimming movements. Because the human subject was much more reliable in her responses than the fish, many fewer trials were required to obtain an estimate of El3 percentcorrect responding. A smooth curve was drawn through the points, as with the fish, and & yerccntwas t-cad off the graph. The subject was a 25-year-old woman with normal color vision. This was established by her mid-point and range scores for the Rayleigh equation measured in a Nagcl anomaloscope. Results Log

1 Ia75

per cent

at each wavelength for this subject is shown in Fig. 5. The measure-

EOw WAVELENGTH

(nml

FIG. 5. Photopic spectral sensitivity for observer FY. Log

1 El5 per cent

at cornea.

mentB have not been corrected for pre-retinal absorption or differences in quantai energy. This function is directly comparable to the fish functions in Fig. 4. The measured curve differs from the familiar foveai spectral sensitivity function in that its peaks at about 484 nm and shows progressively reduced sensitivity at longer wavelengths except for a near-plateau in the region 625-653 nm. The typical photopic fovea1 function

Spectral Sensitivity and Spectral Saturation

in the Goldfish

715

peaks at about 550 to 540 nm. In this experiment, we remember, the subject was not restricted to using her fovea but was deliberately instructed to vary her fixation during stimulus presentations. Indeed, she sometimes reported that she could see the stimulus light when she looked away from it but not when she fixated it directly. Since spectral sensitivity is known to depend on retinal locus, it would have been surprising if the function obtained with varying fixation had duplicated the typical fovea1 curve. Studies by STILES and CRAWFORD (1933) and WEALE (1953) present light-adapted spectral sensitivity functions at different retinal eccentricities. An example of a peripheral function published by Weale is also shown in Fig. 5. Weale’s function is for a heterochromatic brightness match to a constant stimulus of 5636 nm at a luminance of 1.5 e.f.c. and was measured at a fixed retinal locus 45” peripheral to the fovea. Again, it would be surprising if this function were identical in form to that obtained in the present experiment, but it does illustrate the general point that the light-adapted periphery shows heightened shortwave light sensitivity relative to the fovea. EXPERIMENT Spectral

NO. 2

saturation

A monotonically related psychophysical measure of saturation is least calorimetric purity, i.e. the least amount of spectral light that needs to be added to a white light to make it just-detectably different in color from the white alone. This measure is also frequently referred to as a measure of saturation discrimination. The form of this function is strongly dependent on the specific characteristics of the color vision mechanism. If there is no discrimination at any wavelength, then the animal is probably color blind. If the animal can discriminate most wavelengths from a broad-band stimulus (white) except for one or two points in the spectrum where saturation approaches zero, then the animal is probably a dichromat, like the red-green or the yellow-blue blind human. If the animal can discriminate all wavelengths from white and there are no points in the spectrum where saturation approaches zero, as with a normal human, then the animal probably has trichromatic color vision. Conditions in the experimental tank were the same as those that prevailed when the photopic spectral sensitivity function was measured. In addition, the secondary dilution stimulus lamps (described in the apparatus section) were turned on, and density screens were used to bring the luminance of the stimulus patches to l-65 ft-L. On the basis of the threshold photopic sensitivity measures for both monochromatic and broad-band light stimuli, it was possible to determine wedge and density settings which would produce a broad-band stimulus that would be approximately the same luminance for the fish as a given monochromatic light. Two neutral density filters were then chosen which bracketed this monochromatic light luminance over a range of l-0 density unit. Now, when the fish pressed the black target to initiate a trial, monochromatic light was added to the ‘white”2 light already present on one stimulus patch, and additional ‘white’ light was added to the other. The luminance of this ‘white’ light was varied above and below the luminance of the monochromatic light, from trial to trial. The fish was reinforced for pressing the transparent target in front of the stimulus patch illuminated with both monochromatic light and the broad-band diluting light, and not reinforced for pressing the other target not illuminated with spectral light. The timing of light presentations and 2 ‘White’ is used here as a shorthand notation for a broad-band tungsten light distribution. There is no a priori reason to believe that it is perceptually neutral for the fish, but only that it is likely to be perceived as much less saturated than a narrow-band stimulus of the same hue.

DEAN YAGER

716

the

switching of the monochromatic lights were exactly the same as in the photopic spectral sensitivity experiment. Thus, the fish had to choose the ‘colored’ stimulus in order to be reinforced, and luminance differences were eliminated as a consistent cue. Determinations of the energy of monochromatic lights required for 75 per cent correct performance were made just as in the spectral sensitivity experiment. The stimuli Lvere presented in blocks of 25 trials at a given energy, in descending and ascending series. and covered the range of energies that produced about 50 per cent to about 85 per cent correct responding. Results The psychometric functions for the determination of the energy required for 75 per cent correct choice when the fish were responding to mixtures of spectral and broad-band lights were similar to those obtained in the spectral sensitivity experiment. These energy values can be converted to luminance units by the following equation: L,, = 2

(1) ,1 where ETA is the energy at threshold, determined in the photopic spectral sensitivity experiment, and EoAis the energy required for 75 per cent correct responding in the present experiment. Thus, L, is luminance stated in units of threshold energy, which is a standard photometric convention. The same operations can be carried out for the broad-band light (w). With luminance values L,, LAI, LA*, . . , , LA,, the ratio

can be computed for each wavelength; this is the reciprocal of the least calorimetric purity required for detection of a saturation difference, which has become the conventional way of expressing the human saturation discrimination function (e.g., PRIESTand BRICKWEDDE, 1926; MARTIN, WARBURTON, and MORGAN, 1933). In the present experimental procedure, L, is constant, and the value of expression (2) is a function only of the amount of monochromatic light added: the more light added, the smaller the ratio. This expresses the fact that more of a desaturated spectral light is needed than of a saturated spectral light to produce a just-noticeable saturation change. LvfL for each of the three fish used in this experiment. Figure 6 is a plot of log - L A

Each

function was computed individually for each fish whose own threshold energy values were used. It was obvious that saturation discrimination for all three fish is maximally sensitive at the spectral extremes, and that there is a third maximum in the mid-spectral region between about 510 and 535 nm. There is a well-defined minimum at around 600 nm, and a shallower one around 490-500 nm. None of the monochromatic lights used in this experiment were confused by any fish with the broad-band tungsten light. The evidence suggests that all spectral wavelengths elicit color responses that differ from the response to the broad-band tungsten distribution, and that goldfish vision is trichromatic. The relation of these functions to other characteristics of the visual system will be treated later in the discussion section.

Spectral Sensitivity and Spectral Saturation in the Goldfish

717

WAVELENGTH 400

500

600

700 BOO

i0 WAVE N~~ff~rn-‘) FIG. 6. Spectral saturation discrimination for three fish: Log reciprocal of least colorimetricpurity. In the lower part of the graph the range of measurements for the three fish is shown.

SpectraI saturation:

Human cmtrol

experiment

The same human subject from whom the spectral sensitivity data were obtained also observed in a saturation discrimination experiment. The appartitus was the same as for the fish, and the conditions of observation (adaptation, light-shield, and so forth) were identical to those under which photopic spectral sensitivity was measured. Just as in the fish experiment, the monochromatic and the tungsten light luminances were determined from the photopic threshold values for the subject, and two neutral density filters were chosen which bracketed each monochromatic tight Iuminance over a range of 1-O density unit. The subject’s eyes were ~jght-adapted for 30 see, and then she was asked to indicate, at the end of an eight-second stimulus presentation, which of the two stimulus patches was colored. Determinations of the energies of monochromatic lights required for 75 per cent correct performance were made just as in the spectral sensitivity experiment. 3.0

0.0

400

OBSERVER

FY

500

600

700

WAVELE~TH (nml FIG. 7. Spectral saturation discrimination for observer FY, uncorrected (see text).

DEANYAGER

718 Results

E r5pcrcencvalues were converted to luminance units by equation (l), and the reciprocal of least calorimetric purity calculated for each wavelength by expression (2). This function is plotted in Fig. 7. Saturation discrimination is maximally sensitive around 625 nm, and shows a tendency to rise from about 484 nm up to this wavelength. The form of the human function is quite different from the results for the fish (Fig. 6). Whereas the fish showed maximal discrimination at both spectral extremes, the human function tends to be relatively low in the short-wave region and high only at longer wavelengths. It is obvious, furthermore, that this function is very different from those of normal observers reported earlier. Data from MARTIN, WARBURTON, and MORGAN (1933) are reproduced in the upper part of Fig. 8. All three of these curves show maximal sensitivity

0.0.

400

I

0 X)0

1

I

L

700

FIG. 8. Spectral saturation discrimination for observer FY, corrected (see text).

in the shortwave region and a pronounced minimum in the vicinity of 570 nm. The probable reason for the difference is that the subject in the present experiment used different retinal areas in the spectral sensitivity and spectral saturation experiments. As already noted, she reported during the spectral sensitivity experiment that it was often possible to see the stimulus when she looked to one side, but that it disappeared when she fixated it directly. In the saturation experiment, on the other hand, the subject reported that she always saw the color better when looking directly at the stimulus patch, i.e. when she was using her fovea. Since fovea1 and extra-fovea1 photopic spectral sensitivity curves are so different (partly because of macular pigment), the transformation from stimulus energies to luminances will differ for the different retinal areas; derivation of the least calorimetric purity function, which is in turn dependent on these transformations. will consequently be different. Since it is clear from the subject’s reports that she was using her fovea in the saturation discrimination experiment, the least calorimetric purity function was recalculated using an average fovea1 spectral sensitivity curve for the CIE standard observer to convert the stimulus energy data obtained in the present experiment to conventional luminance units. If only the saturation discrimination data were available, this transformation would simply follow the standard photometric procedure for converting energy to luminance

Spectral Sensitivity and Spectral Saturation in the Goldfish

719

units. This function is plotted in the lower half of Fig. 8. Recalculated in this way, it is seen to be very similar in shape to the functions reported by Martin et al, Since the bracketing stimuli for the saturation di~rimination experiment were based on the photopic sensitivity measures, if those measures are not appropriate then we must also question the adequacy of the controls over brightness cues. The subject was instructed to respond to color, and in fact performance reached chance level only when she reported seeing no color in the stimulus patches. Consequently, even if brightness cues had been present, they clearly were not used for the discriminations. Moreover, a logical analysis of the situation (assuming that both bracketing lights appear dimmer or both appear brighter than the monochromatic light) shows that if a subject attempted to respond to either the brighter or dimmer of two stimuli in this experiment, the ‘false criterion’ would be detectable from the psychometric functions, because discrimination would fail at a constant level of the broad-band bracketing stimulus for different wavelel~gths. No such constancy was found, either for the human or the fish. It has also been shown that another species of fish (African mouth breeder), after it has been trained to respond positively to color with the use of bracketing Iuminances for the broad-band negative stimulus, will subsequently give the same psychometric function if the two luminances of the broad-band light are

deliberately chosen so that the luminance of the monochromatic Iight does not lie between them. (Dorothea Jameson, personal communication.) Furthermore, the goldfish retina is much mom homoge~us than that of the human, with no welidefined fovea (BRETT,1957). The shape of the luminosity function is probably not so strongly dependent on retinal locus. DISCUSSION AND THEORETICAL

ANALYSIS

Photopic Spectral Sensitivity

Spectral sensitivity must be related, by some function, to the spectral absorption curves of the visual pigments found in the retina. The light absorbed by these pigments in some way triggers neural responses which are combined at some point in the visual system to produce ‘brightness’ or at least some perceptual quality which enables the fish (or human) to make a discrimination. Recently, LIEBMAN and ENTINE (1964) and MAWS (1965) have succeeded in recording difference spectra from individual cones in goldfish retinas. The following discussion will attempt to relate the spectral sensitivity functions obtained in the present experiment to these cone photo-pigments. WAVELENGTHh)

FIG. 9. Photopic spectral sensitivity for three fish, equal quantum intensity spectrum at the retina.

DEAN YAGER

720

The spectral sensitivities in Fig. 4 are specified with respect to the energy incident on the cornea. To be compared with the absorption spectra of visual pigments, the data have been corrected for pre-retinal absorptions on the basis of BURKHARDT’S (1966) measures. Also. the retinal sensitivities have been expressed in terms of an equal quantum intensity spectrum. Figure 9 is a plot of relative spectral sensitivity for an equal quantum intensity spectrum at the retina, for Fish G6, G7, and G8. In this graph, the abscissa is a frequency. rather than wavelength, scale. Only these three S’s were used in this analysis because these were the three on which a saturation function was also obtained, and the results of this analysis will be used later in discussing the saturation data. WAVELENGTHh-d

25

23

21

19

17

WAVENUMBERbn-‘1

15

J

I:3 x I(13

FIG. 10. Dartnall nomograms for goldfish cone pigments.

Figure 10 shows the Dartnall (1953) nomograms for visual pigments 455, 535, and 620 nm; these are the modal peak wavelengths obtained for spectral absorption in single goldfish cones by Liebman and Entine, and by Marks. Pigment nomograms were used in this analysis rather than the measured difference spectra because of the variability of the microspectrophotometric data, and because of the possibility that photoproducts distorted the difference spectra obtained by measuring absorption before and after light exposure. Notice that these curves are plotted with a frequency scale as the abscissa, and, therefore, are identical in shape. Notice also that the 620 nm pigment curve is shown as a dashed line from 400 to 460 nm, and the 535 nm pigment curve from 400 to 420 nm. Dartnall has applied his generalization only to the primary absorption peak in the visible range, and the behavior of the short-wave absorption band with changes in the spectral locations of the primary maxima has not been fully explored. There are two models that can be used to relate spectral sensitivity to photopigment absorptions. One of these is the envelope model which assumes that the threshold at any wavelength is determined by the single most sensitive mechanism at that wavelength. The other assumes that all mechanisms contribute to the threshold response, in proportion to their respective separate sensitivities. The latter is the additive model which. unlike the envelope model, also applies to supra-threshold spectral luminances. It was decided to attempt to relate pigment absorptions to spectral sensitivity by applying the simple additive model in the present analysis. (A recent experiment by BOYNTON, IKEDA and STILES (1964) questions the validity of using such a model, because events at threshold are more complicated than the simple additive model suggests. However, this model may be used as a first

Spectral Sensitivity and Spectral Saturation

approximation.)

in the Goldfish

121

This model assumes that

B, = .f(~,a,+kzP~+k3~~), (3) where B=brightness, a, fi, and y are percentage absorption functions for three selective photopigments, and kr, k2 and k3 are factors which express the spectral absorptions of the three photopigments relative to each other. Now, for any measure of equal brightness at different wavelengths, A and h’, BA = BAT=> klaA+-k2PA+k3yA = kp,v+k28A*+k3yn,, (4) and so, in order to find the relative heights of the three functions, we need to find kl, k2, and k3, such that B is constant for all 1,. The spectral sensitivity curve obtained in the present experiment can be treated as a measure of the number of quanta needed at each wavelength to produce a constant (threshold) brightness. Then, k,, k2, and k3 are just those factors that give SA = kla~tk219A+k3Y~ (5) where S is the behavioral measure of relative sensitivity. Although no exact fit can be expected because of inter-subject variability, the deviations may be minimized by seeking a least-squares fit to the data. This was done using the data of all three subjects for sensitivity measures from 484 through 690 nm; the factors obtained were kl=O-25, k2=0-14, k3=0*09. The sensitivity function calculated by equation (5) using the estimated k factors is also shown in Fig. 9 along with the behavioral data. The photo-pigment summation function describes a smooth curve which falls fairly close to the median measurements for the three S’s. The close agreement is neither surprising nor particularly important, since the k’s were chosen to minimize the differences. It will be seen, however, in accounting for the saturation function, that these particular k values may indeed represent biological parameters in the goldfish retina. The sensitivity values measured at 401 and 453 nm clearly depart from the theoretical curve. Three reasons for such departure can be cited: (1) As already noticed, there is considerable uncertainty about the shape of the pigment absorption curves in this region of the spectrum; they probably rise at short wavelengths, as does that of ‘cyanopsin’ (WALD, BROWNand SMITH, 1953). (2) Ocular media, as well as the extra-ocular media comprising the fish tank, water, and plexiglass probe may fluoresce when irradiated by short wavelengths of light. This fluorescence would distort the sensitivity data. (HOSOYA,1929). (3) These same materials also scatter short-wave light; in human experiments, this can have the effect of significantly raising shortwave fovea1 sensitivity, due to scatter to more sensitive retinal areas. The control function for the human obtained with the present apparatus (Fig. 5) shows distortions at short wavelengths similar to those in the fish function. Spectral saturation

Spectral saturation must also be related to the spectral absorption curves of retinal visual pigments. There are two theories of vision which are sufficiently quantitative to allow computation of theoretical saturation functions based on cone pigment responses. This first is Hecht’s development of the Young-Helmholtz theory of color vision. In his theory (1934) the total brightness of a spectral light is given by BA= ktaA+k$A+ksyA

(6)

DEANYAGER

722

when a, /S, y are cone response functions which, in the original analysis, corresponded to hypothetical photopi~ent absorption spectra for a human. Whiteness, in this theory, is associated with equal response levels in a, & and y, and the specific color component of the response is then given by CA = [kla,+k~S~+k~~n]-3[minimum of kta,+, k#.,, k3y,,J. (7) Spectral saturation is then the ratio of the color response to the total response:

(8) If hypothetical a, 8, y functions are chosen appropriately, this expression satisfactorily predicts the saturation discrimination functions of PRIESTand BRICKWEDDE (193% where it is assumed that K-

( > Ld-LA

= Sat,,

44

(9)

A theoretical saturation function based on this formulation can be derived with the use of the three best fitting photopigment functions found in the goldfish photopic spectral sensitivity experiment. The resulting theoretical function, plotted on a logarithmic scale. is shown in Fig. Il. WAVELENGTH 400

500

,rt--_

A 3 3

JAMESON

800

AND HURVICH

l.O-

xl_ _.*-

____---..* HECHT

0.0 ! 25

I 23

f

$8 21

I I 19

WAVE NEAR FIG. Ii.

700

---- :_%fiLi

h _2 +z

(nm)

600

I I 17

II 15

(cm-‘)

1 13

xIO3

Theoretical saturation di~rimination functions.

The second theory that can be used to generate a saturation discrimination function is the opponent-process theory of Hering, as developed quantitatively by Hurvich and Jameson (JAMESONand HURVICH, 1955; HURVICH and JAMESON,1957). This theory assumes that for a trichromatic visual system, there are two independent, paired chromatic response processes which are determined by positive and negative interactions among three fundamental processes a, B, y, and the positive component of the achromatic visual response (both brightness and whiteness) is determined by an addition of the responses a, /I, y. The specific equations relating a, 6, y to produce opponent chromatic response are specified for hypothetical photopi~ent absorptions for a human in the Zurich-Jameson analysis.

Spectral Sensitivityand Spectral Saturation in the Goldfish

723

In choosing particular equations to relate the measured photopi~ent abso~tions and opponent neural responses for the fish, it was decided to use the simplest expressions that would produce two opponent systems, and which would, at the same time, produce minima in the saturation function at positions roughly corresponding to those found in the behavioral function. If a, b, (I, d are names for chromatic responses for the fish, and W is the achromatic response, then aA--bA = k2pA-klaA CA-4 = kA -km WA = km +kzBA +km

(10) (11) (12)

where k,a, k#, kp are the best fitting photochemical absorptions for the spectral sensitivity function. These three visual response functions are plotted in Fig. 12.

FIG. 12. Visualresponse functions from the opponent-colors theory. JAMESONand HURVICH(1955) have shown that saturation discrimination is accurately predicted by the ratio of total chromatic response to achromatic response at each wavelength. This function, based on the following expression (13) has been computed for the goldfish. Rewritten on the basis of equations (lo), (1 I), and (12), we have (14) and this function is also shown in a logarithmic plot in Fig. 11. It is obvious that the opponent-process theoretical function bears a much closer relation to the behavioral function (Fig. 6) than does the function derived from the Hecht theory. Two minima are clearly present, and correspond closely to the behavioral minima. The fact that the relative heights of the two minima are in reversed order in the two functions may be due in part to the use of relatively low color-temperature comparison and diluting lights. A further possible reason for the deviations that occur may be in the choice of the interaction equations. The general expression for interaction of the receptor responses for

DEAS

724

YAGER

each of the visual response processes is Visual response = &Ct(kla) -!&Z(k2/3)&Ccj(kjy) (15) For simplicity, only values of unity or zero were chosen for C in the expressions for the three opponent-processes and hence this symbol does not appear in the equations for visual responses (equations IO? 11, 12). The C values could be manipulated to produce a better fit. No manipulation of the values for k 1, kz, k3 would allow the Hecht formulation to produce the double minimum exhibited by the behavioral data. It is recognized that a better-fitting positive three-channel model could be constructed. using the present photopigment functions. However, two a priori considerations lean toward the opponent-process theory as a better model: (1) HURVXH and JAMESON(1957) have had a good deal of success in accounting for several human visua1 functions with the use of this model. (2) As indicated earlier, opponent-type chromatically-coded responses have been discovered at several different levels of the fish visual system. It should be emphasized that the photopigment coeflkients used in the saturation formula were the same ones that were determined in the spectral sensitivity experiment. The opponent-process formulation for saturation discrimination may be tested by obtaining spectral sensitivity and spectral saturation functions under conditions of strong chromatic adaptation. The changes necessary in the coefficients kl, kz, k3 to account for the new spectral sensitivity curve should iead to consistent predictions about the changes in shape of the saturation functions_ This will, of course, include shifts in the locations of minima. From the visual response functions shown in Fig. 12, a theoretical wavelength discrimination function may also be computed, according to the formula of HURVICH and JAMESON (1957). Wavelength discrimination depends on both hue and saturation differences. Spectral saturation and hue coefficients may be calculated (Fig. 13), and variations in wavelength discrimination derived as proportional to the sum of the rates of change

FIG.

of the

two

13.Hueand saturation codicient

coefficient functions.

functions

Such a theoretical

from the opponent colotf theory. function for the conditions

of the

present experiment is shown in Fig. 14. (Figures 13 and 14 are plotted with frequency as

Spectral Sensitivity and Spectral Saturation

in the Goldfish

725

WAVELENGTH bxnl

J 25

23

21

19

I7

WAVE NUMBERkm-’ 1 FIG.

13

Xl03

14. Predicted wavelengthdiscriminationfunction from the opponent-colors theory.

the abscissa because they are photopigment-derived

functions. However, the analysis of the rates of change of the coefficient functions was made with a linear wavelength scale). The form of this function is, of course, specific to the stimulus conditions used in the present study. This awaits an experimental test. Furthermore, the visual response functions in Fig. 12 create strong expectations about the specific forms of retinal and tectal single-cell response functions under the present conditions of adaptation and stimulation. Although the stimulating and adapting conditions were very different, electrophysiological data that have been reported (e.g., SVAETICHIN,1956; MACNICHOL, WOLBARSHTand WAGNER, 1961; JACOBSON,1964)

unquestionably

show qualitative similarities to these theoretical response functions.

Acknowledgements-The

experiments

on

which this paper is based were conducted in the laboratory of

Mrs. DOROTHEAJAMESONHURVICH and

Dr. LEO M. HURVICH at the University of Pennsylvania. Their interest in this study, both in the experimental work and theoretical discussions, has bean invahtable to me, and is gratefully acknowledged. I would also like to thank my wife, Florence, for her assistance as the observer in the human comparison experiments, and Dr. DAVID H. K~UNTZ for his time spent in helpful discussions with me.

REFERENCES AUTRUM, H. (1960). Vergleichende Physiologie des Farbensehens. Fort&r. Zool. 12, 176-205. BERNSTEIN,J. (1961). Loss of hue discrimination in forebrainless fish. ExpZ Neural. 3, l-17. BOYNTON, R., IKEDA, M. and STILEZG, W. (1964). Interaction among chromatic mechanisms. Vision Res. 4,

87-117. BRETT,J. (1957). The eye. In Physiology of Fishes, Vol. 2, Academic Pr&q New York. BURKHARDT, D. (1966). The goldflsh electroretinogram: Relation between photopic spectral sensitivity functions and cone absorption spectra. Vision Res. 6, 517-532. CRESCITELLI, F. (1960). Physiology of vision. Rev. Physiol. 22, 525-578. CRONLY-DILLON, J. and MUNTZ, W. (1965). The spectral sensitivity of the goldfish and the clawad toad

tadpole under photopic conditions.

J. exp. Biol. 42, 481493.

DARTNALL, H. (1953). The interpretation of spectral sensitivity curves. Br. med. Bull. 9, 2430. DARTNALL, H. and L~THGOE, J. (1965). The clustering of fish visual pigments. In Co/our Vision, a Ciba

foundation

symposium.

HECHT, S. (1934).

London;

J. & A. Churchill Ltd.

process. In Murchison, C., Handbook of General Experimental Psychology, Clark University Press, Worcester, Mass. HERTER, K. (1953). Die Fischdressuren und ihre sinnesphysiologischen Grundiugen. Akademie-Verlag, Berlin.

The nature of the photoreceptor

726

DEAN YACER

HOGAN,J. and RozrN, P. (1962). An improved mechanical fish-lever. Am. J. Psycho]. 75, 307-308. HOSOYA,Y. (1929). Fluoreszenz der einzelnen Auaenmedian und Sichtbarkeit des ultraviolette Gebiete de5 Spektruis. T&okrr J. esp. Med. 13, 510-523. HURVICH, L. and JAMESON,D. (1957). An opponent-process theory of color vision. Psychoi. Rev. 64. 384-404. JACOBSON,M. (1964). Spectral sensitivity of single units in the optic tectum of the goldfish. Q. Ji cxp. Physiol. 49, 384-393.

JACOBSON,M. and GAZE, R. (1964). Types of visual response from single units in the optic teztum and optic nerve of the goldfish. Q. JI exp. Physiof. 49, 199-209. JAMESON, D. and HURVlrCH, L. (1955). Some quantitative sspe:ts of an opponent-colors theory. I. Chromatic responses and spearal saturation. J. opt. Sot. Am. 45, 546-552. LIEBMAN,P. and ENTINE,G. (1964). Sensitive low-light-level microspe:trophotometer: Detecrion of photosensitive pigments of retinal cones. J. opt. Sot. Am. 54, 1451-1459. MncNicHoL, E., WOLBARSHT,M. and WAGNER,H. (1961). Eiectrophysiological evidence for a mechanism of color vision in the goldfish. In McElroy and Glass, LigAr and Life, Johns Hopkins Press, Baltimore. MARKS, W. B. (1965). Visual pigments of single goldfish cones. J. Pltysiof., Land. 178, 14-32. MCCLEARY,R. A. and BEaNsTEIN,J. (1959). A unique method for control of brightness cues in the stud? of color vision in fish. Physiol. Zool. 32, 284-292. MARTIN, L., WARBURTON,F. and MORGAN, W. (1933). Determination of the sensitiveness of the eye to differences in the saturation of colors. Med. Res. Council (Brit.) Special Rept. Ser. No. 188. MIJNTZ. W. and CRONLY-DILLON.J. (1966). Color discrimination in rroldfish. Anim. Eehav. 14. 331-355. PatEsT, ‘I. and BRICK~EDDE, F. ‘(19%). ‘The minimal perceptible Eolorimetric purity as a &mction of dominant wave-length. J. opt. Sot. Am. 28, 133-139. STn_Es, W. and CRAWFORD,B. (1933). The liminal brightness increment. Proc. R. Sot. 1138, 496-530. SVAETKHIN.G. (1956). SDectral resDOnse curves from single cones. Acta Dhvsiof. scund. 39. 174%.suaol. 134. VtAuD, G. (196i)). Li vision chromatique chez les animaux (sauf les &&tes). In Mechanisms oj”Colmrr Discrimination. Y. GALIFRET,ed. Pergamon Press, New York. WALD, G., BROWN, P. and SMITH,P. (1953). Cyanopsin, a new pigment of cone vision. Science, X. Y. 118, 505-508. WALLS, Ci. (1963). The Vertebrafe Eye and its Adaptive Radiations. Hafner, New York. WARNER, L. (1931). The problem of c&or vision in fishes. Q. Rev. &of. 6, 329-348. WEALE,R. (1953). Spectral sensitivity and wave-length di~rimination of the peripheral retina. J. PI~.v.~iol., Lo&. 119,175-190. YAGER, D. (1962). Wavelength discrimination in the goldfish. Bachelor’s dissertation, Harvard College. YAGER, D. (1966). Behavioral measures and theoretical analysis of spectral sensitivity and spectral saturation in the goldfish, Curassias aurutus. Doctoral dissertation, University of Pennsylvania. YARCZOWER, M. and BITTERMAN,M. (1965). Stimulus-generalization in the goldfish. In Srirnrrlus Generalization. D. Mostofsky, ed. Stanford University Press, Stanford, California.

Abstract-Isolated information is found in the literature about photo-chemical receptor absorptions and physiological responses from several different levels of the goldfish visual system. To integrate this information into a synthesized account of visual response in the intact organism, precise di~rimination measures are required. An operant conditioning technique was used to obtain di~rimination measures in a two-lever choice situation. To measure photopic spectral sensitivity, the energies of spectral lights required for 75 per cent correct choice responding were determined at 12 points in the spectrum. To measure spectral saturation, the reciprocal of the least calorimetric purity required for 75 per cent correct choice responding to a mixture of spectral light and broadband tungsten light was determined for the same spectral stimuli. The data indicate that goldfish are sensitive to light over the spectral range of at least 401-755 nm. Furthermore, goldfish, on the basis of the measured saturation discrimination function, probably have trichromatic vision. The form of the saturation function is accounted for by a model of the visual system that postulates a summation or responses from different types of cones of known absorptions to yield a brightness response, and that postulates interactions between activities initiated in the different types of cones to yield opponent chromatic responses at the neural level.

Spectral Sensitivity and Spectral Saturation

in the Goldfish

R&am+-On trouve dans la litt~rature des informations &parses au sujet des absorptions phot~himiques des r&epteurs dans le systkme visuel du cyprin do&, ainsi que des reponses physiologiques B diffbrents niveaux. Pour int&rer ces informations dans une description synthbtique de la rCponse visuelle de l’organisme intact, il faut des mesures prkises de discrimination. On opL?re avec une technique de conditionnement, qui permet des mesures de discrimination dans une situation g double choix. On mesure la sensibilite spectrale photopique en 12 points du spectre par les tnergies monochromatiques nCcessaires pour un choix correct & 75 per cent. On mesure la saturation spectrale avec les m&mes stimuli monochromatiques par l’inverse de la purete ColorimCtrique minimale pour un choix correct B 75 per cent, par mtlange de lumitre spectrale et de lumihre du tungstene B large bande. Les r&,ultats indiquent que le cyprin dare est sensible B la lumi&e au moins dans l’intervalle 401 B 755 nm. 11 poss&de en outre probablement une vision trichromatique, &ant donnde la fonction de discrimination de la saturation obtenue. On rend compte de cette fonction de ~turation par un modtle du systkme visuel qui suppose une sommation des r&onses de differents types de c&es ~absorptions don&es pour obtenir une r&ponse de luminance, et des interactions entre les activit&s d&lench&s dans les diff%ents types de c&es pour aboutir B des rkponses chromatiques antagonistes au niveau nerveux. Zusammenfassung-Informationen iiber die photochemische Rezeptorenabsorption und iiber physiologische Reaktionen auf verschiedenen Stufen des visuellen Systems des Goldfisches sind verstreut durch die Literator zu finden. Eine Synthese dieser Information zu einem geschlossenem Bild der visuellen Reaktion im intakten Organismus bedingt genaue Unterscheidungsmal3nahmen. Mit Hilfe einer Lernmethode wurden UnterscheidungsmaBe in einer Zweierwahlsituation gewonnen. Zur Messung der spektralen Empfindlichkeit im photopischen Gebiet wurden die Energien, fiir 75% richtige Aussagen, von schmalbandigen Lichtquellen an 12 Stellen im Spektrum bestimmt. Bei der Bestimmung der spektralen S&ttigung wurde der Reziprokwert der niedrigsten Far~neinheit, die 75% richtige Antworten auf ein Gemisch von mon~hromatis~hemund Wolframgl~hlicht ergab, fiir den gleichen spektrasen Reiz ermittelt. Die MeDwerte weisen darauf hin, daD der Goldfisch auf Licht im Bereich von 401 bis 755 nm reagiert. Aus der gemessenen Siittigungsunterscheidungsfunktion kann geschlossen werden, daf3 der Goldflsch vermutlich ein Trichromat ist. Die Form der Funktion l&l& sich durch ein Model1 des visuellen Systems erkliiren, welches eine Summation der Reaktionen von verschiedenen Zapfenarten bekannter Absorption mit einer dadurch hervorgerufenen Helligkeitsreaktion, postuliert. Es postuliert weiterhin, da0 die Wechselwirkungen zwischen den Aktivitiiten die in den unterschiedlichen Zapfenarten entstehen, entgegengesetzte Farbreaktionen auf neuraler Stufe erzeugen. Pe3loMe B Jnrrepa-rype HM~~TCP pa3po3HeHHaa sf~~@opwwis~ 0 c~mf Mexny nornoxueH%eM u @o~oxa~~eiezr peuen_ropoB, a TaKxe ~~3~onoruq~~~ 0meTami H~p~3n~~HbIXy~OB~~X3p~T~bHO~~uCTeMbI3OROTO~ pbI6~.~~~~~e~~ 3TOfi ~H~OpMaUH~ 0 3puTenbHO~~y~B~~bHOCM:BHT~KTHOrO Opr~Ma~eo6xo~l ToSHble 3xc~ep~~e~r~ B 06nacTa pa3n~qeH~~. 6b~naucnonb3oBaHaoperantconditioningMeTox~acBbIfSopoMH3~~ nonoxcensi&. Ana mMepenHn f&OTOII%iWCKO~ ClleKTpZiJIbHOBVyBCTBEfTenbHOCTH 6bxna OIIpej@SeHa 3Hepran MoHoxpoMaTwxKoro a3nyueHurr B 12 ToqKax CneRTpa, fleo6xoxzisraK nnn npaB&inbHOrO (75%) Bbr60pa. &~a a3MepeHHn CIleKTpFUIbHO& HaCbuueHHOCTK 06” paTHan BenHmHa HanMeHbruefi KonopeMerpmxxoik WCTOT~I ~eo6xo~u~olli LIP npaB&inbHorO (759$BbI6opa, OTHOCIIUULRCIIK CMeCIi MOHOXpOMaTE'IeCKHXHury9eHlfltt U U.IHpOKOnOnOCHOMy If3ny'ieHHKIJIah%IIbI HaKaJlHBiWUiI1OII~#ZJIXJIZlCbAJUI TOrO XCe MOHOXpOMafUYeCKOrO CTnMyna. llonyreHHble nanwe noKa3anu, 9~0 30noTaK pH6Ka 06nanaeT muTenbHOcTbw K CBeTy B Cllt2KT&XiJlbHOM JlSiana3OHeKaKMHHIlMYM OTml X0 755HM. EOJIee TOrONa 0cHoBe si3MepenHbrx @~HKUH& pa3nwiewfs no HacbnnewocTu, 30noTaK pbI6Ka, BepOSiTHO, HMtXTTfX?XUBeTHOe3peHHe. ~OpMa~ynK~~HaC~~eHH~THo6~XcH~eTc~ MOa~nbK,3~LIT~JlbHO~CETCT~MbI,BKOTOpO~~p~H~~CyMMa~~OTBeTOBOT xon609eK pa3nWEHblX THfIOB c H3BeCTHbIM norno~eH~eM, ~cne~~~ CB~T~OTHYK) peaKIll% II nOn5lpHOe XpOMaT~~~K~ B3a~MO~e~CTB~e peaXU@ Of pa!3JIWiHblX T~~nOBKon6o~eK Hii He~Tp~bHOMypOBHe.

727