Psychometric and psychophysical hue discrimination functions for the pigeon

Vision I&T. Vol. 12. pp. 1447-1464. Pergsmon Press 1972. Printed in Grtat Britain.

PSYCHOMETRIC DISCRIMINATION

AND PSYCHOPHYSICAL HUE FUNCTIONS FOR THE PIGEON’ ANTHONYA. WRIGHTS

Department of Psychology,ColumbiaUniversity,New York, U.S.A. (Received 13 July

1971; in revisedform 3 January 1972)

PSYCHOPHYSICAL researchers have traditionally employed four approaches in attempting to determine the underlying processes of color vision (BOYNTON,1963): (1) color discrimination; (2) color mixture; (3) chromatic adaptation; (4) spectral sensitivity. Data on color discrimination contribute to our understanding of color vision because those spectral regionswhere color discriminationis best identify regionswhere the relative contributions of the underlying processes of color vision must be changing most rapidly (WRICH’I’, 1947). WRIGHT(1947) in a volume summarizing 20 years of his research on normal and defective color vision said about the hue disc~~nation function: The curvesas a wholehave the characteristicsthat mightbe anticipatedfrom a qualitativeexaminationof the spectrum.The part of the spectrumwhere a minimumexistsmust obviouslyoccur where there is a rapid change of hue; thus in the yellow where the colour turns redder on one side and greener on the other, in the blue-green where it turns bluer on one side and greener on the other, in the violet where it becomes redder or bluer, minimum steps would be expected.

The hue discrimination function may thus locate interhue transition points, and identify the number of hues the organism perceives. The present experiment was conducted to obtain a hue discrimination function for the pigeon and to determine whether or not minima of the hue discrimination function would coincide with those spectral regions of hue transition identified from slopes of wavelength generalization gradients (GUTTMANand KALISH, 1956; BLOUGH,1961) and from the intersection of the pigeon’s color naming functions (WRIGHT and GUMMING,1971). If the relation between color naming and hue discrimination is found to be the same for other organisms as for humans, minima of hue discrimination functions can be assumed to mark transitions between hues in organisms for which color naming data are not available. Related to the question of whether or not discrimination is best at color boundaries is the question of whether or not di~rimination is the inverse process of generalization. Do subjects generalize their responses to the degree that they cannot discriminate among the stimuli ? GUTTMANand KALISH(1956) obtained wavelength generalization gradients from pigeons and compared them to the pigeon’s hue discrimination function (HAMILTONand COLEMAN, 1 This research was based on an experiment submitted in partial fulfillment of the Ph.D. degree from Columbia University. The author is grateful for the suggestions and encouragement from Professors John A. Nevin, Bruce A. Schneider, Robert N. Lanson, and the late William W. Cumming. Support of the research was provided in part from Mental Health Grant MH 18624 to John A. Nevin, principal investigator. * Present address: University of Texas, Graduate School of Biomedical Sciences, 6420 Lamar Fleming Blvd., Texas Medical Center, Houston, Texas 77025, U.S.A. 1447

1448

ANTHONY A. WRICM

1933). Regions of the spectrum where there was feast generaIization did not correspond to spectral regions of best hue discrimination. Transition points between pigeon hues, as determined from color naming functions (WRIGHTand GUMMING,1971) and from slopes of generalization gradients (GIJTTMANand KALISH,1956; BLOUGH,1961), were likewise not those spectral regions where the pigeon was shown to have ‘best hue discrimination (HAMILTON and COLEMAN,1933). As BLOUGH(I 961) has said, “. . . the need for more discriminability data is pressing”. METHOD Subjects

Subjects were four white Guneaux pigeons (Columbia livia) from the Palmetto Pigeon Plant, Sumter, South Carolina. Birds 285 and 287 were 7 years old, bird 286 was 8 yrs old, and bird 53 was 11 years dd at the begin&g of the oxperkat. Bird 53 had previous’experience mat&& monocluornatk stimuli (WIIOHT and CUMMIM, 1971); the other subjec& wem experimentally naive. Experimental -ions were conducted seven days per week if the subjects were 77-83 per cent of their free-feeding weight. Apparatus Optical system. The optical system which produced the bipartite stimulus is shown in Fig. 1. The light source (1) was an Ostam XBO 150 W/l xenon arc lamp. Two beams of light were taken from the source to form the separate halves of the bipartite stimulus. Light from the source passed through i&a-red reflectors (2.2’) and heat absorbing glaaaea (3.3’). It was then collimated (4,43 to form separate beams of Wt. After being rclkctcd from front surface mirrors (S,Y), these collimated beams of light passed through Alter boxes (6.6‘) containing polar&s, neutral density filters, and blocking filters. Then they passed tbrougb Bausch k Lomb series No.U78 intieraza liters (7,9).The multing monochromatkbeamswae unltal by the front surface mirror (IO) to form the split &Id. A solenoid operated device (8, of Fir. 1) allowed tha radiance of the right &Id half to be varied automatkalky by actuation of individual channels containing Kodak Wratten neutral density filters.

Chamber wall__... __,, _____ _______

/ FLeed;: Ll__ Stimulus panel

FIG. 1. Schematic of optil system for producing a bipartite stimulus. (1) source, (2.2’) infra-red rdktors, (3.37 beat aghgcp, (4.41 knaf, (s.53 mirror4 (6.61 filter boxes with p&&en, (7) intertkraps m ti rdmacs wav&ssgth. (8) daadc)r fll# actuator, (9) interference Mar for compa&un Wrvaacagth, (10) mirror, (11) lens. (12) Stmttet, (13) lens, (14) screen. (15) aperture.

Psychometric

and Psychophysical

Hue Discrimination

Functions for the Pigeon

1449

The collimated light was condensed by lens (11). and the edge of the front surface mirror (10) was focused by a photographic triplet (13) onto the ground glass screen (14). Two pieces of ground glass, placed so that their ground surfaces were adjacent, were used as a screen. Placed against the front of the screen was a O-2500 in. (6.350 mm) aperture (15) which limited the diameter of the bipartite stimulus. The screen was located 2.93 in. (74.4 mm) behind a color-clear glass pecking key, and the visual angle of the bipartite stimulus was approximately 3”14’ of arc when the pigeon’s beak was against the front surface of the pecking key. Distance from the tip of a pigeon’s beak to its eye is approximately 1.50 in. (38.1 mm). Bausch & Lomb series No. 44-78 interference filters passed three modes of energy: one in the infra-red, one in the visible, and one in the ultra-violet. The ultra-violet mode was eliminated by blocking filters placed in the filter boxes. The wavelength of the ultra-violet mode determined which blocking filter (Kodak Wratten No. 2A or No. 15 filter) was used. The infra-red reflectors (22’) eliminated the infra-red mode of energy. Wavelength of monochromatic light was varied by rotating the interference filters. Polarizers were placed in the filter boxes (6,6’) so that large angles of incidence could be employed with no band pass distortion (HEAVENS,1965). The interference filter (9) was rotated by joining the mounted interference filter to a precision gear reducer (Planet No. A-l 13) and the gear reducer to a precision stepping motor (Superior Electric No. HSSOE). Backlash in the rotational system was eliminated by a weight applying constant torque counterclockwise to the shaft of the interference filter mount, and by all angular positions being approached by nine steps in a clockwise direction. No variability in wavelength produced by the system could be detected with the calibration equipment available. Experimental chamber. The subject’s portion of the experimental chamber was 12 in. (304.8 mm) long x 14.8 in. (375.9 mm) wide x 13.8 in. (3505 mm) high. A stimulus panel formed one wall to which was attached a grain feeder with a 2 x 2 in. (50.8 x 50.8 mm) opening located 5.8 in. (146.1 mm) below the center key. Also on the stimulus panel were three horizontally aligned color-clear glass pigeon pecking keys (modified Lehigh Valley No. 121-16).,. 9.8 in. (247.6 mm) from the chamber floor and snaced 2.50 in. (63.5 mm) as measured from center to center. The pecking keys were positioned behind 1 in. dia. holes in the stimulus panel. In order to close the microswitch of each pecking key, 20 g of force had to be applied through a distance of 1 mm. The side key stimuii were @50 in. (12.7 mm) achromatic circles from IEE in-line units positioned 3.50 in. (88.9 mm) behind the colorclear glass pecking keys. The chamber was lighted by four GE No. 1820 bulbs mounted in the ceiling. A 6.0 x 6.0 in. (152 x 152 mm) piece of opal glass diffused the light from the bulbs and produced an illuminance of 0.63 ft-L (2.16 cd/m’) on the gray chamber walls as calibrated with an Ilford SE1 photometer. A fan mounted in the chamber wall opposite the entrance door provided ventilation and masking noise. The calibration instruments could be positioned in front of the bipartite stimulus by removing the chamber wall opposite the stimulus panel. Calibration of the bipartite stimulus Wavelength calibrations. An Edgerton, Germeshausen & Grier 580-585 spectroradiometer was used to calibrate wavelengths. The stimulus panel and ground glass screen were removed during wavelength calibrations and a platform positioned the spectroradiometer in front of the stimulus. A scanning method was used to determine peak wavelength for each position of the interference filter (9). The peak wavelength was defined as the monochromator setting (k O-1 nm) which produced the greatest response. Four determinations of peak wavelength were made at each position; two of them by rotating the monochromator grating in a clockwise direction, and two of them by rotating it counterclockwise. These individual determinations were usually within f 0-l nm of each other. Peak wavelengths were calibrated before and after data collection at each reference wavelength.The means from the two calibrations typically were equal within the resolution (& 0.1 nm) of the calibration equipment. Means from the two calibrations contributed equally to the fmal estimate of peak wavelength. Mean wave-number was the mean of the reciprocals for the individual wavelength calibrations. Slight differences between the wave-number shown for a particularpeakwavelength and the reciprocalof that peak wavelength might occur because the mean of the reciprocals is not equal to the reciprocal of the mean. Wavelengths of the two field halves were equalized by moving the interference filter (9) in discrete steps until its wavelength was as close as possible to the desired reference wavelength. Then the wavelength of filter (7) was adjusted to equal the particular value of (9). Interference filter (7) was mounted on a Mica (640-A) rotary table and its adjustment was continuous, as opposed to (9) which was adjusted in discrete steps. When monochromatic light from filter (7) was calibrated, the front surface mirror (10) was moved by its rack-and-pinion mount so that the only light incident upon the spectroradiometer came from this filter. As an added precaution, the collimated light beam to filter (9) was blocked as well. In a like manner during calibrations of monochromatic light from filter (9). mirror (1O)‘was moved out of the beam passed by filter (9) and the light to filter (7) was blocked. Being able to move mirror (10) was particularly useful when calibrating radiance; otherwise the value of radiance would depend on the accuracy to which the split field could be divided. Radiance calibrations. The radiance of each part of the bipartite stimulus was calibrated with an Edgerton,

1450

ANTHONY A. WRIGHT

Germeshausen & Grier 580 radiometer. A platform was used to position the radiometer pm&&y 4.75 in. (120.6 mm) in Front of the ground glass semen. The stimuIus panel was removed from the chamber and a 0.50 in. (12.7 mm)aperture was placed in front of the ground glass screen. Mirror (10) was positioned so thatonly one-half of the bipartite stimulus was incident upon the screen at any one time. Radiance calibrations were performed each time wavelength calibrations were made. The stimuli were calibmted and adjusted with Kodak Wratten neutrat density tihers to be photometrically equal (for the pigeon) to a 560 nm stimulus at 24C~x 10’aW/cm2. This value of irradiant power is specific to the position of the radiometer from the dhfusing screen. The 560 nm stimulus, at 240 x IO-‘@W/cm2,was 8.6 mL (27.2 cd/m? as calibrated with an Ilford SEE photometer. Desired radiometer readings for other tohecor&ed wz&ei>hs were then computed. The radiometer reading at the 560 nm refer-e&e was corrected for the Spectral response of the radiometer and for the light adapted pigeon (Btouc~, 1957). The corrected 560 nm reading was set equal to the radiometer reading (unknown) at the to-be-corrected wavelength, multiplied by the spectral response of the radiometer and thecoetlicient for theligbt adapted pigeon(&ouc%r, 1957)at this to-be-corrected wavelength. The equation was then solved for the unknown radiom&er reading. Next, the radiometer was placed in front of the stimuIus and density (Kodak Wratten neutral density @tens) was added to the Iii beam until the desired radiometer readhtg was obtained. The obtained radiance was at least within f 042 log units of the desired value. Wavelength differences of one or two nanometers were frequently used during the experiment; at these smaflwavelength differences, photometric equality is identical to radiometric equality. Therefore, at small wavelength differences the\ accuracy to which brightness of stimuli was equated depended upon the accuracy of calibration equipment and not the pigeon’s ~rn~i~ coefficients. Density needed to equate several stimuli photometrically was compared to the density indicated by a variant of a more traditional procedure. The Edgerton, Germeshausen and Grier 580-585 speetroradiometer was used to calibrate successive 5.0 nm bands of the monochromatic stimulus. Slit widths of the spectraradiometer were set for a 5Onm band pass. These calibrations were adjusted for the spectral response of the spectroradiometer and for photopic sensitivity of the pigeon (RLOUOH,1957). The resulting energies were then summed, using a trapezoidal rule. Density values for the two methods were similar, but the former method of “‘looking in” with the radiometer and recording the gross response of thesystem was found to be more reliable. The split field was halved by moving the mirror (IO) until the field appeared to be equally divided. A rack-and-pinion drive moved mirror (10) along its axis so that it did not change its a&e to the light beam. The angle of mirror (10) to the light beam from filter (7) was set so that the two halves of the split &Id were separated by a thin dark line 05 mm wide, as calibrated with an optical comparator(Edmund Scientlhe No. 30,585 6 x). PROCEDURE

Final procedure The subjects’ task and the scheduled consequences are d&rammed in Fig. 2. A trial began with the onset of the @25 in. (6-4 mm) bipartite stimulus. A peck on the giass disk in front of the bipartite stimulus closed a microswitch, turned on side key stimuli, and allowed side key pecks to produce 3 set access to grain or an-8 set intertriai interval. Right side key pecks were correct when the two field halves differed in wavelength, and left side key pecks were correct when the two halves were of equal wave~en~. Correct side key choices were occasionally followed by access to grain; reinforcement probabilities generally were changed from session to session to manipulate side key bias. Unreinforced correct side key pecks were followed by a 0.41 set gash of the feeder Iight. Reinforcement or feedback (feeder light flash) was folfowed by an 8 see intertrial interval. All incorrect side key pecks (either right or left) were followed by a O-38 set extinction of the overhead chamber light, and then an 8 set intertriai interval, Generally, reinforcement probabilities fcir the two side keys were varied so that their sum was 0~40. Re~ur~ment pro~b~~~~ were programmed by tape readers, one for each side key. The sequence for each probabihty was drawn from a random number table (RAND, 1955) and run lengths were adjusted according to binomial probabilities to yield geometric distributions. Each of the six bipartite stimub shown in Fig. 2 was presented for l~-t~~s in mixed order within a session, The comparison wavelength was either shorter than the reference

Psychometric and Psychophysical Hue Discrimination Functions for the Pigeon

P(SR)>O-+

RK

CK

LK

x,

x, x0

1451

1

Xl i

FIG. 2. Schematic of response options and ~infor~ement conditions for hue disc~mination procedure. ho is the reference wavelength and r\& are comparison wavelengths which are shorter than the reference wavelength.

or equal to it. Wavelengths of the comparison stimuli depended upon the subjects’ performance. At each shift to a new reference wavelength, differences were made initially very large, thenprogressivelyreduced within thefirst few sessions. After performance stabilized, the largest difference selected was one where the subject’s performance was just short of perfect, and the smallest difference was one where performance was just above chance. Fifteen sessions were usually conducted at each reference wavelength. Data from the first five sessions were not used in the analysis. During the last two sessions, the radiance of one comparison stimulus was varied; in the next to the last session it was increased 0.3 of a density unit, and in the final session it was decreased by 0.3 of a density unit. Discrimination performance was determined every 10 nm. The reference wavelength was initially 570 nm. It was increased in IO nm steps to 660 nm, changed back to 570 nm, and then decreased in 10 nm steps to 470 nm. wavelength,

Preliminary procedure

Before data presented for this experiment was collected, the subjects had been given extended discrimination training. Approximately 350 preliminary sessions were conducted with each subject. Initially only two comparison stimuli were used at a reference wavelength of 5550 nm. In the presence of the bipartite stimulus 555G-535.0 nm, right key (different) responses were correct. In the presence of the bipartite stimulus 5556 555.0 nm, left key (same) responses were correct. Fifty trials of each type constituted a session. The subjects were reinforced for each correct side key choice; they were blacked-out (all chamber lights off) for each incorrect side key choice, and all trials were separated with a 15 set intertrial interval. After approximately 30 sessions, reinforcement was made intermittently available for correct choices Cp = 0*20), the intertrial interval was reduced to 8 WC,and the session was lengthened to 500 trials. Preliminary assessment of discrimination performance was made at reference wavelengths of 555.0, 54&O, 525.0 and 570.0 nm. At 555.0 and 5409nm di~~~nation performance was determined for wavelength differences presented singly within a session and in combina~ons of five. Performance when only one wavelength difference was presented within a single session was found to be comparable to performance to the same differences when five of them were presented within the same session. Twelve subjects were initially trained. Eight were eliminated from the experiment before the final procedure; six subjects never acquired appropriate identification of the stimuli, and two were inordinately slow to reacquire performance at new reference wavelengths.

1452

APUHONY A.

WRIGHT

RESULTS Analytic method Wavelength discrimination was asseised by manipulating bias and extracting a bias free discrimination index (d’) with the analytic methods of signal detection theory. Bias was manipulated by changing the reinforcement probabilities. An increase in reinforcement probability for correct left-key “same” responses relative to correct right-key “different” responses would increase the subject’s bias toward making “same” responses. Such bias change would be shown by the data points being closer to A’ along A-A’ in Fig. 3. For an easier discrimination, one where the difference in wavelength was ‘greater, manipulation of the bias might map out the function from B to B’ in Fig. 3. When data from such bias manipulations are plotted on normal-normal coordinates (as in the lower panel of Fig. 3), the resulting functions are usually linear (CLARKE, BIRDSALLand TANNER,1959). The discrimination index d’ can be computed for each data point by subtracting $he normal deviate value ofp (“di&rent”/wavelength difference) from the normal deviate value

:r

A

B

so -

:;I

70

/’

/’

B’

50

z El & =r 5

/‘ /’

/’

30

1‘

n’

IO

/’

,’

,’

I’

,’

/’

/’

I’

I’

I.

I’

/’

I’

,.

/’

,.’

P L”different”/no wovelength differmel

I

2

J.’

.’

./

0

-I

-2

.’ ,-2 5

IO

P (“diffemt”/no

30

MZO

90 95

wawlength difference)

FIG. 3. Hypothetical isosensitivity curyes for a di&ult discrimination A-A’, and for an easier one B-B’, They are plotted on iii co&iaatcs in the uppst panel and on normalnormal coordinates in the lower panel.

Psychometric and

Psychophysical

Hue Discrimination

Functions for the Pigeon

I453

of p (“different”/no wavelength difference). If the linear function (on normal-normal coordinates) is of unit slope, then the value of d’ at all points along the line will be equal. It is not unco~on, however, to find isosen~tivity functions with slopes less than one (GREEN and WETS, 1966). For functions which are not of unit slope, d’ wiIl vary depending upon the particular point selected. When a point must be chosen to compute the discrimination index (d’) the one frequently chosen (and the one used in this study) is that point where the linear function crosses the negatively sloping diagonal (CLARKEer al., 1959; EGAN, SCHULMAN and GREENBERG, 1961; POLLACK, 1959). EGAN and CLARKE(1966) have noted that day-to-day variability in isosensitivity functions is manifested by changes in slope, while the index d’ at the negative diagonal remains relatively constant. Isosensitivity functions Figure 4 is a typical example of isosensitivity functions obtained from this experiment3 Figure 4 shows five isosensitivity functions (from bird 287) plotted on normal-normal Normal deviote 2

99

I

0

_I

-2

2

I

0

-I

-2

95 90 70 50 30 In

1 I’.I I

/ I i !

Itcorrect

/) ,, I ,..

rfght

I

I

key chotces,

:b

FIG. 4. Isosensitivity functions for bird 287 for five comparison stimuli at a reference wavelength of 64@4 nm. Arrows over data points indicate 100 per cent correct. Unfilled square is for a decrease of 0.3 density unit of the 629-6 nm comparison stimulus. Untilled triangle is for an increase of O-3 density unit of the 629.6 nm comparison stimulus. Untilled circles are from sessions where the only wavelength difference was produced by a comparison stimulus of 631.4 nm. 3 A limited number of copies of isosensitivity functions at each reference wavelength together with the raw data in tabular form are available upon request to the author.

1454

ANTHONYA. WRIOHT

coordinates. In all five cases the reference stimulus was 640.4 nm. The most difticult discrimination is shown in the upper left hand panel for the 635.4 nm comparison stimulus to be discriminated from the 640.4 nm reference stimulus. The easiest discrimination is shown in the lower right hand panel for the 627.8 nm comparison stimulus to be discriminated from the MO.4 nm reference stimulus. One point in each of the five panels was obtained from each session. The proportion of incorrect right key choices is necessarily the same for all five wavelength differences and so the data points from each individual session have the same abscissa value. Discrimination data for radiance changes in the 629.6 nm comparison stimulus are shown in Fig. 4 by the unfilled triangle and the unfilled square. Changing the radiance by Bird

3;

I+

.’

660

i:_IO

20

30

21 660.1

“m

I

40 .’

. *

6497nm

10

; /I

20r30

4nm* 20

.

.50

60 t

: : ! ,,’ 640

4 “m

. -

'5&5nm

10

20

L....539+

L__

--

:*

50

:

. ~‘560 IO

40.

I rim

L-L--_._

L.

3r

.,

.

2j 31

267 _,’

30

IO

nm

20

3.

2:

2:

.

:

*

.

i3:

‘5300nm IO

20

-20

’ 10

5200nm 20

3ci

2’

Wave-number

difference,

mm-’

FIG. 5. Psychometric hue discrimination functions for birds 53, and 287. The discrimination index (d’) for each data point is a measure of the subject’s ability to detect the dWannce between that comparison wave-number (recipmcel of wave-h) and the refm wavenumber.

Psychometric and Psychophysical Hue Discrimination Functions for the Pigeon

14.55

f 0.3 density units did not affect discrimination performance (at 640.4 nm reference wavelength) which suggests that at 640.4 nm discriminations were not being made on the bases of brightness differences. Filled data points in each panel of Fig. 4 were fitted by Ftraight lines drawn by eye. Generally the closeness of fit to the straight line diminishes somewhat for increasing ease of discrimination. This is not an unexpected finding; the binomial variance, an underestimation of the actual variance (GREEN and SWETS,1966), increases with closer approximations to the upper left hand corner of the normal-normal square. Once discrimination performance to the five comparison stimuli was tested {filled data points), six sessions were conducted where the bipartite stimulus was either 640*4-640-4 nm or [email protected] nm. The unfilled circles are within the range of the filled ones, indicating that there were no prominent interaction effects from testing discrimination performance to all five comparison stimuli in the same session. Following hue discrimination testing at the first reference wavelength (569~4 nm), it was progressively increased in IO nm steps to 660 nm. It was then changed back to 569.4 nm, and discrimination performance retested. Birds 287,286 and 53 showed no change in discrimination performance over the I1 month period, suggesting that their discrimination performance was probably asymptotic. Bird 285 showed a slight improvement over the 11 month period. Ps~~~~~etr~c ~~s~r~~i~~tion~~~ti~~~

The discrimination index (d’) was calculated for the point where each isosensitivity function intersected the negative diagonal. The normal deviate value of the horizontai projection was subtracted from the normal deviate value of the vertical projection (EGAN and CLARKE,1966). Each d’ value was then plotted as a function of the stimulus difference used to generate the isosensit~vity function. One such function per reference waveIength was obtained for each subject as shown in Figs. 5 and 6. Lines were fitted to data points in Figs. 5 and 6 by a least squares method. Straight lines appear to provide a good fit to the data, and the correlation coefficients (r) were usually greater than O-99. The slope of the lines indicates the rate at which ‘discriminability changes with changes in wave-number difference between the two halves’ of the split field. A steep slope, e.g. at 600 nm, means that dis~minabjlity increases vel y rapidly for increases in wave-number difference between the two field halves. At 660 nm. the relatively shallow slope means that discriminability increases only slowly with increases in wave-number difference between the two field halves. Differences between the reference stimulus and the comparison stimulus are scaled in wave-number difference units rather than in the more conventional scale of wavelength difference (/x_A).A scale of wave-number differences was chosen becausewave-number, which is a constant proportion of light frequency, is invariant over all media whereas wavelength changes.4 4 Frequenzy (14 and wavelength (A) are related by their product being equal to the speed of light (c): VA= c. As light enters denser media, its speed decreases, and so the product (VA)must also decrease. Either A must decrease, or Y must decrease, or both must decrease. Frequency invariance is dictated by the enerw relations of light. The energy (4 per quantum is equal to rlz;k; constant (h) times the frequency (v): If the frequency of light were to change upon entering each new medium, then the energy per quantum would also have to vary according to the medium. This obviously is not the case. Frequency cannot be directly measured,it must be calculatedby dividinglight speed by wavelength. But light speed cannot be determined as accurately as wavelength, and the resolution of frequency needed for most applications would involve more than the possible number of significant digits. Therefore it is more pragmatic to use the r&procal of wavelength, wave-number, rather than frequency.

ANTHONY A. WRIGHT

1456

Bird

3;

266

:[

*-*

Bird

285

.

; I

I , _x’ 2

IO

,20

660 30

40

50

63 ,/”

I nm 70

60

, _,. ” 2: /

J ‘- 6497nm” &---.--IO 20 30 40 50 60

:t

. :‘,” ,

3i

I

I 10 20

30

40

I

.

IO

2G

3s

30

40

Wove-number

,. L-_-i

I ;

-*

649

7nm

____-

IO

20

30

IO

20

30

difference,

40

50

~‘6&4mn

I60

IO

P

IO:

20

30

40

‘.

30

mm-’

FIG. 6. Psychometric hue discrimixWion functions for bii 286 and 285. The discrim&&on indw; (d’) for each data point is a meas? of the sub&t’s ability to detect the dS&ence meen that comparison wave-number (rz;gl of waveiemh) -and the reference wave-

Psychophy&al discrimination functions: constant d’ contours The linear functions of Figs. 5 and 6 were intersected at d’ = 2.0, and the wave-number dilkences at these intersection points were plotted in the lowest panel of Fig. 7 at the reference wave-number. The dot-dashed line connects the me811 points at each reference wavelength; at 590,600, and 610 nm onIy the data from birds 287,286, and 2Q15contributed to the mean. These psychophysical functions in the lower panel of Fig. 7 are equal discriminability contours because the wave-number di&rences are those di&r@zes which, if presented to the subject, would result in the same discrimination index (d’). The functions for the four

Psychometric and Psychophysical Hue Discriminalion Functions fur the P&eon

3457

Wavelength A, nm 500

550

subjects are of the same absolute value and of the same general shape. Troughs (or dips) in the function (600~ 540 and 500 nm) indicate relatively good discr~m~~abi~ity. Bird 286 (f&d squares) shows the most accentuated peaks and troughs with &early defined minima at 500,550 and 600 nm. 3ird 287 (&fed triangles) with a very smooth function shows the same trends and locations of the minima as does bird 286, afthough the trough to peak differences are less than they are for bird 286. Bird 285’s friction (Etlled diamonds) is more irregular than is 286% or 287’s, and the 530 maximum is questionable as it is an isolated, elevated point. Bird 53’s ftmction (@Ied circles) stops at 510 nm beoause this (X2-year-old) subject developed a cataract, The anom~o~s points for 590, 600 and 610 nm seem to be due to a progressive dete~ora~o~ of bird 53’s performance at these ispetiral points, in spi#e of the precautions to provide conditions conducive to reacquisition at new reference wavelengths. Reacquisition at 620 nm was possible by making the wavelength difference initially very much larger for bird 53 than for the other subjects. The slopes of tile ps#xomettic fwztions are skwn in the mid&e panel of Fig 7, and are plotted as a function of reference wave-number. The slopes are pbtted on a descending

s&e, so that dips in the slope function will correspond to dips in the equal 6” function (lowest panel of Fig. 7). The slope function shows virtually a one-to-one correspondence to the equal d’ function. One difference is that the slope function shows a minimum in the 550 nm region as opposed to the 540 nm region. There is less variability among subjects Y,R.12?9-a

1458

ANTHONY A. WRIGHT

and the trends are smoother for the slope function than for the equal d’ function. A SimuItaneous Test Procedure (SOKAL and ROHLF, 1969) was employed to test for differences among the set of regression coefficients for each subject. Differences between each minimum (600,550 and 500 nm) and m~imum (660,570,530 and 480 nm) of the slope functions were highly significant at the p < 0.01 level for each subject. The uppermost panel in Fig. 7 shows the intercepts of the psychometric functions, Intercepts do not show any discernible trend related in any simple way to the slope function or to the equal d’ function, and the intercept values are small relative to the largest wavelength differences employed at each spectral point. A slope function may be more informative and appropriate conceptuahxation of hue discriminabihty than an equal d’ contour. By assuming that nonzero intercepts of the psychometric functions are artifactual and that each function passes through the origin, then hue discriminability can be entirely summarized by a slope constant Cm): discriminability

(d’) = m x wave-number difference.

The slope of the psychomet~c function represents the growth of d~~ri~abil~ty at each spectral point. A steep slope means that the hue difference increases very rapidiy for increases in the difference in wave-number. A shallow one means that the hue difference increases only slowly with increases in the difference in wave-number. The slope values may be used in order to select a set of wave-numbers (or wavelengths) which have an equal psychological spacing for the pigeon. Table I gives the mean slope for each reference wavelength. TABLE 1. MEAN SLOPE AND

WAVE-LUBBER DIFFERENCE AT EACH REFERENCE WAVE-NUMBER(l/A)

slope

l/A in mm-* (Ain nm) -

_ ___ ___ __

1/A diff. (A diff.) at d’ = 2.0

l/h in mm-’ (A in nm)

slope

--.~ -

1514-9 (660.1)

0.049

45.8 (19.4)

1784.5 (560.4)

0.105

1539.2 (649.7)

O-052

45.6 (18.7)

18168 (5505)

O-129

1561*5 (640.4)

o-102

27.1 (1 l-2)

18524 (539.8)

0.113

1586.5

0.104

20.4 (8.0) 13-o (5.1) 10.4 (3.9)

188&7 (.530*0)

O-099

1923-l (520~0)

O-105

1962.4 (509*6) 2002*0 (4995)

0*106

2W4 (490-l)

0~100

2083.1 (4tW1) 2125-l (470.6)

0.071

(630.3) 1612-i (620.4)

0.166

1639.0 (61O-2)

0.218

1666.6 (600*1)

O-297

1693.5 (590.4)

0,165

1723.2 (580.3)

O-ill

1756-2 (569.4)

0.090

(“21:) 11-6 (3.9) 22-o (7.3) 19-3 (6.2)

o-141

o-073

1/A diff. (A diff.) at d’ = 2-o 19.3 (60) 17.9 (5.5) 15.8 (4-5) 21.0 (5.8) 19-o (5-l) 19.7 (5.1) 17-I (42) 22.4 (5-3) 29.5 (6.8) 27-O (5.9)

Note: Mean slopes and wave-number differences for reference wa~~m~~ of 4706, W*l, 49&l, 4995,590-4,6Ot&l and 610.2 nm are means for three subjects: 286,287 and 285; means for other reference wave-numbers are means for ah four subjects.

Psychometric and Psychophysical Hue Discrimination Functions for the Pigeon

3459

DISCLJSSION In this experiment the ~s~r~~na~on indices (d’) were found to be linearly related to wave-number difference at each spectral point. Extrapolations of the linear functions generally showed zero intercepts, indicating that huediscri~nabi~ityis a ratio scale of wavenumber difference. An absolute threshold theory would predict that changes in the intercept, rather than changes in the growth of discrimination (slope) would result from changes in reference wavelength. The concept of the absolute threshold assumes some wavelength difference which cannot be ~stinguished (d’ = 0.0) from physical equality. As BIRCHand Wmsr (1962) said in their review of the literature on human hue dis~rimina~on~ “Their wavelength difference [between the two field halvesJ must have a certain finite value before the difference in their colour can be detected.” In spectral regions where the hue discrimination is relatively poor the undetectable wavelength difference would be greater than in those regions where it is better. On the other hand, there may be no absolute threshold (WET.%, 1961). That is, ~sc~~na~on performance may be ordered for ah wave~en~h differences as shown by zero intercepts for the psychometric functions. The psy~hophy~cal hue dis~ri~nation functions (equal d’ function and slope function of Fig. 7) showed that hue discrimination was best at 500, 540-550 and 600 nm spectral regions. Results from this experiment do not corroborate the findings of HAMILTON and COLEMAN (1933). HAMILTON and COLEMAN (1933) obtained a hue discrimination function for the pigeon using a Iashley jumping stand. Their results and the comparable psychophysical function from this experiment are shown in the upper two panels of Fig. 8. The Wavelength A, nm

2100

a300

1900

1800

Wave-,number !$ ,

no0

IS00

mm-’

FIG.8,Top panel: Pigeon’s hue discrimination function (~M~TG~ and COLEMAN, 1933). Middle panel: Pigeon’s hue discrimination function (present experiment). Bottom panel: Cob-naming functions from a matching-to-sample procedure (WRIGHTand &tdMING, 1971).

1460

Arrrrio~~ A. WR~OHT

hue discrimination function from I-~MILTON and COLEMAN (1933) shows a minimum (best discrimination) at 570-580 nm regio,n, which is the spectral point where there is a maximum in the function from the present experiment. Hamilton and Coleman’s function shows a rather broad m~mum in the 540 nm region, which is the spectral point where there is a minimum in the function from the present experiment. Only for short wavelengths (< 510 nm) do the two functions show general agreement with each other. Several peculiar aspects of %WILToN and COLEMAN's (1933) data may indicate reasons for the discrepancy between their results and the results from the present experiment. HAMILTON and COLEMAN (1933) trained three pigeons in a Lashley jumping stand with two doors transilluminated with monochromatic light from a prism monochromatar. Radiance was adjusted for sensitivity of the ‘pigeon’s eye according to pupillary measurements (LAURENS,1923). Two pigeons were trained to jump to the door of shorter wavelength, the third to the door of longer wavelength. Jumps to the correct door allowed the pigeon to enter; jumps to the incorrect door resulted in no entry and the pigeon fell into a net. The wavelength difference between the two doors was initially very large and was reduced progressively by increasing the shorter wavelength until the pigeon made an incorrect choice. The wavelength dif%rence to which the incorrect choice was made was the discriminable step (A A), plotted as a function of the reference wavelength (the longer of the two wave~n~s). The authors summarized their procedure for exploring the wavelength continuum as follows: The spectrum was gone over ‘continuously’ that is, if the pigeon were just able to distinguish between 670 and 620 rnp (nm), one would beginning at 620 nyl (nm) determine the next least perceptible diierence between 620 and 600 w (nm), say, and the next between 600 and 590 w (nm), etc. Thus all parts of the spectrum between the points 700 and 460 rn@(nmf were survey&i. (HAML.T~Nand COLEMAH, 19331. Discussing

their results, HAMILYDN and COLEMAN(1933) noted:

A bird (NOS.1 and 3) which had been jumping to the shorter of two lights consistently at all wavelengths above 530 nq (nm) will, when presented di~~~nab~e wa~l~gths below 530 mtc b@ Jump at the iorqper. The second pigeon, trained oppositely, reversed his behavior in an analogous fashion at the pair (5@0490 ml*). This peculiar result cannot be easily reconciled with the subjects’ hue discrimination function. If the pigeons, trained to jump to the shorter of the two wavelengths, “consistently” jumped to the longer one at wavelengths below 530 nm, then at wavelengths of 530 nm or less the subjects should be “consistently” incorrect, and hence their discriminable step should be irately large. But the di~~minable step is not infinitely large. In fact, for wavelengths somewhat shorter than 530nm, the discriminable step decreases and discrimination is shown to improve. It is not clear whether the experimenters redefined a colreet response as the longer of the two wavelengths, or whether there might have beena prolonged rea~u~tion phase so that the subjects would once again “con~stently”j~p to the shorter wavelength. In any case, it is the subjects’ reversal at 530 nm which is of interest. Another peculiarity of the ELTON and COLEMAN(1933) discrimination function is found by computing the value of the comparison wavelength for ref&ence wavelengths 620-700 nm. Comp~tion reveals that the comparison waviest remained the same (608-612 nm) for all of the reference wavelengths in the range 620-750 nm. This means that discrimination was independent of the wavelength value of the reference stimulus. IIAMILToN and COLE&N’s (1933) procedure did not allow the subjeots to make a “same color” versus “different color” response as did the procedure in the pi%@nt-experiment. Rather, the subjects were trained to jump to the shorter wavelength do&, 3ut what

Psychometric and Psychophysical Hue Discrimination Functions for the Pigeon

1461

does it mean to select the shorter of two wavelengths? It is only from a knowledge of the physical nature of light that yellow is shorter than red, or green is shorter than yellow. HA~LTON and COLEMAN (1933) may simply have trained their subjects to jump to a particular hue. When the wavelength of the compa~son stimulus was increased above 600 nm, the subjects made errors. When the wavelength of the reference stimulus was decreased below 540 nm, the subjects made errors. Thus, either the reference stimulus or comparison stimulus (or both) had to be in the range 540-600 nm for the subjects to play the game correctly. These wavelengths 540 and 600 nm may he the bounda~es of a pigeon hue. These are the boundaries of a hue identified by WRIGHT and GUMMING(1971) from the pigeon’s color-naming functions. WRIGHT and GUMMING (1971) trained pigeons to match-to-sample. In matching-tosample the subject’s task is to select the side key stimulus which is identical to the center key stimulus. The subjects were tested by presenting test wavelengths on the center key and allowing the subjects to choose between two of the training stimuli on the side keys. For each test wavelength the proportion of choices to each of two possibie training wavelengths was tabulated, and the mean proportions for the six subjects in the experiment are shown as filled data points in Fig. 8. Test wavelengths are shown on the abscissa, and wavelengths of the three training stimuli are indicated in Fig. 8 by filled arrows below the abscissa. For a second test, training wavelengths were changed as shown by the position of the unfilled arrows below the abscissa in Fig. 8. Intersection of these gradients (unfilled data points) occur at the same spectral points (600 and 540 nm) as did the intersection of the gradients from the first test. If the subjects were matching the stimuli per se, then a change in the wavelength values of the training stimuli (e.g. a uniform decrease of 20 nm) ought to result in intersection points some nanometers less than the former ones. Changing the training wavelengths, however, did not result in a change in intersection points. The subjects seem to have been matching hues rather than wavelengths, and 540 and 600 nm are the transition points between three pigeon hues. These hue transition points are essentially two of the three minima (600,550-540 and 500 nm) of the hue discrimination function from this experiment. The correspondence of hue transition points to the minima of the hue discrimination function corroborates the conjecture by WRIGHT (1947) that the part of the spectrum where a minimum exits “must obviously occur where there is a rapid change of hue.“5 Color naming functions are generalization gradients where presumably the color name generalizes freely within a hue, but very little across hue boundaries. It was suggested (LASHLEYand WADE, 1946; GUTTMANand KALISH, 1956) that subjects’ responses generalize to stimuli to the degree that they cannot discriminate among them, that generalization is the inverse process of discrimination. GUTTMANand KALISH(1956) trained pigeons to peck a disk transillumina~d with mon~hromati~ light, and conducted gene~li~tion tests in extinction. They compared the slopes of the resulting generalization gradients to HAMILTON and COLEMAN’S(1933) hue discrimination function, but they found no simple relation between the slopes of their gradients and minima of the hue discrimination function. If subjects generaiize their responses to the degree that they cannot discriminate among the stimuli (inverse hypothesis) then steepest slopes of thegeneralization gradients (least generalization) should have occurred at those spectral points where hue discrimination was best. ’ NOintersectiOnPoint is shown co~~pondi~ to the 500nm minimum of the hue di~ri~nation fun+ tion. For an intersection point at 500nm one training stimulus shoufd be located in the ~00_540 nm wavelength range and another in the -C500 nm range. This was not the case for either test 1 or test 2.

1462

ANTHONY A. WRIGHT

It seems that controversy over the validity of the inverse hypothesis has been perpetuated through a dearth of adequate data on the-pigeon’s hue discriminability. Spectral regions where the slopes of CWTTMAN and KALISH’S(1956) and BLOUGH’S(1961) generalization gradients were steepest are those regions (500, 540 and 600 nm) where hue dis~rimina~on was shown (this experiment) to be best.$ So it seems that the present experiment. together with the experiments of WRIGHT and CU~WNG (1971), BLOUGH(1961) and GUTTMANand KALISH (1956), are in agreement that the regions of 500, 540-550 and 600 nm probably contain hue ~ansitions (color boundaries) for the pigeon. The inverse hypothesis has, in some cases, been accepted as a “given” rather than a theory to be validated. GRAHAM(1962) used the human hue discrimination function to derive hypothetical color naming functions. This derivation was based upon the inverse hypothesis; regions of best hue discri~nabi~~ were hypothesized to be regions where generalization of color names would be least. The general features of G-M’s (1962) color naming functions have since been confirmed by BEARE(1963); BOYNTON, SCHAFER and NEUH (1964); B~YNTONand GORDON(1965); STEKNHEIMand BOWW~N(1966); BEAREand SIEGEL (1967); LUMA (1967). The comparison between human data and pigeon data is particularly interesting because the boundary between two pigeon hues in the 540 nm region is located in the middle of a human hue. For increasing wavelength, the human color spectrum changes from blue to green at about 495 nm, and from green to yellow at about 565 nm (Bw, 1963). The 600 nm transition point for pigeons, likewise, does not correspond to the transition point from human yellow to human red, but there are several factors which make this, distinction less clear: (1) It is questionable whether orange should be given a status ~de~ndent from red; (2) Yellow is a narrow hue in terms of its wavelength range (565-585 me); (3) The minimum of the human hue discrimination function is rather broad for this spectral region (LAUREM and HAMILTON,1923; WRKE~, 1947; WLMLE,1951; BEDFORDand WYSZECICI, 1958). Researches relevant to the ph~iolo~ of the pigeon’s color vision are not in close agreement with one another. There is little agreement as to the spectral characteristics of pigeon cone pigments (WALD, 1958 ; BRIDGES,1962), and there is even some doubt(SrLLMAN, 1969) that they exist. One of the complicating factors is that pigeons (Iike most sauropsida) have colored oil droplets in the outer segments of the cones. Any attempt to construct a unifying picture of the physiological mechanisms of the pigeon’s color vision will have to take into account the role of these oil droplets. There is somewhat more agreement among the results from ~~~~~~ recording studies than those for photopigments. D~NNF+R(1953) recorded the massed discharge of the pigeon’s retinal ganglion cells, and those units which gave relatively narrow-band spectral responses (modulators) tended to be of one of three types: (1) Peak response at 480 nm; (2) Peak response at 540 nm; (3) Peak response at 600 nm. As pointed out by WRIGIFFand CUM,MING (1971) the peaks of these response functions correspond to the intersection points of the color naming functions, rather than to the peaks of them. In a recent experiment examining single units of the nucleus rotundus of pigeon’s thalamus GRANDAandYAZ~LLA (1971) found that those units which did not have a peak response at 500 nm, generally showed peaks centered at 540 and 600 nm. These experimenters have recently (personal communication) isolated opponent units, where 500 and 540 nm- represent (1956) and BWUGH 6 See ~~~PHAW’S(1955) anaIysis of the experiments by OUTT!%N and ICAUSE? (1961).

Psychometric

and Psychophysical

Hue Discrimination

Functions for the Pigeon

1463

crossover points in the spectrum between wavelengths that excite the unit and those that inhibit the unit. It is hoped that the reiationships among functions for the pigeon, obtained through various approaches to color vision, wili indicate which retationships among functions reported for humans are fortuitous, and which have functional significance. The ability to separate the functional relationships from the fortuitous ones will help advance our understanding of the underlying mechanisms of color vision. The pigeon is particularly interesting as a color vision subject, because its hues do not correspond to human hues. The present experiment has shown pigeon’s hue discrimination to be best at its hue boundaries, corroborating the same finding for human subjects.

REFERENCES BEARE,A. C. (1963). C&or-name as a function of wavelength. Am. f. P.~ychof. %,248-256. BEARE,A. C. and SIEGEL, M. H. (1967). Color name as a function of wavelength and it&t-u&on. F&e@. & Psyckopkys. 2,521-527. BEDFORD,R. D. and WYSZECKI,G. (1958). Wavelength discrimination for point sources, .L opt. Sot. Am. 48, 129-135. BIRCH,J. and WRIGHT, W. I). (1962). Color discrimination. Pkys. Med. Biol. 6, 3-24. BLOUGH, D. S.(1957) Spectral sensitivity in the pigeon. .r. opt. sot. Am. 47, 823-833. BLCSUGH,I).S.(t96l). The shape of some wavelength generafization gradients. J. exp. Ana!. De&. Q 31-40. &XXTOX, R. M, (1963). Contributions of threshold measurements to cofor discrimination theory. .T. opi. sue. Anz. x+,165-178. BOYNTON,R. M.,SCHAFER,W. and NEUN, M. E, (1964). Hue-wavelength relation measured by color-naming method for three retinal locations. Science, N. Y. 145, 666-668. BRIDGES,C, D. B. (1962). Visual pigments of the pigeon, Columba livia. Vision Res. 2, 1%137. CLARKE,F, R., BIRIXALL,T. G. and TANNER,W. F. JR. (1959). Two types of RGC curves and definitions of parameters. J. (icotls. Sot. Am. 31,629-630. DONNER, K. 0, f1953). The spectral sensitivity of the pigeon’s retinal elements. I. P&J&?!. bnd. 122,524-537. Eonrj, 3. P. and CLARKE,F. R. (f%Q. Psychophysics and signal detection. En Experimental ~eth~s nnd ~nstr~~~~~t~tion in Psyckology. (edited by J. B. Smows~l,)pp. 211-241. McGraw-Will,New York, EGAN, J. P., SCHULMAN, A. I. and GREENBERG,G. Z. (1961). Memory for wave-formand time uncertainty in auditory detection. J. acous. Sot. Am. 33,779-781. GRAHAM,C. H. and RATOOSH,P. (1962). Notes on some interrelations of sensory psychology, perception and behavior. In Psychology: A Study of Science. Vol. 4. Biologically oriented fields : their pIace in psy&fogy and in biological sciettce (edited by S. KOCH). Gnnrm~, A. M. and YA~ULLA, S. (1971). The spectral sensitivity of single units in the nucleus rotundus of pigeon, Columbn fiuia, J. gen, Pkysiol. n, 363-384. GREEN,D. M. and SWE~S,J. A. (1966). Signal Detection Theory and Psychophysics. Wiley, New York. GUITMAN,N. and KALL?.H, H. 1. (1956). Discriminability and stimulus generalization. J. exp. Psychof. 51, 79-88.

HAMILTON, W. F. and COLEMAN, T. B. (1933). Trichromatic vision in the pigeon as illustrated by the spectral hue discrimination curve. J, camp. Psychof. 13, X83-191. HEAY~X~, 0. S. (1965). Qpficar Properties of Thin Solid F&s. Dover, New York. EnsHLxY,K. S. and WADE, M. f1946). The Pavlovian theory of generalization. Psychof. Rev. s&72--87. H. (1923). Studies on the relative physiological value of spectral hght III. The pupillomotor effects of wave-lengths of equal energy content. Am. J. Physiol. 64, 97-l 19.

LAmENS,

LAURENS, H. and HAMILTON, W. F. (1923). The sensibility of the eye to differences in wavelength. Am, J. Physiol. 65, 547-568.

LURIA,S. M. (1967). Color-name as a function of stimulus-intensity and duration. Am. S. Psy~ltol. SO, 14-27. POXLACK,1. (1959). Message uncertainty

and message reception, J. arousr. Sec. Am, 31, 1500_1508.

THERAND CORPORATION. (1955). A M~~~~~~ Random Digits with IO.300 Norm& De&tes, Free Press, Glencoe, 111. SHEPARD, R. N. (1965). Approximation to uniform gradientsof generalization by monotone transformation of scale. In Stimulus Generalization(edited by D. 1. MOSTOFSKY). pp. 94-l 11. Stanford University Press, Stanford.

SILLMAN,A. J. (1969). The visual pigments of several species of birds. Vision Res. 9, 1063-1077. SOKM., R. R. and ROHLF, F. J, (1969).Biometry p. 457. Freeman, San Francisco.

1464

ANTHONY A. WRIGHT

%ERNHEZM, C.

E. and &WNlDN, R. M. (1966). Uniqueness of perceived hues investigated with a continuous judgemental technique. J. exp. Psychol. 72, 770-776, Swe~s, J. A. (1361). Is there a sensory threshold? Science, N. Y. 334, 168477. WALQ G. (1958). Retinal chemistry and the physiology of vision. In Vis~i~bIe~ of Colour. Symposium No. 8. National P~ys~I~ratory (EJ.E& pp. 741, H.M.S.O. London. WEAL&R. A. (1951). Hue-&c&ination in par¢ral parts of the human retina measured at different hnninance levels. J. Physiol., Land. 113, 115-122. WRIGHT, A. A. and Cubnfl~G, W. W. (1971). Color-naming functions for the pigeon. J. exp. Anal. Behav. 15, 7-17. WRIGHT, W. D. (1947). Reseurches on Normal and Defective CoioBr Vision. Mosby, St. Louis. Abstract-Four pigeons were trained on a “yes-no” signal detection procedure to detect the presence or absence of hue difference in a split field. Response bias was maniptdated by varying reinforcement probability. Discrimination indices (d’), computed for each reference wave-number (l/h), were linearly related to wave-number di&rence, and were shown to be a ratio scak of wave-number difference. Mtersection of these functions at d’ = 211 yielded a hue disc&&nation function. This psychophysical hue d’is&mination function and t&e.slope&of the psychometric fimctions showed that hub-divination was best at !KX$5404fa and 660 nm spectral regions. R&B&-On dresse quatre pigeons & une technique de detaction par “oui-non” pour d&eider de la prdsence ou de l’absenee d’une di@&eBoede tonal&d dans un champ a deuxpkges. On manipulait les r&onses en moditiant la prob&ifit~ de renforcemen t. Les indieos de discrimination
rony6a 6arw BarpeBBpoBaBH o6Bapyxmaallr BaJBPiBeBJIB QTCyTCTBEf~

&X33-# B IIRGTOBOM TOHCB U0J-N3$X!HES~ACJRXiBOhf HB J#H lTOlIOBSfRk& @X#S&3kaJIaCb T. H, ‘Lna-Ber”-nponeztypa o6BapypbBHII cmrrana. Ha pe+rmmm oxa3xBarmcB anmrnife “‘yc%?JIem*’ (rc@forcement). l’lOMOU0K) MarnmymrpOBaHSfJl pa3BO3MOXE-

c

Hoxa3a3wtB pa3nmremm (@), o6pa6orasmbre ~o~mbrorepoBr w rtaBuIoro~~or0 BonBoaoro mmna (l/h), 6hunr mrne&rro coor~ecema B pa3mnrBro BbOIXBoBo!dBsK‘JteH &ma ~0~eHa rmuura orrtomemia K pa3mmx10 B mwrmonohi -sfuze. mpcuwmme 3rax ~#ymmBii npB d’ = 2.0 Eaaarro ocno~ pa== &BBnmr pasmcrcaaylt BBerouoro rona. 3Ta rmx*~wKan fpy~utrur pa3mmmR (P)WU&i

IXOWW~~T,

bHblX 06~amtx

!m,

rlT0 pa3mneRosc

s40,sso

A 600 AM.

IWlTOBGrO-

ToR8

-