spectral characteristics of the response of the pupil to retinal stimulation.

Transcription

1 478 J. Physiol. (1962), 164, pp With 8 text-figurem Printed in Great Britain THE SPECTRAL SENSITIVITY OF THE CONSENSUAL LIGHT REFLEX BY M. ALPERN* AND F. W. CAMPBELL From the Physiological Laboratory, University of Cambridge (Received 4 June 1962) Although it is well known that light falling on the human retina will reduce the size of the pupil of each eye, it still cannot be unequivocally stated that this effect is obtained by the excitation of only cones, only rods, or both rods and cones. Since the spectral sensitivities of the photopic and scotopic eye are well established the identification of the photoreceptors activating the pupil should be readily achieved by measuring the spectral characteristics of the response of the pupil to retinal stimulation. Laurens (1923) was able to demonstrate that the curve relating pupil size to wave-length (equal energy spectrum) depended upon the adaptation state of the eye. The curve for the light-adapted eye had a minimum at m,u, the dark-adapted curve had one at m,u. This suggests that both rods and cones may influence the size of the pupil. Since, however, the relation between pupil size and intensity of the stimulating light is non-linear, such curves tell us nothing about the amount of energy necessary at each wave-length in the spectrum to produce a constant response. It is the information obtained from this kind of curve (an action spectrum) that is required in order to compare any non-linear physiological-response characteristic to an absorption spectrum of a visual pigment or to the well-documented spectral sensitivity curves of the visual rods and cones. It is in the attempt to demonstrate just this kind of parallel between the spectral sensitivity curves of the photo-receptors for vision and for pupillary motion that the data available until now give such a confused picture. If one adapts the eye to a very dim light and measures the size of the pupil, then increases the intensity of the light a little, allowing the eye to adapt to this new level, and measures the size of the pupil again, and continues the process from a very low to a very high light intensity in equal logarithmic steps, the pupil shows very little, if any, change in size until the cone threshold is exceeded (de Launay, 1949). This result is evidence tn support of cones exclusively as the pupillomotor photoreceptor. On the other hand, if this experiment is repeated with various monochromatic * Assisted by Special Fellowship BT-712 U.S.P.H.S.; on leave from the University of Michigan.

2 PUPILLARY SPECTRAL SENSITIVITY 479 lights and, once the pupil begins to respond, the action spectrum required to produce a standard response is measured, a curve which is similar to, although not quite identical with, the scotopic spectral sensitivity curve is obtained (Wagman & Gullberg, 1942; Bouma, 1962). Furthermore, and this is evidence in support of the idea that only rods can excite the pupil response, these authors reported that the action spectrum was independent of the intensity of the adapting light. A second method of identifying rods and cones is on the basis of their directional sensitivity. Flamant & Stiles (1948), in man, and Donner & Rushton (1959b), in the frog, have both shown that while the effect of light is very much dependent upon its angle of incidence on the cones, rods have very little, if any, directional sensitivity. Spring & Stiles (1948) showed that for a very large field (520) the pupil size was only very slightly larger when the incident light passed through the edge rather than the centre of the dilated pupil (of the other eye). This result might be interpreted either as a verification of the 'rods only' hypothesis or of the ' mixed rod and cone' hypothesis, since with a large field the contribution of the cones to the pupil response might well have been masked. Alpern & Benson (1953) showed that this latter was the most reasonable interpretation, by repeating this experiment with a very small (less than 10) centrally fixated test light. They found as marked a directional sensitivity of the pupillary photoreceptors, under these conditions, as is found with psychophysical measurements of cone directional sensitivity. Moreover, evidence from the influence ofthe state ofdark-adaptation on the response of the pupil to the stimulation of the retina by light flashes also suggests that both rods and cones activate the pupil (Alpern, Kitai & Isaacson, 1959). Since this conclusion is, however, in direct conflict with the finding that at all levels of adaptation the action spectrum of the pupil is a scotopic one, the entire question of the response of the pupil to monochromatic light has been reinvestigated. In describing the stimulus for the light reflex it is convenient to differentiate between steady-state and transient light stimulation. The first of these refers to experiments in which the eye is adapted to a given light level and the size of the pupil at that level is determined, when equilibrium has been reached. On the other hand, transient light stimulation means any other time sequence of light stimulation, such as a step or sinusoidal sequence, presented to the eye under a given adaptation condition. As a number of different techniques have been used in this paper to establish the spectral sensitivity curves of the pupil response, and as separate introductions and discussions are required for each group of experiments, these have been described in six Parts: Part I: Spectral sensitivity curve by the method of equivalent colour substitution.

3 480 4M. ALPERN AND F. W. CAMPBELL Part II: Steady-state pupil spectral sensitivity. Part III: Scotopic response. Part-IV: Reflexion coefficient of the fundus. Part V: Photopic response. Part VI: Synthesis of rod and cone contributions. PART I Spectral sensitivity curve by the method of equivalent colour substitution Because of the wide variety of non-retinal factors which contribute to variations in the size of the pupil it was important to design a method to determine the spectral response curves which would be reasonably accurate, quite consistent and allow for the determination of values at a large number of different wave-lengths in a single observer in a relatively short period of time. We measured, at each wave-band in the spectrum, the energy required by the monochromatic light such that sinusoidal transitions between it and a standard light were not associated with any variations in pupil size. This method gave quite consistent results and permitted the determination of the action spectrum (from 450 to 650 m,, in 10 mp steps) in about 10 min. It was remarkably free from artifacts induced by light-adaptation and the order in the spectrum the measurements were made. In some experiments the standard was of zero light intensity and in this case the test reduced to the criterion of the minimum amount of energy required for a detectable pupil response. While such a threshold value is a perfectly valid criterion of spectral sensitivity, measurements obtained in this way do not necessarily represent the minimum amounts of energy required to produce the smallest possible response of the fully dilated pupil. METHODS The apparatus (Fig. 1) used may be most conveniently described in two main sections; the stimulating light system and the pupillometer. Stimulating light system The image of a 6V 18A ribbon filament (S 1) was focused by lens L 1, after passing through a neutral wedge (N.W.), on the entrance slit of a Hilger and Watts D290 Grating Monochromator. The light leaving the exit slit was then plane-polarized. It also passed through a large disk of Polaroid (R.P.) which could be rotated slowly by a controlled-speed motor. Lens L2 focused the exit slit in the plane of the subject's left eye. The pupil, of this eye, was dilated with 2 % homatropine. Lens L2 was seen in Maxwellian view and the size of the field controlled by placing a variable diaphragm (D2) in such a position that its image, formed by lens L2 and the optics of the eye, was focused on the retina. Rotating the Polaroid, R.P., permitted the stimulus intensity to be varied sinusoidally at any selected frequency.

4 PUPILLARY SPECTRAL SENSITIVITY 481 A second light path from lamp S1 (by way of prisms Pr 1, 2 and 3) circumvented the monochromator but was mixed with the light from it after reflexion at the glass plate, G.P., placed at 450 to the axis. In this second path lens L 3 formed an aerial image of the lamp filament in an image plane so that lens L 2 could form a further image in the plane of the pupil of the subject. Neutral and coloured filters could be placed in this second path at F 1 and F 2. A Polaroid, P 2, was positioned in this path, oriented to plane-polarize the light at right angles to that of the monochromatic beam. P3 Pd2 52 P2 i~~~~j ~~F2, Fl I~~~~~~V IGVP VVsi4 F4 L5 D1 L4 SW. Fig. 1. Diagram of the apparatus. See text for explanation. The wave-length indicator on the monochromator was checked against spectral lines produced by mercury and neon sources. The scale was accurate to m,u The monochromator had a dispersion which was independent of wave-length. The dispersion for the slit widths used throughout these experiments was measured photoelectrically using mercury line 546 m,ut and the half band-width was found to be 9*8 m,u. At either end of the visible spectrum where the sensitivity of the retina is decreasing rapidly, the nominal wave-length may be an erroneous description of the centroid of light flux emerging from the slit due to slit-width errors. The extenit of these errors was calculated in the case of the Commission Internationale de l'eclairage (C.I.E.) photopic curve for the two most extreme cases, i.e. nominal wave-lengths of 450 and 650 ma.the actual computed mean wave-lengths after correction for these errors were and my, respectively. These discrepancies are so small that no corrections for slit-width errors have been necessary. The relative spectral energy output of the monochromatic stimulating system within the range m,u was measured independently (with good agreement) by two photocells calibrated by the N.P.L. At the peak of the photopic spectral sensitivity curve (555 mil), the stray light was estimated to be 001 of the monochromatic luminance. All the data were limited to regions of the spectrumn where the monochromator stray light was an insignificant fraction of the light emerging from the slit. All the 'neutral' filters and the wedge used in the experiments were calibrated for density between 400 and 700 m,u using a Hilger Uvispek photo-electric spectrophotometer. The wedge was calibrated at 10 mu steps at four different densities and for one wave-length in 16 steps, each of 0-2 log.density. Theset measurements revealed a linear relationship

5 482 M. ALPERN AND F. W. CAMPBELL between wedge position and wedge density over the entire range used. The necessary corrections have been applied to all data presented. As an over-all check on the calibration of the monochromator, neutral filters and wedge, the photopic luminosity curve was measured with this apparatus. The mean results are shown in Fig. 2, where they are compared with the C.I.E. photopic curve. The agreement is good. II II 1.9 / 00 0'~~~~~C >**15 _._5 >1 GJ _ Ib I'Q Wave-length (m,) Fig. 2. Psychophysical measurements of photopic luminosity by the method of flicker photometry. Centrally fixated field was 10 in diameter; the rate of alternation, 20 c/s; the standard field had an Ilford 622 filter in the light path. These results are the mean from eight young adults with normal colour vision. The pupillometer The principle used to detect changes in pupil size was to focus a high-intensity infra-red light on the iris, and then to detect photo-electrically changes in the amount of light reflected from the iris. The amount of reflected light decreases as the pupil area increases. This method does not permit the precise measurement of the magnitude of large changes in pupil area, as the iris does not necessarily reflect light uniformly. In this study, however, intensity was varied to obtain a null (i.e. absence of changes in pupil area) or a minimum detectable pupil response, and an accurate knowledge of the actual pupil size was never necessary. The infra-red light source (S4) was a Phillips, 8V, 50W, lamp with reflecting surfaces deposited on the lamp envelope. The lamp was run at a constant voltage. The reflectors in 650

6 PUPILLARY SPECTRAL SENSITIVITY 483 this lamp produce a small image of the filament about 2 cm before the front window of the lamp. In the plane of this aerial image the sector wheel (S.W.) chopped the light beam at a frequency of 400 cls. Lens L4 was a 1 in. (2.5 cm) focal length, 1 in. aperture, compound lens, so placed that another aerial image of the filament was formed in the plane of a variable circular diaphragm (D1). Lens L5 was similar to lens L4 and was mounted so that it could be moved along the axis of the optical system, and also vertically and horizontally. These adjustments allowed one to form an image of D 1 on the iris of the subject's right eye and to illuminate a circular area 7 mm in diameter concentric with the centre of the pupil. An infra-red filter (Polaroid XRX 30) was placed before the eye at F4 so as to render this illuminating beam invisible to the subject. The light reflected from the iris was monitored by placing a lead sulphide photo-detector about 3 in. (7.5 cm) from the eye, in a position which did not obscure the illuminating beam. As the illuminating beam was chopped at 400 c/s, the electrical output from the photodetector (Pd 1) alternated at this frequency. This permitted the use of a.c. amplification with its attendant advantages over d.c. amplification. The condenser-coupled amplifier was tuned to 400 c/s to improve the signal-to-noise ratio. The amplified signal was then full-wave rectified and the resulting 800 c/s signal passed through a low-pass filter with a 33 c/s corner frequency and 36 db/octave high-frequency cut-off. The resulting d.c. signal was then backed off with a constant voltage and changes in the difference signal could then be applied to the recording or observing system. The output from the amplifier was displayed on the Y axis of an oscilloscope. The X axis of the oscilloscope was fed from a signal obtained from a photo-detector (Pd 2). This photocell was placed on one side of the rotating polaroid (R.P.). On the other side was placed a small torch lamp (S2) with a fixed Polaroid (P3) in front of it. The output from Pd2 thus monitored the phase of the main monochromatic light beam. By suitable adjustment of the angle of P3, the frequency of rotation of the Polaroid and the amplitudes of the X and Y signals, it was possible to display movements of the iris on the oscilloscope as a series of Lissajous figures. The low frequencies used to stimulate the pupil light reflex (from 0 5 to 1 c/s) necessitated the use of a cathode-ray oscilloscope with a long persistence phosphor for convenient observation. In this way a movement, or a series of movements, of the iris which occurred in a fixed phase relation with the stimulating light beam were readily identified, even in the presence of random movements of the pupil. In the experiments where one coloured light beam was substituted by another, Lissajous figures could be obtained rotating in one direction when the intensity of one beam was high and in the opposite direction when it was low. By adjusting the wedge up and down several times between the positions -where clockwise or anticlockwise motions were clearly obtained the neutral point could be quickly determined. In some experiments the miimum energy required for a detectable pupillary constriction was measured. In these experiments, only the light path through the monochromator was used as a stimulus and in this case the rotating Polaroid provided an oscillation between zero and some finite intensity. In this case when the pupil responded a Lissajous figure appeared and the intensity of the monochromatic beam could be adjusted until this Lissajous figure barely disappeared or appeared, whichever was required. RESULTS The results of a series of experiments using the method of equivalent colour substitution are illustrated in Fig. 3A and B and in Fig. 4. All these curves are very similar; they have a Amax 530 to 540 m,. The lines drawn through the points are theoretical and are discussed in Part VI. In these experiments it is perhaps a misnomer to describe the procedure

7 484 M. ALPERN AND F. W. CAMP. BELL A B o 2 _.J.W. C CI: Wave-length (m,u) Fig 3. For legend see opposite page.

8 PUPILLARY SPECTRAL SENSITIVITY 485 as substitution, since at the extreme red end of the spectrum there was a suggestion of a phase difference between the two stimulus colours, even when the intensities were so equated that neither a clockwise nor an anticlockwise Lissajous figure was visible. The null under these conditions was a straight line slightly tilted to the horizontal. This suggests a difference in latency between lights of different colours but of the same pupillary sensitivity. The experimental result obtained by the methods just described was a quite accurate measurement which was consistently obtained even under a variety of stimulus conditions. In Fig. 4 the results of three experiments are presented. Despite the variety of stimulus conditions under which these various curves were obtained they all agree reasonably well with one another when plotted on a relative basis. They are all very similar to results obtained by Schweitzer & Bouman (1958) for the spectral sensitivity of the pupil when the differential threshold was determined for 10 diameter test field seen against a red background. The curves in Figs. 3 and 4 are characteristic neither of purely scotopic (rod) vision nor of purely photopic (cone) vision. This was found even when the geometrical image of the test light was confined to the rod-free fovea. This result by itself emphasizes the importance of light reaching the retina outside the geometrical image (non-focal light) when observing the pupil responses to lights seen against a totally dark background. The way in which this comes about is discussed in Part IV. Legend to Fig. 3 Fig. 3. A and B. Spectral sensitivity curves for the pupil of two colour-normal observers by the method of equivalent colour substitution. The data are the mean + S.E. for ten repetitions; the standard beam contained an Ilford 621 filter and the intensity of retinal illuminance was 20 td. Diameter of the centrally fixated field was 20. C. Steady-state measurements at high-intensity criteria. x, relative sensitivity results obtained by drawing independent curves through results at each wave-length, and determining from them the radiance necessary for 2-5 mm change in pupil size. *, the sensitivity as measured by the amount each wavelength set was laterally shifted for superpositioning with minimum scatter about the smooth curve in Fig. 5 once the dark control values were equated. The open circles, 0, represent the sensitivity as measured by the amount the continuous smooth curve in Fig. 5 was shifted for the fit at each wave-length. +, the sensitivity obtained when the method used to obtain the open circles was applied to the high-intensity data of Wagman & Gullberg (1942). In all three figures the lines are theoretical (see p. 502): the continuous lines according to eqn. 2 (ineachcase ac/,b is very nearly unity and K = 0); the interrupted lines according to eqn. 1 (a/b has the values 0 73, 0-78 and 1-58, in A, B, and C, respectively). In producing these theoretical curves the value for o-a was the C.I.E. scotopic curve modified by the smooth curve in the top graph of Fig. 7; OA was the observer's photopic luminosity curve except in C, where it was the photopic luminosity curve of the C.I.E. standard observer.

9 486 M. ALPERN AND F. W. CAMPBELL While the results in Figs. 3 and 4 were obtained for a variety of differing stimulus conditions and observers, the curve (An,ax = 530 to 540 m,u) is somewhat different from the one described by Wagman & Gullberg (1942) (Amax = 511 m,u) for steady-state spectral sensitivity. Is this because the physiology of the responses to steady-state stimulation is essentially different from that of the responses to transient stimulation? 2*0H x 1'6[- x 6' 0 x c 4) M L.._ 43 4) '4 ba 0 -A 0-4 H Oj- J.w I I I I I I Wave-length (m,) Fig. 4. Pupil spectral sensitivity curves obtained by using three different methods on the same observer. The open circles, 0, together with the solid lines are from the experiment whose results are given in Fig. 3B. The x 's are for threshold criterion with a centrally fixated field, 200 in diameter. The absolute values for these latter results are about 1 log1o units higher than for the open circles. The dots, 0, are for the equivalent colour substitution (white light standard) for a 20 test patch centred on the fovea with zero background. The absolute values for the fovea curve are about 1 log1o units lower than the open circles. The three curves are vertically shifted so as to superimpose with miimum scatter.

10 PUPILLARY SPECTRAL SENSITIVITY 487 PART II Steady-state pupil spectral sensitivity In order to decide this matter the experiments of Wagman & Gullberg (1942) have been repeated, except that the 16-6' field was replaced by a smaller (about 80) diameter stimulus area. The size of the pupil of the contralateral eye was measured after steady-state adaptation to monochromatic lights with intensities varying from very low to very high levels. METHODS The right eye of the observer was adapted to a monochromatic light, provided by a tungsten ribbon filament, which was collimated, allowed to pass through narrow-band Bausch and Lomb interference filters with suitable gelatin subsidiary filters to cut off secondary peaks, and then brought to a focus on a small round aperture, which in turn was imaged in Maxwellian view on the centre of the entrance pupil of the observer's right eye. Nine such combination filters with dominant wave-lengths of: 480, 502, 512, 532, 550, 575, 594, 620 and 650 m,u were used. The energy of the light source was measured at various wavelengths in the visual spectrum with a calibrated photocell and spectrometer and found to agree quite closely with the distribution of a tungsten ribbon filament with a colour temperature of K. The transmission characteristics of each interference-gelatin filter combination was measured spectrophotometrically. Wratton 96 'neutral' filters were used to vary the luminance. Since it was necessary to have very dim fields the filters were used in combination up to 8-0 density, which is much too dense to be measured by a spectrophotometer. Instead, at each wave-length in the spectrum the transmission characteristics of a sample 1-0 density of this filter, as supplied by the manufacturer, was assumed to be characteristic of all of the filters and the assumption was made that the filters which were used related to this sample filter in direct proportion to their respective rated densities. Any errors induced by these assumptions are very small for the wave-lengths used in these experiments, because within this range the filters are almost neutral. The energy of radiation emerging from each filter combination was then computed in the usual way. The power to the lamp was supplied by a constant-voltage transformer. In this experiment, the observer was dark-adapted for 30 min, at which time several photographs were made of the diameter of the pupil of the left eye exposed only to the very faint red glow from the light of a tungsten lamp passing through the almost visually opaque (Wratten 87) infra-red filter, the illumination source for the photography. After this the observer adapted to a very dim monochromatic field. A measurement was made after 30 sec adaptation and a second one 30 sec later. The intensity of the adapting field was then increased by 0-5 log. unit and the entire process was repeated. The process was again repeated at the next intensity, and so on. In this way measurements of the relation between pupil diameter and intensity were obtained for one monochromatic light at a single session. The whole process had to be repeated for each of the other monochromatic lights in separate sessions. The results are from four observers, one of whom repeated the entire procedure. RESULTS The results of these experiments are illustrated in Fig. 5. This figure shows the pupil diameter as a function of radiance of the adapting field (W/steradian x cm2), for each monochromatic wave-band. In this figure 32 Physiol. 164

11 488 M. ALPERN AND F. W. CAMPBELL Log10 PA (W/steradian x cm2+15) Fig. 5. Steady-state pupil diameter as a function of radiance for various monochromatic wave-lengths. The top set of results is correctly positioned on the graph, each of the others has been vertically displaced 0.5 mm downwards with respect to the set immediately above it, except the lowermost curve which is displaced 1-0 mm downwards. To obtain the curves the various sets of results have been equated at the control values (measured in the dark) and then displaced laterally to obtain superimposition with minimum scatter. A smooth graph, estimated by eye, was then drawn through the results to show the average trend. The solid curve is a lateral displacement of this smooth graph in order to fit the high-intensity results. The interrupted curve is a lateral displacement of the smooth graph in order to fit the low-intensity results. The curves were all fitted by eye.

12 PUPILLARY SPECTRAL SENSITIVITY 489 the topmost set of results is correctly positioned on the graph but each succeeding set is progressively displaced downwards from the one above it. In order to decide about the spectral sensitivity from data of this kind it is best to fit smooth curves to the results. The following procedure has been used to fit the curves illustrated in Fig. 5. The sets of results were all vertically displaced until the measurements of the dark control values at each wave-length were coincident. Each set was then laterally shifted until the results were superimposed. The amount of lateral displacement necessary for superimposition with minimum scatter provides one measurement of the spectral sensitivity of the steady-state pupillomotor system, and the results using this criterion are illustrated in Fig. 3 C as the filled circles. A single smooth graph could then be drawn by eye to fit all the data, although a certain amount of scatter remained. This single graph was then used to fit the data for each individual wave-length and in following this procedure the amount of lateral shift required for the best fit of this single smooth graph to the results at any given wave-length provided a second independent criterion for pupillomotor spectral sensitivity. In performing this latter operation, however, it became apparent (particularly at the extremes of the spectrum) that a different amount of lateral shift of this single graph was required to fit the data at the highintensity (small pupil diameter) part of the results (solid lines in Fig. 5) than was needed to fit the data at the low-intensity (wide pupil diameter) part of the results (interrupted lines in Fig. 5). In this sense then, the results demonstrated a Purkinje shift. The results for the high light intensities (open circles, Fig. 3C) agree with the filled circles in the same figure which were obtained by the criterion already described. The results for the low light intensity are illustrated in Fig. 6C as the open circles. While this procedure yields reasonably accurate measurements for the high intensities, fitting the wide pupil data in Fig. 5 by lateral shift of a single smooth curve (interrupted lines) leads to a good deal of scatter about the curve even when the optimum fit is achieved. In order to avoid being misled by a mistake in the curve-fitting process a second procedure was also used. Individual smooth curves were drawn which, as judged by eye, provided the best fit to each individual monochromatic set of results. From these curves it was then possible to measure the energy at each wave-band required to produce a 2-5 mm decrease in pupil diameter (over the dark control) and these results are illustrated as x's in Fig. 3C. Similarly, the energies required to produce a 0*5 mm change in pupil size are illustrated in Fig. 6 C as the filled circles. While differences between the two curve-fitting procedures do appear, they are small, and in general the two methods agree reasonably well. The results in Fig. 3 C do not differ in any detail from those obtained 32-2

13 490 M. ALPERN AND F. W. CAMPBIuLL 20,0*-- A 0 A.S. 00. B C~~~~~C 10-6_ \ + ~ \ 000 \~~~~~~ Fig. 6. Wave-length (m,) For legend see opposite page.

14 PUPILLARY SPECTRAL SENSITIVITY 491 with transient stimuli (Fig. 3A and B) and the similarity between the various curves in Fig. 3 therefore suggests that similar principles must be operating in the two cases. Furthermore, the results illustrated in Fig. 6 C show that at low intensities the spectral response is dominated by rods. These data agree well with the results obtained by Wagman & Gullberg (1942) for 0*5 mm decrease in pupil diameter (+ 's in Fig. 6C). (It seems likely that the deviation of Wagman & Gullberg's result at 450 m,u in this figure was due to an inappropriate assumption that their neutral filters attenuated equally at all wave-lengths.) Wagman & Gullberg state that since their curves are more or less parallel the increase of the intensity of the light is not associated with any change in characteristics of the spectral sensitivity. But on careful examination, their data do show a mild Purkinje shift at very small pupil diameters and if a similar procedure is used to fit their data to that employed to determine the open circles in Fig. 3 C, we find the results indicated by the + 's in Fig. 3 C. These results are sufficiently similar to those derived from Fig. 5 to be certain that the same mechanisms operate in the two cases, although, perhaps because of the smaller stimulus field in our experiment, the highintensity curve was a satisfactory fit over a smaller range of intensities in the case of Wagman & Gullberg's data than in Fig. 5. The data in Fig. 6 C are not a completely satisfactory fit to the scotopic spectral sensitivity curve and an explanation for the discrepancy is discussed in Part IV. These curves are quite similar to those obtained by Schweitzer (1956) at the absolute pupillary threshold, and to those obtained in the experiments in Part III, in both of which transient stimuli are employed. Thus, again, the relations which govern transient stimulation are not essentially different from those governing steady-state responses. In the theoretical formulation which is developed in Part IV to explain the deviations from the scotopic curve illustrated in Fig. 6 it is of some importance to know whether or not the discrepancies in Fig. 6C are less Legend to Fig. 6 Fig. 6. A. Measurements with the test light focused on the blind spot of A.S., a severely protanomalous observer. 0, pupil results; 0, psychophysical absolute threshold on the blind spot (this latter is typical of the results, even on normal eyes). B. Measurements (pupil only) with test light focused on the blind spot of J. W., who has normal colour vision. C. Pupil spectral sensitivity curves from the steady-state results of Fig. 5. 0, sensitivity results as measured by the lateral shift of the interrupted curve of Fig. 5; 0, sensitivity results as measured by drawing individual smooth curves through the set of results for each wave-length and then determining the amount of radiance required at each wave-length to produce 0 5 mm change in pupil diameter over the dark control; +, results from Wagman & Gullberg (1942) by the same method used to obtain the solid circles. The smooth curve is the C.I.E. scotopic spectral sensitivity curve.

15 492 M. ALPERN AND F. W. CAMPBELL pronounced than those in Fig. 6A, B (Campbell & Alpern, 1962). But the uncertainties of the methods by which the interrupted curves have been fitted to the results in Fig. 5 are too large to permit any quantitative statement about this with the available data. Inspection of the data in Fig. 5 readily reveals that small errors in estimation of the fit of the smooth curves to the experimental points can lead to somewhat different spectral response curves. Furthermore, the scatter of these results is so large, particularly at the low intensities, that even the rather elaborate curve-fitting methods used to fit these results give an unsatisfactory agreement between the points and the curve. For these reasons this method of measuring pupillary spectral sensitivity is not very reliable, and one hesitates to draw any conclusion from these experiments, with the possible exception that the processes which determine spectral sensitivity under these conditions are not obviously different from those which determine the spectral sensitivity under conditions of transient stimulation. PART III Scotopic response In the previous parts it was possible to show that the usual spectral sensitivity curve obtained for the steady-state or transient pupillary response has a Ama1 at mix, and it will be shown that this is a mixture of photopic and scotopic components. At very dim light levels it was found that for steady-state conditions (Part II) a more scotopic curve could be obtained, and Schweitzer (1956) obtained a similar result by measuring the minimum amount of light required for the smallest detectable iris change in the dark-adapted eye. The pupillometer used in the present study would not have detected this kind of a change in a fully dilated pupil without introducing artifacts due to eye movement. However, it turns out that it is possible to study the same kind of receptor process by measuring the threshold energy required for a detectable response when the stimulus beam is confined to the blind spot. METHODS The apparatus was essentially the same as that described in Part I, with the exception that light passing only through the monochromator was used in the stimulus beam. A small mirror was attached to the nasal edge of the lens (L 2) on the surface close to the eye. A small very dim tungsten lamp was then moved about in the field of this mirror until the light from the stimulus beam (about 2 in diameter) fell on the subject's blind spot when he fixated on the mirror image of the lamp filament. Measurements were made of the threshold response energy when the stimulus light was exposed against the zero background at a frequency of exposure of approximately 1-0 c/s.

16 PUPILLARY SPECTRAL SENSITIVITY 493 RESULTS The results of these experiments on two observers are illustrated in Fig. 6A and B. The curve is the C.I.E. scotopic spectral sensitivity curve. Hess (1908) showed that light focused on the blind spot can cause the pupil to contract. This result demonstrates that scattered light can play an important part in evoking pupil responses. For a number of reasons, light scattered by the blind spot is favourable for the excitation of rods. In the first place there are approximately 20 times as many rods in the primate retina as there are cones, and these are most numerous in the peripheral retina while the cone population is most dense in the small fovea area. In the second place the area over which spatial summation can occur is much larger for rods than for cones and for this reason large areas of weak light are efficacious in stimulating rods. Finally, the finding (Flamant & Stiles, 1948; Donner & Rushton, 1959b) that the cones have a marked directional sensitivity whereas the rods do not, means that this stray light tends to excite rods rather than cones. Because of these considerations the spectral sensitivity curves illustrated in Fig. 6A and B are more or less scotopic. However, it is apparent that the results are not in perfect agreement with the C.I.E. scotopic spectral sensitivity curve, and therefore with what might be expected if the response to blind spot stimulation was due to stimulation of rods, and rods alone. The agreement that is illustrated in Fig. 6 is somewhat less than that obtained with visual thresholds on the blind spot, but about the same as those Schweitzer (1956) obtained for a minimum detectable iris change in the fully dark-adapted eye. These pupil response curves are somewhat higher in the red and blue ends of the spectrum and somewhat reduced in the green as compared with the C.I.E. scotopic curve. Two possibilities to explain these discrepancies have been considered. The first is that they are due to the activity of another photosensitive visual substance in addition to rhodopsin; the second is that they are due to the screening by a photostable coloured filter somewhere between the light source and the rods, which was not present for the measurements of the C.I.E. curve. In measurements of the absolute visual threshold, for small test areas in the parafovea, an increased sensitivity in the red end of the spectrum similar to that in Fig. 6 appears (Flamant & Stiles, 1948). They showed that in this case the cone sensitivity is higher in the red end of the spectrum than that of the rods, even at the absolute threshold. A similar explanation is possible for the results of Fig. 6, but a number of considerations make this interpretation unlikely. In the first place, the directional sensitivity of the cones (which in the case of the Flamant & Stiles result was sufficient to elevate the cone threshold in red light above

17 494 M. ALPERN AND F. W. CAMPBELL that of the rods) is already acting in the blind spot experiments to weigh heavily in favour of rod responses. Secondly, this experiment was carried out on a protanomalous observer (A. S.) whose photopic spectral sensitivity in the red end of the spectrum is far below that of the normal eye (approximately 0-8 logl0 units at A = 625 m,). Only a very small amount of erythrolabe can be measured in his fovea (W. A. H. Rushton, personal communication). But his pupiliary spectral sensitivity for blind-spot stimulation shows as prominent an elevation in the red end of the spectrum as that found in observers with normal colour vision (Fig. 6A). Measurements of his pupillary spectral sensitivity with equivalent colour substitution methods, however, are much reduced (about 0 4 logl0 unit at 625 m,) in the red end of the spectrum compared to the normal. These observations seem to rule out the possibility that the deviations in Fig. 6 are due to cones, and thus suggest that the discrepancies between the pupillary measurements of the blind-spot spectral sensitivity and the C.I.E. scotopic curve may be due to a photostable coloured filter through which the light passes before excitation of the retinal rods. While it is not possible to be specific as to the way in which diffusion of stray light through the globe brings about such a distortion of the rods spectral-sensitivity curve, it seems likely that halation from structures behind the retina may be an important factor to be considered (Campbell & Alpern, 1962). Certainly scatter in the eye media is practically independent of wave-length in the relevant range (Boettner, personal communication; De Mott & Boynton, 1958). PART IV Reflexion coefficient of the fundus In order to relate the distortions from the C.I.E. scotopic curve (Fig. 6) to the coloured filter action of the fundus, its reflexion coefficient for light of various wave-lengths in the spectrum has been measured. METHODS The technique employed was very similar to that used by Brindley & Willmer (1952) to measure the amount of macular pigment in the living eye. The subject observed the light emerging from a monochromator directly. Between the subject's eye and the monochromator exit slit was a glass plate, so arranged that it reflected some of the light emerging from the monochromator into a roof prism through a neutral wedge, and thence into the experimenter's eye. This served as a comparison beam by means of which the experimenter was able to make a visual comparison with the light emerging from the subject's fully dilated pupil, which the experimenter could view by reflexion at the glass plate. The fundus reflexion coefficient could then be computed for each wave-length. Measurements were made in the inferior temporal region of the retina 27 from the fovea.

18 PUPILLARY SPECTRAL SENSITIVITY 495 RESULTS The computation of the reflexion coefficient from measurements of this kind has been described in detail by Brindley & Wilimer (1952). Briefly, the reflexion coefficient is the ratio of the luminance of the fundus to the retinal illuminance, when the quantities are measured in appropriate units. The former quantity could be determined by multiplying the luminance of the exit slit and the square of the wedge transmission required for equation of the brightness of the fundus and the comparison field (doublewedge traverse). The retinal illuminance would be computed by finding the product of the exit slit luminance and the solid angle which the exit pupil of the eye subtends at the retina. The losses in the optical system were measured by replacing the subject's eye with a second roof prism and a calibrated neutral filter. The reflexion coefficient values computed in this way (illustrated in the middle graph of Fig. 7 as the interrupted line) neglect any losses by absorption in the ocular media, as the light traverses the eye twice. In order to know that fraction of light reaching the retina which is reflected from it, these raw measurements must be corrected for the losses by absorption during a double passage through the media of the eye. These corrected values, in which the transmission characteristics of the media of the eye measured by Ludvigh & McCarthy (1938) are used, are presented in the middle graph of Fig. 7 as the solid line and the open circles. In Fig. 7 the top graph shows the deviations of the results in Fig. 6 from the C.I.E. scotopic curve. The smooth curve drawn through these points represents the general trends of the results as estimated by eye. This same curve has also been drawn through the measurements of reflexion of the fundus in the middle graph of Fig. 7. The agreement between the two sets of measurements is reasonably good. Such differences as persist are probably due to a composite of other factors, including measurement errors, individual differences and the invalidity of applying the data of Ludvigh & McCarthy to the eye under consideration. The human scotopic electroretinogram (e.r.g.) poses problems very similar to those presented by the data in Fig. 6, in so far as stray light in the eye is concerned. If one attempts to compare the spectral sensitivity of the scotopic e.r.g. to the C.I.E. scotopic curve, similar discrepancies appear. The bottom curve in Fig. 7 illustrated results from the dark-adapted eye of two observers of Armington & Thiede (1954). The solid lines in this graph are the measurements of reflexion coefficient from the middle graph of the same figure. It is evident that the agreement is rather good although the e.r.g. data show an even more prominent increase in sensitivity in the red end of the spectrum than the reflexion values predict. In the case of the e.r.g. the relation between the deviations from the

19 496 M. ALPERN AND F. W. CAMPBELL A 0 UCL 0 U) u I U BEo 0) u 0L.) uj LLi ZE MUJ 0 CL 0 - ba 0 -j C 4)A 0._ x 0 W J Fig. 7. Wave-length (mu) For legend see opposite page.

20 PUPILLARY SPECTRAL SENSITIVITY 497 visual-purple absorption curve have already been quite directly related to the reflexion characteristics ofthe fundus. For example, in the albino rabbit a very prominent 'red' hump appears which is absent in the pigmented rabbit (Dodt & Walther, 1958). Hagins's (1958) measurements of the reflexion coefficient of the rabbit fundus indicate that the human fundus is somewhat more reflective than that of the pigmented rabbit, and somewhat less than that of the albino rabbit. Moreover, the albino rabbit's pronounced red hump may be removed by injecting a blue eye into the vascular system (Dodt & Walther, 1958). Finally, Dodt, Copenhaver & Gunkel (1959) found in the human (photopic) e.r.g. that the prominence of the red hump could be directly related to the fundus reflectivity. In blonds, with little melanin and a bright fundus, a very large red hump was evident, but in heavily pigmented persons no very prominent hump could be demonstrated. All these observations are consistent with the suggestion that the discrepancies of the results in Fig. 6 from the C.I.E. curve are in some way related to the reflexion characteristis of the back of the eye. None the less, examination of the implications of this idea soon reveals a number of difficulties. Theoretically the reflexion characteristics of the fundus should have the same colour as the deviations from the C.I.E. scotopic curve, but it is surprising that the two sets of results seem to indicate the same density of filter as well. Further work is needed to elucidate this coincidence of density. In the second place, the e.r.g. and steady-state pupil data were both obtained under stimulus conditions in which the focal light passes through large areas of functioning retina before it is reflected at all; yet discrepancies from the scotopic curve in these cases are not obviously different from those obtained for the pupil when the stimulus is on the blind spot (Campbell & Alpern, 1962). But the initial traverse through functioning retina must be of some importance, if for no other reason than that this incident beam contains more light than the reflected ones. The low values of the reflexion coefficient even in the red end of the spectrum Legend to Fig. 7 Fig. 7. A. Deviations of the pupil results in Fig. 6A and B from the C.I.E. scotopic spectral sensitivity curve. The smooth curve shows the trends of the results as estimated by eye. B. Measurements of the reflexion coefficient of the fundus for an observer with blue eyes. Interrupted line, uncorrected for absorption by the media of the eye; open circles, 0, corrected for the absorption in the eye media (double passage) by the data of Ludvigh & McCarthy (1938). The smooth curve is from Fig. 6A vertically shifted in each case to show optimum agreement. C. Solid lineopen circle curve from Fig. 6B shifted vertically for optimum agreement with the deviations from the C.I.E. scotopic spectral sensitivity data (25 1V criterion) of Armington & Thiede (1954); 3, the b-wave of the e.r.g. of the dark-adapted eye. Mean results from two observers.

21 498 M. ALPERN AND F. W. CAMPBELL make it highly unlikely that the discrepancies in Fig. 6 result from multiple reflexions from the fundus. Why does halation play such an important part in pupillary motion and the e.r.g., but not in vision? Evidently the e.r.g. and the pupil do not have the same ability to differentiate between focal and scattered light that the visual mechanism does, but even this is not the entire answer, since even blind-spot results show smaller deviations from the C.I.E. curve for vision than for the pupil. Still another difficulty becomes evident when one asks what the colour of the fundus represents. The measurements in Fig. 7 do not reveal certain maxima (575 and 540 mp) or minima (560 and 510 m,u) characteristic of oxyhaemoglobin, even though the method is sensitive enough to detect them (and does so in the fundus of the albino rabbit). Undoubtedly melanin, rhodopsin and perhaps other substances acting merely as screening pigments can also contribute to the shape of the reflexion coefficient curve. The solutions to such problems should become more obvious once a better understanding is at hand.of the physics of halation within the fundus, including a precise knowledge of the location of the reflecting surface or surfaces and many of their other properties. Without this it seems pointless to attempt to build a more quantitative explanation for the way in which the colour of the fundus contributes to pupillary spectral sensitivity, although there can be little doubt that it does so. PART V Photopic response While the results described in Part I strongly suggest that cones do contribute to the response of the pupil to light, a spectral sensitivity curve of the light reflex which is at all similar to the photopic visibility curve has never been demonstrated. From the results which have already been described in this paper it seemed apparent that the reason for this was that the stray light in the eye always resulted in the addition of rod responses even when the light was focused exclusively on the fovea. If this inference is correct it should be possible to obtain a pure cone response if the rod contribution could be masked. If, instead of presenting the foveal stimulus to the dark-adapted eye, the parafoveal rods could be made insensitive by keeping them adapted to a bright blue light acting as a background upon which the foveal test stimulus was exposed, it should be possible to obtain a purely photopic pupillary spectral sensitivity curve. The following experiment was undertaken to verify this prediction.

22 PUPILLARY SPECTRAL SENSITIVITY 499 METHODS The apparatus was the same as that described for Part I, except that only the stimulus beam traversing the monochromator was used. A blue background was provided by mounting a 2 mm diameter polished steel sphere at the anterior focal plane of the eye, and illuminating it by tungsten light passed through an Ilford 622 filter (S3). The sphere was mounted as close to the line of sight as possible without interfering with the Maxwellian view of the stimulus light. The subject's pupil for this eye was well dilated with 2 % homatropine. The arrangement provided a background field whose angular diameter was defined by the entrance pupil size divided by the focal length of the eye. The subjects were able to position themselves in such a way that the small monochromatic stimulus light was seen virtually in the centre of this background field. The retinal illuminance of this background was estimated by binocular matching with a S.E.I. photometer fitted with a 1-5 mm diameter artificial pupil. To simplify the detection of a resulting pupil contraction the X axis of the oscilloscope was triggered by means of a photocell which monitored the light flash. The Y axis displayed the pupil size. The speed of scan in the X axis was adjusted so that the reaction time of the pupil response could readily be seen. The experimenter, by rejecting all iris contraction that did not commence at the usual reaction time, could thus detect small pupil movements which were correlated with the stimulus. The use of a cathode-ray tube with a long-persistence phosphor permitted the experimenter sufficient time to observe each response. RESULTS The results of these experiments on two observers are illustrated in Fig. 8. The curve forms a peak at approximately 560 m,u and is in reasonable agreement with the photopic spectral sensitivity of the C.I.E. standard observer (interrupted line in this figure). An even better agreement is found, however, when these pupillomotor spectral-sensitivity data are compared with the photopic luminosity curve of the same observers, measured on the same apparatus by flicker photometry (solid line). This similarity between psychophysical and pupillomotor measurements is striking and there can be little doubt that the photoreceptors inducing pupil changes under the conditions of this experiment were cones, and cones alone. To the dark-adapted eye stimulus flashes which are below pupillary threshold are still quite visible (Schweitzer, 1956). (For flashes including the entire visual field this is no longer true, and so De Launay's (1949) result was probably fortuitous for the size of the field (2 ) which he selected.) For the field sizes used in the present experiment (200 or smaller) light intensities at the pupillary threshold (or at the higher levels in the colour substitution experiments) excite photoreceptors outside the focal image by entopic scatter. For reasons enumerated in Part III these photoreceptors are mainly, if not exclusively, rods. This scattering, which (for reasons given in Part IV) is probably due to halation in the fundus, is important enough to produce some modification of the spectral sensitivity curve of

23 500 M. ALPERN AND F. W. CAMPBELL the pupil. While the characteristics of this process in the lving eye are not well understood, the experiments illustrated in Fig. 8 show that a background of about 150 in diameter sufficed to mask any effects from rod stimulation by scattering when the fovea was stimulated with a small 20 field. Hence the halation effects appear to be confined to a rather limited region around the test patch, although the possibility of further masking due to scatter of the light from the blue background outside of the 15 area cannot be excluded. These experiments, when considered together with 2.0 I II., 4, o Mean data D.H. and J.W Wave-length (mnu) Fig. 8. Mean spectral sensitivity curve for the pupil response of two subjects. 0, Differential threshold measurements are plotted for 2 sec flashes of a 2 test patch centrally fixated and seen against a continuous blue background approximately 15 in diameter, which produced a retinal illuminance somewhere between 100 and 200 td. Interrupted line, C.I.E. photopic luminosity curve; solid line, mean results of psychophysical measurements of photopic luminosity (flicker photometry) on the same two subjects with the same apparatus. the experiments in Part III, show that the relative role of focal and nonfocal light in evoking pupil responses is subject to a certain amount of experimental manipulation. In Part III the entire effect was due to scattered light, since there are no photoreceptors on the blind spot. In Fig. 8 the entire effect was presumably due to focal light, since no evidence of a rod response appeared when the retinal image of the test patch was confined to the rod-free fovea and the directional sensitivity and small

24 PUPILLARY SPECTRAL SENSITIVITY 501 summation area made cones relatively insensitive to light scattered by halation. The other methods of measuring the spectral sensitivity of the pupil response, such as equivalent colour substitution or a high-intensity threshold criterion, result in a curve (Fig. 4) which is neither purely photopic nor purely scotopic, and it is now time to find an explanation for the shape of this curve. PART VI Synthesis of rod and cone contributions The first possibility is that this curve is the consequence ofthe absorption of light by a single photosensitive pigment. Indeed, the shape of the curve and the wave-length at which the maximum occurs are similar to the absorption spectrum of a green photosensitive pigment, chlorolabe, which Rushton (1958) has measured in the living fovea. There are two reasons for believing that this is not the explanation for the equivalent substitution curve. In the first place the measurements on the fovea in the absence of a background give curves (Fig. 4) identical with measurements with larger areas (Am,a = 530 to 540 m,). But the presence of the blue background suffices to yield a photopic pupil curve for the fovea similar to, if not identical with, the photopic luminosity curve (Fig. 8). Secondly, measurements were made on a severely protanomalous observer, who has as much chlorolabe in his retina (although he is very deficient in erythrolabe) as does the normal: yet the method of equivalent substitution for the pupil of his eyes gives a curve which, when compared to the normal, is very remarkably reduced (about 0 4 log units at 625 m,) in the red end of the spectrum. This would not be expected if absorption of light by chlorolabe was the only photochemical reaction contributing to the action spectrum of his pupil. Since chlorolabe is the only visual pigment in the human eye which could possibly account for the results of Figs. 3 and 4 by itself, these reasons are sufficient evidence for rejecting the hypotheses that these curves are a consequence of (a) the decomposition by light of only one photosensitive substance and of (b) the activity of a single class of photoreceptors as well. But if more than one class of photoreceptors contribute to this combined response, what are these classes? As a first step, a reasonable assumption would be that the curves in Figs. 3 and 4 are obtained by the combination of the contributions of two classes of receptors. The action spectra of these two classes are those measured by the two limiting conditions of the present work, namely, the C.I.E. scotopic curve when modified by the smooth curve drawn through the top graph in Fig. 7 on the one hand, and the fovea luminosity curve on the other. For convenience,

25 502 M. ALPERN AND F. W. CAMPBELL these two action spectra will be referred to as representing those of rods and cones respectively, even though the former has the absorption spectrum of visual purple considerably distorted by the colour of the fundus and the latter may well be obtained by some form of combination of three (or more) different kinds of cones. In what way do the contributions from the rods and cones, as defined above, combine in order to synthesize the mixed spectral sensitivity curve of Figs. 3 and 4? Three different possibilities will be considered: 1. The sensitivity for the red end of the spectrum is determined exclusively by cones, that for the blue end of the spectrum exclusively by rods. Bridgman (1953) has used this method to explain the mesopic psychophysical spectral sensitivity curves. He allows for a certain amount of (unspecified) interaction of rods and cones in the neighbourhood of the wave-lengths where the two curves intersect. 2. The next method is to regard the measured spectral sensitivity (EA) as determined by the simple arithmetic sum of the individual rod (CA) and cone (#A) contributions. That is EA = ata +boa, (1) in which a and b are weighting factors. 3. A third method regards the measured sensitivity at each wave-length as the weighted mean of the logarithms of the individual rod and cone sensitivities at that wave-length. According to this, (a +f) log EA = a log CA +logo0 +K, (2) in which a and,b are weighting factors and K is a constant. Rushton (1959) has used exactly this method to describe the mesopic spectral sensitivity curves of the frog's ganglion cell. It is based on the fact that the physiological response to light stimulation becomes related to the logarithm of light intensity very soon after the decomposition of the photosensitive pigments (Donner & Rushton, 1959a). It should be emphasized that, according to Rushton's hypothesis, eqn. (2) is valid only for the range of intensity levels where physiological consequences of rod and cone stimulation are each linearly related to the logarithm of the light intensity. Whenever this method is discussed below it will be assumed that the intensity level of relevance is within this range. The first step in evaluating these different possibilities was to fit the empirical results with theoretical curves predicted by the different methods. In this the second and third methods were clearly more successful than the first and the curves predicted by these latter have been drawn in Fig. 3, the dotted line being drawn according to eqn. (1) the solid line according to eqn. (2). The agreement between the empirical values and

26 PUPILLARY SPECTRAL SENSITIVITY 503 those predicted by either theory is less satisfactory than one might wish. In evaluating this, however, it should be remembered that these theoretical curves are made very approximate by the assumption of the nature of the underlying rod and cone systems, both of which were subject to experimental errors. While the theoretical curves tend to favour the third method in preference to the other two, and the second method in preference to the first, it is doubtful whether a decision about the matter based only on fitting these curves is of much value. Consequently, two other attempts to differentiate between these possibilities were tried, but only the second was successful in providing an unequivocal answer. The unsuccessful attempt consisted in adopting for the photopupillary response, a modification of an experiment that Weaver (1949) carried out on psychophysical brightness matches. The dark-adapted eye was presented with two lights, one of A = 650 m,u emerging from the monochromator, the other of tungsten light attenuated by an Ilford 621 filter. The blue beam was reflected from a thin glass plate into the red beam and the two focused on the pupil of the eye. The subject saw a field 200 in diameter and fixated its centre. A neutral wedge was then placed in the common path. Each beam was polarized in opposing directions and the intensities of the two were adjusted by adding neutral filters in the blue beam (in a preliminary experiment) in such a way that the amount of light in the two beams were at threshold for the pupil. A second polaroid in the common beam could then be quickly adjusted in such a way that the light entering the eye was either 100 % blue, 100% red or 50 % blue and 50 % red (in practice this mixture was 44-5 % blue and 55.5 % red because of the difference in wedge density for the two colours). The subject was then exposed to a one-second flash of light of one of seven different intensities of the common wedge and one of the three spectral compositions, and the experimenter decided whether or not a response of the pupil occurred to the flash. A period of 1 min was allowed for recovery of dark-adaptation before the next flash was exposed, and so on. In all, the experiment was repeated 15 times for each intensity and spectral composition. If this criterion of response was determined by a process of synthesis described by eqn. (1) then one would expect to find no difference between the wedge setting at threshold for 100 % blue and that for the mixed blue and red. On the other hand, eqn. (2) leads to the expectation that, under the conditions of this experiment (ac =,), the wedge density for the 100 % blue would be 0*24 log10 units larger than would be the setting for the mixture, provided the red and blue beams were exactly equated for threshold. The data were analysed by probit analysis (Finney, 1947) but 33 Physiol. 164

27 504 M. ALPERN AND F. W. CAMPBELL the results were not completely satisfactory, since only the blue and the mixed responses could be reasonably approximated by parallel probit lines. Hence one cannot be certain that the thresholds for the pure red and pure blue stimuli were exactly the same (although they must have been nearly so). Comparison of the wedge densities for 50 % response of the pure blue and the mixed cases shows that the setting for blue was in fact denser than the wedge setting for the mixed stimulus. The difference in these medians, ± 0*13 (standard error), was almost half way between the values predicted by eqns. (1) and (2). However, the 5 % fiducial limits of this difference were and - 0*12, from which it is evident that the experiment was not precise enough to allow any decisions between the alternative additivity principles represented by eqn. (1) and eqn. (2). The final attempts to decide between these various methods of synthesizing the mixed curve were based upon the fact that both the first two methods imply that the response at threshold to red light is largely (or exclusively) determined by cones and the response to blue light is largely (or exclusively) determined by rods, where the third method implies that the response at every wave-length at thresholds is determined by a fixed proportion of rod and cone contributions (in the case under consideration about 1: 1). Since the cones have a marked directional sensitivity and the rods do not (Flamant & Stiles, 1948; Donner & Rushton, 1959b), comparing thresholds for red and blue light directed through the edge and through the centre of the pupil should give different answers depending upon the way the mixed curve is synthesized (Alpern & Campbell, 1962). The ratio (R) of the sensitivities for peripheral and central entry of the pupil according to methods one and two would be 1-0 in the case of blue light and, in the case of red light. (,u is the extent to which the cone threshold is raised by causing the light to pass through the edge of the pupil, compared to the threshold at the centre.) According to method three, however, the value R at every wave-length would be,ufi'+). If one determines the pupillary threshold for a rather large field (200 centrally fixed), when the light enters through the centre of the pupil it will be about one half logarithmic step lower than when the light enters through the edge of the dilated pupil. This confirms the idea that cones contribute to the pupillomotor threshold under these conditions. For red light (650 m,u) the difference between the threshold for central and peripheral pupil entry was 0f (s.e. of mean) log. units, a value which differed from zero at a significant level (P < 0-01; simple sign test). In the case of blue light (450 mu) the threshold for peripheral pupillary entry was 0* log. units higher than the threshold for central entry, a value which again differed significantly from zero (P < 0.02). It was impossible,

28 PUPILLARY SPECTRAL SENSITIVITY 505 however, to find any significance at all for the difference between the threshold differences for the two colours. This result would not be expected at all, provided the rod and cone contributions synthesized the mixed curve by either method 1 or method 2; it is precisely what is to be expected if the rod and cone contributions were mixed according to eqn. (2), method 3. To be certain of this conclusion, a supplementary and somewhat more sensitive experiment has been carried out, which is based on the result of the previous experiment, namely that light going through the edge of the pupil is less effective in evoking a pupillary response than light going through its centre. The question is, can one detect any difference at all between red and blue lights in this regard. In order to detect very small differences the method of equivalent substitution was again adopted. The intensity of the red light required to make an equivalent colour substitution with blue light was measured both when light entered the centre and when it entered the edge of the widely dilated pupil. Now if the responses at the long and short wave-lengths at the null of the substitution are each made up of the same amount of rod and cone contributions (logarithmic addition) there will be no difference between the wedge setting at the null, regardless of where the stimulus beam enters the dilated pupil. On the other hand, if the response to red light originated mostly in cones, while that to blue originated mostly in rods at the null point in substitution (linear addition), then the wedge in the red beam should be about 0 5 density units larger when the stimulus beam enters the centre than when it enters the edge of the dilated pupil. This experiment proved to be sensitive enough for it now to be necessary to be certain that the cones had identical directional sensitivity for the specific red and blue lights used. This was very nearly so (as judged by visual matching) if the red light from the monochromator was 660 mu and the blue light supplied by the tungsten lamp was filtered with an Ilford 622 filter (A. = 480). Measurements were then made for light of these two colours going through the centre, the edge, and then the centre again, of the fully dilated pupil, for a 200 field fixated at the centre. The mean wedge setting for the red beam (10 measurements in each position) were: (s.e. of mean) (central pupil entry), (peripheral entry) and (central entry). None of these values differ significantly from the others. Thus these results confirm those of the previous experiment and the conclusion to be drawn from both of them is that the mixed spectral sensitivity curve of the light reflex is determined by the combination of the weighted mean of the logarithm of the sensitivities of the rod and cone responses, in exacly the same way as the mesopic spectral sensitivity of the ganglion cell of the frog's retina (Rushton, 1959). 33-2

29 506 M. ALPERN AND F. W. CAMPBELL SUMMARY 1. The spectral sensitivities of the consensual pupil light reflex were determined with an infra-red photo-electric pupillometer for threshold and for equivalent colour substitution (transient stimuli) and with infra-red photography (for steady-state stimuli). 2. The spectral response curve usually obtained (transient stimulation), if the intensity of the standard is at photopic level, or the threshold criterion is large, is neither purely photopic nor purely scotopic. It has a A = 530 to 540 m,u. 3. For a steady-state stimulation and a large change in pupil diameter, the spectral sensitivity curve obtained is quite similar to the curve obtained by equivalent colour substitution. 4. Such curves are not the consequence of the absorption by light of a single photosensitive material. 5. These curves can be synthesized by obtaining the weighted means of the logarithms of the rod and cone sensitivities, and by obtaining the weighted sum of rod and cone sensitivities. The fact that light going through the edge of the dilated pupil is as effective in evoking a pupil response at threshold, whether it is red or blue, although each is less effective through the edge than it is through the centre of the pupil, supports the logarithmic additivity process. 6. In the cases studied the relative contributions of rods and cones were always about equal. 7. When the test patch is confined to the blind spot (transient stimulation) the spectral response curve obtained is a rod curve modified by scattered light. 8. Similarly, using steady-state stimulation and measuring the intensity required to produce a small (0.5 mm) change in pupil size, also resulted in a rod spectral response curve modified by scattered light. 9. The discrepancy between such curves and the C.I.E. scotopic spectral sensitivity curve corresponds to the colour of the fundus. This suggests that the scatter comes about by halation in the fundus. 10. When the effects of halation are avoided by an intense blue background, which masks the effects of rod stimulation, foveal spectral sensitivity measurements of the pupil agree quite well with photopic (i.e. cone) visibility measurements, but not otherwise. 11. The size of the background required to avoid any obvious effects of scatter coming into a stimulation by a 20 foveal target is only about 150, and may be even smaller than this. We are indebted to W. A. H. Rushton, F.R.S., W. S. Stiles, F.R.S., and H. B. Barlow for many helpful suggestions in various stages of this work. A. Safir, D. Hamasaki and

30 PUPILLARY SPECTRAL SENSITIVITY 507 J. Wales, among others, endured many long hours as patient observers. Mr Roy Rumble provided unfailing assistance. The Royal Society and the W. H. Ross Foundation provided funds for equipment. REFERENCES ALPERN, M. & BENSON, D. J. (1953). Directional sensitivity of the pupillomotor photoreceptors. Amer. J. Optom. 30, ALPERN, M. & CAMPBELL, F. W. (1962). The physiology of mixed rod and cone spectral response curves. J. Physiol. 164, 12-13P. ALPERN, M., KITAI, S. & ISAACSON, J. D. (1959). The dark-adaptation process of the pupillomotor photoreceptors. Amer. J. Ophthal. 48, (Nov., Pt. 2), ARMINGTON, J. C. & THIEDE, F. C. (1954). Selective adaptation of components of the human electroretinogram. J. opt. Soc. Amer. 44, BOUMA, H. (1962). Size of the static pupil as a function of wave-length and luminosity of the light incident on the human eye. Nature, Lond., 193, BRIDGMAN, C. S. (1953). The luminosity curve as affected by the relation between rod and cone adaptation. J. opt. Soc. Amer. 43, BRINDLEY, G. S. & WILLMER, E. N. (1952). The reflexion of light from the macular and peripheral fundus oculi in man. J. Physiol. 116, CAMPBELL, F. W. & ALPERN, M. (1962). The pupillomotor spectral sensitivity curve and the colour of the fundus. J. opt. Soc. Amer. 52, DE LAUNAY, J. (1949). A note on the photo-pupil response. J. opt. Soc. Amer. 39, DE MOTT, D. W. & BOYNTON, R. M. (1958). Retinal distribution of entopic stray light. J. opt. Soc. Amer. 48, DODT, E., COPENHAVER, R. M. & GUNKEL, R. D. (1959). Electroretinographic measurements of the spectral sensitivity in albinos, caucasians and negroes. Arch. Ophthal., N.Y., 62, DODT, E. & WALTHER, J. B. (1958). Spektrale Sensitivitat und Blutreflexion. Pflug. Arch. ges. Physiol. 266, DONNER, K. 0. & RUSHTON, W. A. H. (1959a). Retinal stimulation by light substitution. J. Physiol. 149, DONNER, K. 0. & RUSHTON, W. A. H. (1959b). Rod-cone interaction in the frog's retina analysed by the Stiles-Crawford effect and by dark adaptation. J. Physiol. 149, FINNEY, D. J. (1947). Probit Analysis, pp Cambridge University Press. FLAMANT, F. & STILES, W. S. (1948). The directional and spectral sensitivities of the retinal rods to adapting fields of different wave-lengths. J. Physiol. 107, HAGINS, W. A. (1958). Rhodopsin in a mammalian retina. Ph.D. thesis, University of Cambridge. HESS, C. (1908). Untersuchungen zur Physiologie und Pathologie des Pupillenspieles. Arch. Augenheilk. 60, LAURENS, H. (1923). Studies on the relative physiological value of spectral lights. III. The pupillomotor effects of wave-lengths of equal energy content. Amer. J. Physiol. 64, LUDVIGH, E. & MCCARTHY, E. F. (1938). Absorption of visible light by the refractive media of the human eye. Arch. Ophthal., N.Y., 20, RUSHTON, W. A. H. (1958). Kinetics of cone pigments measured objectively on the living human fovea. Ann. N.Y. Acad. Sci. 74, RUSHTON, W. A. H. (1959). Excitation pools in the frog's retina. J. Physiol. 149, SCHWEITZER, N. M. J. (1956). Threshold measurements on the light reflex of the pupil in the dark adapted eye. Docum. ophthal. 10, SCHWEITZER, N. M. J. & BOUMAN, M. A. (1958). Differential threshold measurements on the light reflex of the human pupil. Arch. Ophthal., N. Y., 59, SPRING, K. H. & STILES, W. S. (1948). Variation of pupil size with change in the angles at which the light stimulus strikes the retina. Brit. J. Ophthal. 32, WAGMAN, I. H. & GULLBERG, J. E. (1942). The relationship between monochromatic light and pupil diameter. The low intensity visibility curve as measured by pupillary measurements. Amer. J. Physiol. 137, WEAVER, K. J. (1949). A provisional standard observer for low level photometry. J. opt. Soc. A4mer. 39,