Functional Role of Efferents to the Avian Retina I. ANALYSIS OF RETINAL GANGLION CELL RECEPTIVE FIELDS1 ALAN L. PEARLMAN 2 AND CHARLES P. HUGHES 3 Department of Neurology and Neurological Surgery ( N e u r o l o g y ) and Department of Physiology and Biophysics, W a s h i n g t o n University School of Medicine, S t . Louis, Missouri 63110
ABSTRACT Receptive fields of retinal ganglion cells were analyzed during extracellular microelectrode recordings in the optic tract of the lightly anesthetized pigeon. Four major types of receptive field can be distinguished among the 359 fibers studied. Twenty-five percent of the receptive fields are relatively simple, responding at on and at off to stationary spots of light in the central region. All of the receptive fields have inhibitory surrounds of varying strength that do not produce a response when illuminated alone, but antagonize responses from the central region. Motion sensitive units comprise 15% of the recorded population; they are similar to the o n d f center type except that responses to stationary stimuli are absent or very weak while responses to moving stimuli are vigorous. Directionally selective units also have the basic features of on-off, inhibitory surround cells, but respond to moving stimuli well from the preferred direction and not at all from the null direction. Directional cells have a broad range of null directions; in about one-third of the units the range becomes broader when the stimulus involves both center and surround of the receptive field, thus enhancing directional selectivity. Directionally selective units are common, comprising 38% of the units studied. Cells unresponsive to stimuli moving from anterior in the visual field are much more common than other types, while cells unresponsive to stimuli from posterior in the field are rare. A few units (11%) respond only at on or at off to stationary stimuli in their receptive field centers; they also have antagonistic but unresponsive receptive field surrounds. The area of the visual field sampled is uniform in regard to the relative numbers of the four major receptive field types. Centrifugal fibers to the retina were first described by Cajal (1889) in the avian retina; the bird retina remains the only well substantiated example of a vertebrate retina receiving efferent fibers (Ogden, '68; Cowan, '70). Although efferent fibers have been described in the retinas of the fish (Witkovsky, '71), the dog (Cajal, 1894), the chimpanzee (Polyak, '57), the cat and monkey (Brooke et al., '65), and the human (Honrubia and Elliott, '68), a great deal of controversy continues to surround the question of the exsistence of efferent fibers in these species. The nucleus of origin of the centrifugal fibers in the bird has been clearly demonstrated; no such demonstration is available for other vertebrates. Efferent fibers to the retina in the bird arise in the isthmo-optic nucleus, a distinct cell mass in the caudal midbrain, (WallenJ. COMP. NEUR.,166: 111-122
burg, 1898; Cowan and Powell, '63) and form the final link in an anatomical closed loop. Each point on the retina projects to a restricted area of the tectum and the projection from tectum to isthmo-optic nucleus is also organized in a distinct retinotopic manner. The isthmo-optic nucleus in turn projects back to the retina in a retinotopic fashion, each quadrant of the nucleus sending fibers to the same quadrant of the retina that provides its input via the tectum (McGill et al., '66a,b). The terminations of the efferent fibers 1 This investigation was supported by NIH Research Grant R01-EY-00621from the National Eye Institute. A preliminary report of this work was published earlier in abstract form (Pearlman and Hughes, '73). 2 Gordon R. and Thelma B . Coates Scholar in Neurology, Washington University. 3 Supported by NIH Special Fellowship F11 NS 02437 from the National Institute of Neurological Diseases and Stroke.
111
112
ALAN L. PEARLMAN A N D CHARLES P. HUGHES
to the retina have also been carefully studied in the bird. Cajal (1889) and Dogiel (1895) demonstrated that these fibers terminate on amacrine cells along the inner aspect of the inner nuclear layer. Maturana and Frenk ('65) extended these observations and described two distinct patterns of efferent terminations: (1) a divergent type in which the fibers branch and contact several different cells and (2) a convergent type, in which the fibers also branch but after the branches separate they ultimately converge to form a pericellular nest near the soma of one or a few cells. There are an estimated 100,000 centrifugal axon terminals in the pigeon retina; since there are about 12,000 fibers in the isthmo-optic tract each isthmo-optic tract fiber gives rise to eight to ten terminals (Cowan, '70). The efferent terminals are quite distinct; they form the largest synaptic endings in the junction of inner plexiform and inner nuclear layers, and contain tight aggregates of synaptic vesicles and many mitochondria. These endings are found on the proximal portion of the main processes of the amacrine cells, and occasionally on the cell soma (Dowling and Cowan, '66). In the present study a neurophysiological analysis of this anatomically well described system has been carried out in two parts. The first is a study of the receptive field properties of individual pigeon retinal ganglion cells; it extends the observations of earlier reports (Maturana, '62; Maturana and Frenk, '63; Holden, '69) and is in essential agreement with the recent work of Miles ('72) in the chick. The second part (Pearlman and Hughes, '76) is concerned with the effects of reversible removal of efferent influences emanating from the isthmo-optic nucleus on the physiologic properties of retinal ganglion cells.
of small receptive fields over 1-2 hours. Initially the animals were ventilated with a respiratory pump (Harvard Apparatus) and the percentage of COz in the expired air measured with an infrared COz meter (Beckman Instruments). In later experiments the unidirectional airflow system of bird respiration (Burger and Lorenz, '60) was substituted, consider ably reducing brain pulsations. A mixture of nitrous oxide (40 % ), oxygen (55 % ), and COB(5 % ) was humidified and passed in a steady stream (1,000 cc/mm) into the trachea and out through a hole in a major abdominal air sac. The gas mixture and flow rate were determined by serial measurements of arterial p02, pCOz, and pH on two animals over several hours. Body temperature was monitored by a thermistor deep in the pectoral muscle, and maintained by circulating warm water in a heating pad (Gorman-Rupp). The EKG was recorded at varying intervals on a polygraph. The bird's eye was refracted at 30" with a slit retinoscope, and fitted to within kO.5 diopters with a contact lens. The bird's head was placed in a stereotaxic head holder built to conform to the co-ordinates of the Karten and Hodos ('67) atlas for the pigeon. The head holder was designed so that it blocked only a small portion of the most posterior visual fields. Electrode carriers (Kopf) were mounted on a bar parallel to the ear bars behind the bird's head, so that they were not in the field of view. In early experiments fibers were recorded in the optic chiasm or tract by stereotaxic placement of tungsten or platinum microelectrodes. Electrode placements were subsequently checked in serial frozen or paraffin sections. In later experiments the left cerebral hemisphere was removed acutely by suction to permit direct visualization of the optic tract, thereby greatly enhancing the yield in a given exMETHODS periment. There were no detectable differAdult White Carneaux pigeons (Columba ences in the units recorded in the two livia, Palmetto Pigeon Plant, Sumter , S.C.) preparations. were anesthetized with halothane for surExtracellular action potentials of single gery and then maintained with nitrous optic tract fibers were recorded convenoxide-oxygenmixtures. They were paralyzed tionally, displayed on an oscilloscope, (Tekwith intramuscular curare (7.5 mg initially tronix 565) and monitored through an and 1.5-3 mg every three hours thereafter) audio amplifier and loud speaker. Ampliand artificially respired. With curarization, fied action potentials were led into a winno eye movement could be detected by the dow discriminator that produced a brief method of repeatedly plotting the borders rectangular pulse for each spike falling
AVIAN RETINAL GANGLION CELL RECEPTIVE FIELDS
within the variable upper and lower limits of the window. Action potentials, window limits and window output were monitored on an oscilloscope (Tektronix 564B) and checked frequently to be certain that only the action potentials of the unit under study were producing a pulse at the window output. The brief rectangular pulses provided the input to a LINC computer (Computer Systems Laboratory, Washington University) programmed to produce post-stimulus time histograms. Histogram collection was initiated when the stimulus beam moved across a photo-electric sensor mounted on the tangent screen near the receptive field. Histograms were photographed from an oscilloscope face, and also stored on magnetic tape for later analysis or for display on an x-y plotter. Receptive fields were mapped on the tangent screen with a projector (Leitz Prado) mounted on a tripod and moved by hand. More precise moving stimuli were provided by another projector whose beam was directed at a front-silvered mirror mounted on a pen-motor (Brush-Clevite). The pen motor was mounted in turn on a mechanical swivel that allowed rotations of the motor through 360 *, providing moving stimuli from any direction. The mirror motor was driven by a trapezoid generator with variable ramp slope, ramp duration, and repetition rate. An electronic shutter on the projector was opened and closed at appropriate times in the stimulus sweep on signal from the computer. RESULTS
The receptive field properties of 359 retinal ganglion cells were characterized during microelectrode recordings from single units in the optic tract or chiasm of 69 pigeons. Since the types of receptive field identified are at least in part a function of the stimuli employed in searching for units, it is important to specify the search procedures in some detail. A distinct multiunit response was present when a flashlight beam was shown across the retina. This was useful for indicating a functioning electrode, but since so many units responded it was of no value in indicating the presence of cells sufficiently isolated for proper study. A similar but smaller multi-unit response was often evident when bars of light, dark bars, or more complex
113
objects moved across a specific part of the visual field. Once such an area was located, the electrode was advanced slowly while the area was being stimulated by light and dark bars of varying sizes, both stationary and moving, until a single unit was clearly isolated above the background. Spontaneous activity was not a prominent feature of retinal ganglion cells in our preparation, although roughly 50% had some spontaneous firing, and a small percentage fired vigorously when unstimulated. The presence or absence of spontaneous activity could not be correlated with any of the specific receptive field types outlined below. The receptive field of each isolated unit was studied with stationary squares or rectangles of varying size and orientation turned on and off, with light and dark bars of varying size moved through the receptive field from each of eight directions at 45" intervals, and colored stimuli of the same type produced by placing red, green, and blue Wratten filters in the projector beam. Receptive fields defined with these stimuli can be distributed into four major classes, with several other types occurring less frequently. h - o f f center, inhibitory surround units In the recorded population of optic tract axons, 25 % have relatively simple receptive fields with central regions that respond transiently to small spots of light when turned on and when turned off. The central region is always surrounded by an antagonistic zone that inhibits the center response when the stimulus extends into it, but does not produce a response when illuminated by itself. The inhibitory surround region is quite variable in the strength of its effect from one unit to another; a stimulus that extends only a short distance into the surround may produce profound inhibition in one cell, while a stimulus encompassing the entire inhibitory surround may produce only slight inhibition in another. A typical on-off center, inhibitory surround unit is displayed in figure 1. The receptive field is shown on the left with the responsive central region outlined by a solid line; the f signs indicate responses at on and at off to a small stationary test spot. The open circles indicate no response to a small test spot. A bar of light (hatched
114
ALAN L. PEARLMAN AND CHARLES P. HUGHES
Fig. 1 On-off center, inhibitory surround unit. Receptive field on the left; on responses to small stationary spots shown as + , Off responses -, and no response as 0. The 1 calibration mark refers to both the receptive field and the bar of stimulus light, shown as hatched bars to the right, moving in direction indicated by arrow. Post-stimulus time histograms at bottom of figure show responses to leading and trailing edges of the stimulus bar. Each histogram is the sum of 15 sweeps of the bar across the field at 12.6"/sec,one sweep every 4.8 seconds.
rectangle) moving across the receptive field produces a transient response as the leading edge crosses the field, and another when the trailing edge crosses. The responses are indicated in the post-stimulus time histograms beneath each bar. The marked decrease in the response when the stimulus extends outside the receptive field center illustrates the effect of a strong inhibitory surround. Also evident is the fact that the on (or leading edge) response and off (or trailing edge) response are not always of equal amplitude. This feature is quite variable from cell to cell, with some cells responding more strongly at on, some at off, and some nearly equally to both. When the stimulus is reduced in intensity nearly to threshold for a given unit, the stronger of the two responses usually occurs exclusively. Most of the features described thus far for the on-off center, inhibitory surround cells are also present in the next two types of cell to be described, and therefore in about 80% of the cells studied; these two cell types differ only in their special features in response to moving stimuli. Motion sensitive units
Although all units respond to some degree to moving stimuli, we have somewhat arbitrarily divided off a group of units (15%) that respond far better to moving stimuli than to any others. The receptive
field maps of these units look identical to those of the on-off, inhibitory surround types described above, except that the responses to stationary stimuli are weak or occasionally entirely absent, while the responses to moving stimuli are vigorous. Responses to both leading and trailing edges of a moving bar are the rule, but again with variability in relative amplitude of the two responses from unit to unit. A small spot moving within the receptive field center is frequently a very effective stimulus, as is a small spot moving back and forth through the field. The orientation of the moving edge is unimportant. These units respond to a moving edge no matter in what direction it is moving, thus distinguishing them from the directionally selective cells to be described next. Directionally selective units
The largest single class (38%) of units encountered in this study are those exhibiting directional selectivity in their responses to moving stimuli. An example is shown in figure 2,where the receptive field is displayed in the center, and the poststimulus time histograms generated by a bar of light moving across it are shown around the circumference. Each histogram is displayed alongside an arrow representing the direction of movement of the bar; the bar was a rectangle whose short side was about the same size (2.6")as the short
AVIAN RETINAL GANGLION CELL RECEPTIVE FIELDS
0 O(-9O
0
0
115
0
-
0
0
L 1 0
liPi7 b
Fig. 2 Directionally selective unit.Receptive field in the center of the figure; on responses to small stationary spots shown as off as -, and no response as a circle; l o calibration refers to receptive field; 5-spike4.5 sec calibration refers to post stimulus time histograms elicited by 15 sweeps of a 2.6" X 6.6' bar moving parallel to its long axis in the direction indicated by adjacent arrow at 16.4O/sec, one sweep every five seconds.
+,
side of the field center. The long side of the rectangle was long (6.6') relative to the field center and was oriented parallel to the direction of movement in each case. Bar speed was constant (16.4'/sec), as was the repetition rate (one sweep every 5 seconds). The superior aspect of the bird's visual field is represented at the top of the figure, and the posterior aspect to the right; the receptive field was in the right eye.
For this unit stimuli moving from posterior-inferior are virtually ineffective, and those moving from anterior-superior are quite effective. A quantitative representation of directional selectivity for the same unit is shown in the polar plot of figure 3. Each point on a solid line axis represents the average number of action potentials produced by a sweep of the stimulus bar through the re-
116
ALAN L. PEARLMAN AND CHARLES P. HUGHES Superior
!
\
\ Posterior
----------
Small Bar Medium Bar Large Bar -I-...-------
12 spikes /sweep Inferior
Fig. 3 Polar plot of directional unit shown in figure 2. Points on a given axis are the average number of action potentials produced by 15 sweeps of a stimulus of the indicated size through the field center along that axis. Small bar: 0.66 X 6 . 6 O ; medium bar: 2.6O X 6.6": large bar: 6 . 7 " x 6.6O. Superior, posterior, etc. refer to the visual field of the bird's right eye.
ceptive field along that axis. For example, points on the superior axis represent the number of spikes produced by a stimulus sweep from superior to inferior. The null and preferred directions are again evident, but i t is also evident that the range of directions giving no response is related to the size of the stimulus bar. A small bar, smaller than the receptive field center, produces a good response when moved through the field along the 45" axis from posterior-superior to anterior-inferior; a bar of the same size as the field center (medium bar, fig. 3 ) is about half as effective from the same direction, and a large bar, involving the field surround, is virtually ineffective. The directional selectivity of the unit is thus enhanced for large stimuli, since the cell responds to large stimuli over a narrower range of directions than it does to small stimuli. About one-third of the directional cells in our sample had enhanced directiond selectivity to large stim-
uli. This phenomenon is also present in the receptive fields of chick retinal ganglion cells (Miles, '72), and has been carefully studied in the directional ganglion cells of the rabbit (Daw and Wyatt, '74; Wyatt and Daw, '75). When the entire group of directionally selective cells is analysed, it is apparent that the null directions of the receptive fields are not evenly distributed. More than one-third of directionally selective units were not responsive to stimuli moving from anterior in the visual field, while only one unit had its null direction from posterior in the field. Figure 4 is a polar plot of the number of units with null directions from each of the eight directions tested, with null direction defined as that direction of movement of a small bar that produces the least response. (In the cell illustrated in figures 2 and 3, the null direction is from posterior-inferior). The observed distribution of null directions is significantly dif-
AVIAN RETINAL GANGLION CELL RECEPTIVE FIELDS
ferent than would be expected if null directions were distributed randomly (chisquare test, p < 0.001).
SUPERIOR
On or off center, inhibitory surround units A relatively small number (1 1% ) of units respond exclusively at light on or light off to small or large stationary spots. Like all other receptive fields encountered, they have an inhibitory surround that does not produce a response when stimulated alone. None were more sensitive to moving then to stationary spots, and none were directionally selective.
Miscellaneous units (a) Color units Only 3% of the enitre group of units studied responded differentially to wavelength. The action potentials of these cells were often quite small and the recording time usually less than average, thus none were characterized carefully with spectral sensitivities or with colored adapting lights. The usual situation was to encounter a unit that responded rather poorly to white but showed a better response to a particular color (usually red and occasionally blue) suggesting a n opponent color mechanism. None of the units had surround responses to any wavelength. The small numbers of cells encountered and the relatively short periods of time available for study made more quantitative observations impossible. (b) "Dimming" units Only two cells were encountered in the population that responded to a decrease in diffuse illumination and could not be made to respond to other stimuli. The receptive fields were large and not easily localized; both units were first detected by the presence of sustained activity that increased greatly when the background illumination was decreased. Neither had detectable inhibitory surrounds, and neither were sensitive to movement or direction of movement. Receptive field positions The visual field positions of the four major types of receptive field are plotted in figure 5. Because of the difficulty in locating the pigeon's fovea ophthalmoscopically, we centered our co-ordinate system
117
INFERIOR
Fig. 4 A polar plot of the number of directionally selective cells with null directions from each of the eight directions tested. A cell with its minimum response to a small stimulus moving from superior to inferior in the visual field would be placed in the group with null directions f ~ o msuperior, and be plotted on the axis labelled superior. The horizontal axis corresponds to the bird's horizon when the head is positioned with the long axis of the beak 30° downward from horizontal (see text).
on a line passing through the center of the pupil of the right eye, perpendicular to the birds midline plane. During experiments, the receptive fields were plotted with the head in the position specified by the stereotaxic atlas for the pigeon (Karten and Hodos, '67), with the line between mouth bar and ear bars at an angle of 45" to the horizontal. This position puts the long axis of the beak a t about 75" downward from the horizontal. In order to compare our data with other studies of the pigeon visual system, the visual field coordinates of figures 4 and 5 have been transformed to correspond to a head position that places the beak at an angle of 30 downward from the horizontal (Hamdi and Whitteridge, '54; Bilge, '71; Clarke and W hitteridge, per son a1 communication). This position would be obtained in a stereotaxic instrument that had mouth bar and ear bars level, and is close to the normal head position of a walking or flying pigeon (Duijm, '51; Gray, '53). A transformation of visual field co-ordinates based on head position is possible because there is little if a n y static counter-rotation of the eye with changes in head position (Benjamins and Huizinga, '27). With the beak at 30" downward from the horizontal, the O
10 :13 d nS
AVIAN RETINAL GANGLION CELL RECEPTIVE FIELDS
long axis of the pecten forms an angel of 60" to the horizontal, and serves as a useful reference. The pigeon retina has two specialized regions, the fovea and the red field, that contain a relatively large number of cells in the inner nuclear layer (Galifret, '68). We have superimposed Galifret's outline of the fovea and the red field on our plot of receptive field positions in figure 5. The superimposition was accomplished by matching the magnifications in Galifret's ('68) figure 6 and our figure 5, then orienting the pecten in his figure at 60" to our horizontal axis. We placed the center of the fovea (Galifret, '68: fig. 6) at 15" anterior to our vertical axis and 5" above our horizontal axis, corresponding approximately to the position given by Barlow and Ostwald ('72) and Clarke and Whitteridge (personal communication). The dashed line in figure 5 encloses the projection of the dorsal temporal retinal zone that contains a relatively large number of cells in the inner nuclear layer (more than 900 cells in an area (50 X 50 p , from Galifret, '68: fig. 6). This region corresponds roughly to the region of the retina called the red field because of its high concentration of red oil droplets in the receptors. The placement of these outlines on our co-ordinate system must be taken as approximate; they are nevertheless useful in indicating the retinal regions that we have sampled. Each point in figure 5 represents the position of a single optic tract receptive field in the visual field of the right eye. On-off, inhibitory surround units are shown as filled circles, motion sensitive units as open circles, directional units as filled triangles, and on or off center units as open triangles. The sample is biased towards cells above the horizontal and behind the vertical meridian, presumably because of the position of the usual electrode placement in the optic tract. Each receptive field type was also plotted on an individual visual field map. No clustering of a particular type in a particular part of the retina was evident. Each type occurs in about the same proportions in the nasal and temporal halves of the field, and in each of the four field quadrants. Since the number of units from the fovea and the red field is relatively small, we are unable
119
to make meaningful comparisons between these functionally important regions and the rest of the retina. DISCUSSION
Two major studies of receptive field characteristics of avian retinal ganglion cells have been published (Maturana, '62; Maturana and Frenk, '63; Miles, '72a). The present study is in substantial agreement with the more recent of the two (Miles, '72a), published as the present work was nearing completion; the studies are thus quite independent but without major disagreement. The preparation used by Miles was the decerebrate chick, 1-2 weeks after hatching. In that preparation he found little or no spontaneous activity among retinal ganglion cells studied with micropipettes; spontaneous activity was present in about 50% of the units in our study, and varied greatly from cell to cell. The other major procedural difference between the two studies was that Miles sectioned the isthmo-optic tract prior to study of retinal ganglion cells; we did not carry out tract section. As will become clear in the next paper, which deals with the effects of reversible removal of the efferents (Pearlman and Hughes, '76), basic receptive field properties are not altered by interrupting the isthmo-optic tract. The receptive field types described in Miles' study and in ours are remarkably similar despite these differences in preparations. There are, however, several quantitative differences between the two studies worthy of comment. Miles places 56% of his 159 units in the on-off, non-directional category, at first glance considerably more than the 25% of the 359 units in our study. He does not have a separate category for motion sensitive units, but rather states that most non-directional units were motion sensitive. When the units in our motion sensitive category (15% ) are added, the figures become more nearly equal. A greater numerical difference arises in the consideration of directional cells: Miles described 1 2 % , whereas we find 38%. Miles also found a group of units that respond only at on or at off to stationary spots; the percentages are again rather different, however, in that 32% of his units were on
120
ALAN L. PEARLMAN A N D CHARLES P. HUGHES
or off center, whereas only 11% of our detectors” have features making them sample was in this category. Miles’ non- compatible with our on-off center inhibitory directional off center units are very much surround units and motion sensitive units. like the units that we classify as “dimming” His “directional moving edge detectors” units, in that they have large receptive are the same as our directional units. fields, no inhibitory surround, and dis- Maturana also mentions “luminosity decharge during dimming of ambient light. tectors” and includes in this group cells He also describes two off-center units that that “respond to color”; these are not deare directional; we did not find any such scribed further in his published work. We found a disproportionately large numunits. Color coded retinal ganglion cells occurred in our population, but were not ber of directional cells with null directions well characterized; Miles apparently did from anterior in the birds visual field, and very few with null directions from postenot find any. The differences between the findings of rior. A similar non-uniform distribution of Miles in the chick retinal ganglion cells, null-preferred axes is present in Holden’s and our findings in the pigeon, point out (‘69) preliminary report of a small number some of the problems inherent in receptive of pigeon retinal and tectal cells, in Frost field analysis. In the absence of more com- and Thomsen’s (’72) study of tectal units, pelling explanations, most quantitative dif- and in two studies of receptive fields of the ferences can probably be accounted for by isthmo-optic nucleus (Holden and Powell, differences in sampling. There are an es- ’72; Miles, ’72b). Our own study of pigeon timated 1,000,000 fibers (Cowan, ’70) in tectum (Hughes and Pearlman, ’74) sugthe pigeon’s optic nerve; taking the case gested that a similar distribution exists for of directional cells, the difference between directional cells recorded with metal elec38% of a sample of 359 and 12% of a trodes, but not for those recorded with sample of 159 is quite likely meaningless. glass pipettes, again raising the problem Since the two samples were obtained in of sample selection. If an artifact of samdifferent manners (glass pipettes in the ple selection has given rise to the non-uniretina vs. metal microelectrodes in the form distribution of directionalities that optic tract), there is abundant opportunity we find among retinal ganglion cells, it is to record from slightly different populations an artifact of some interest; sample selecdepending on the vagaries of electrode tion in this case would presumably be reconfiguration (Stone, ’73; Levick and Cle- lated to some basic property of the cells land, ’74), cell and fiber geometry, and under study, such as axon diameter. The rabbit retina also has a large numelectrode position. Such difficulties probably also account for the absence of color- ber of directional retinal ganglion cells coded ganglion cells in some studies (Ma- that respond best to stimuli moving from turana, ’62; Miles, ’72a) and the paucity posterior to anterior in the visual field. in ours. Donner (‘53) described a number The preferred directions of rabbit on-off of retinal ganglion cells with narrow spec- directional cells cluster into four groups tral sensitivities, but gave no indication along the vertical and horizontal axes of of an opponent-color process. The diffi- the visual field, with about 40% preferring culty in finding color opponent cells in the stimuli moving from posterior to anterior retina of an animal known to have color (Oyster, ’68; Daw and Wyatt, ’74), and vision could relate to a problem in sam- only 20% responding best to stimuli movpling; perhaps color-coded retinal gan- ing posteriorly. Directional cells of the on glion cells are simply among the smaller type cluster into three groups, none of which respond to stimuli moving posteriorcells, and thus not easily recorded. ‘ Maturana (‘62) and Maturana and ly (Oyster, ’68). There is a very interesting Frenk (‘63) described six classes of retinal parallel between the physiologic observaganglion cells. They found a small number tions in rabbit and pigeon, and the studies of orientation specific units that they called of optokinetic eye and head movements in “verticality and horizontality detectors.” these animals. Optokinetic eye movements These were not present in Miles’ (‘72a) in the rabbit show directional preference study, in Holden’s (’69) brief account of when the stimuli are presented to only one his work in pigeon, nor in ours. Maturana’s eye. Monocular stimuli moving forward in “general edge detectors” and “convex edge the visual field are much better than those
AVIAN RETINAL GANGLION CELL RECEPTIVE FIELDS
moving backward in eliciting optokinetic nystagmus (Collewijn, '69, Ter Braak, '36). Optokinetic head movement in the pigeon is also elicited preferentially by anteriorly moving monocular stimuli (Huizinga and van der Meulin, '51; Frost and Thomsen, '72; Corbalis and Luthe, '73). It has frequently been pointed out that the ganglion cell receptive field patterns of birds and frogs are more complex than those of "higher" vertebrates like monkey and cat. Dowling ('68) and Dubin ('70) have suggested that this increased complexity is related to the large number of amacrine synapses that are interposed between bipolar and ganglion cells in pigeon and frog retinas, and that are not as prevalent in cat or monkey. Our study of the physiology of retinal ganglion cells conh s the distinction between the more complex receptive fields of the pigeon on the one hand, and the apparently simple center-surround arrangement of the cat on the other. Although this distinction has become less sharp with the description of receptive field types in the cat that are quite like those of the pigeon (Stone and Hoffmann, '72; Cleland and Levick, '74), receptive field properties such as directional selectivity are common in the pigeon but only rarely encountered in the cat (Cleland and Levick, '74), so that the distinction continues to hold on a quantitative if not a qualitative basis. There are also distinct anatomical differences between various regions of the pigeon retina. The fovea and the region of the dorsal nasal retina called the red field contain greater numbers of cells in the ganglion cell layer (Binggeli and Paule, '69) and in the inner nuclear layer (Galifret, '68). The inner plexiform layer of the red field has a more complex synaptic arrangement than the rest of the retina, as judged by a high density of amacrine synapses and a high ratio of amacrine to bipolar synapses (Yazulla, '74). By these criteria the fovea and parafovea are less complex than other retinal regions, despite the large number of cells in ganglion cell and inner nuclear layers (Yazulla, '74). Yazulla has put forward the hypothesis that since the distribution between synaptic complexity in the inner plexiform layer of lower and higher vertebrates correlates reasonably well with the differences in receptive field complexity in these two groups,
12 1
there might also be differences in receptive field complexity in different regions of the pigeon retina to correspond to the anatomical differences. Although there are differences between nasal and temporal yellow fields in the anatomical complexity of the inner plexiform layer (Yazulla, '74), we found no difference in the relative numbers of the various receptive field types in these two regions. A more meaningful test of the hypothesis would be to compare the red field and the fovea with each other and with the rest of the retina, since these are the regions that are anatomically most distinct. Unfortunately, our sample does not contain sufficient numbers of cells from fovea or red field to make these comparisons. The question would best be approached by directly sampling ganglion cells from these regions within the retina rather than in the optic tract. ACKNOWLEDGMENTS
We are grateful to Nigel W. Daw and Harry J. Wyatt for many helpful suggestions during the course of these studies. Dr. Daw also took part in several of the initial experiments of this series. Gerald C. Johns of the Computer Systems Laboratory, Washington University, designed the electronic equipment used in the stimulus apparatus, and wrote programs for the LINC computer. Georgia Harper and Robert White provided valuable technical assistance. LITERATURE CITED Barlow, H . B., and T. J . Ostwald 1972 Pecten of the pigeon's eye as an inter-ocular eye shade. Nature (New Biol.), 236: 88-90, Benjamins, C. E., and E. Huizinga 1927 Untersuchungen uber die Funktion des Vestibularapparatus bei der Taube. Pflugers Archiv, 217: 105-123. Bilge, M. 1971 Electrophysiological investigations on the pigeon's optic tectum. Quart. J. Exp. Physiol., 56: 242-249. Binggeli, R. L., and W. J. Paule 1969 The pigeon retina: quantitative aspects of the optic nerve and ganglion cell layer. J. Comp. Neur., 137: 1-18. Brooke, R. N. L., J. C. Downer and T. P. S . Powell 1965 Centrifugal fibers to the retina in monkey and cat. Nature, 207: 1365-1367. Burger, R. E., and F. W. Loren2 1960 Artificial respiration in bird by unidirectional air flow. Poultry Sci., 39: 236-237. Cajal, S. Ramon y 1889 Sur la morphologie et les connexions des elements de la retine des oiseaux. Anat. Anz., 4: 111-121. 1894 Die Retina der Wirbelthiere. Wiesbaden: Bergmann.
122
ALAN L. PEARLMAN AND CHARLES P. HUGHES
Cleland, B. G., and W. R. Levick 1974 Properties of rarely encountered types of ganglion cells in the cat’s retina and an overall classification. J. Physiol., 240: 4 5 7 4 9 2 . Collewijn, H. 1969 Optokinetic eye movements in the rabbit: input-output relations. Vision Res., 9: 117-132. Corballis, M. C., and L. Luthe 1973 Perception of lateral movement by monocularly viewing pigeons. Per. Psy., 14: 4 1 4 4 . Cowan, W. M., and T. P. S. Powell 1963 Centrifugal fibers i n the avian visual system. Roc. Roy. SOC.(Ser. B), 158: 232-259. Cowan, W. M. 1970 Centrifugal fibers to the avian retina. Brit. Med. Bull., 26: 112-118. Daw, N. W., and H. J. Wyatt 1974 Raising rabbits in a moving visual environment: an attempt to modify directional sensitivity in the retina. J. Physiol., 240: 3 0 9 3 3 0 . Dogiel, A. S. 1895 Die retina der vogel. Arch. mikr. Anat., 44; 622-648. Donner, K. 0. 1953 The spectral sensitivity of the pigeon’s retinal elements. J. Physiol., 122: 524537. Dowling, J. E. 1968 Synaptic organization of the frog retina: An electron microscopic analysis comparing the retinas of frogs and primates. R O C .ROY.SOC.(B), 170: 205-228. Dowling, J. E., and W. M. Cowan 1966 A n electron microscope study of normal and degenerating centrifugal fiber terminals in the pigeon retina. Z. Zellforsch., 71: 14-28. Dubin, M. W. 1970 The inner plexiform layer of the vertebrate retina: A quantitative and comparitive electron microscope analysis. J. Comp. Neur., 140: 4 7 9 4 0 6 . Duijm, M. 1951 On the head position in birds and its relation to some anatomical features. I. Proceedings koninklijke Nederlands. Akudamie van wetenschoppen, 54; 202-211. Frost, B., and D. Thomson 1972 Directional asymmetry of cells in the pigeon tectum. Paper presented to Canadian Psychological Association Meeting, Montreal, June. Galifret, Y. 1968 Les diverse airesfonctionnelles de la retine du Pigeon. Zeitschrift fur Zellforschung, 85: 5 3 5 5 4 5 . Gray, J. 1953 How Animals Move. Cambridge University Press. Hamdi, F. A,, and D. Whitteridge 1954 The representation of the retina o n the optic tectum of the pigeon. Quart. J. Exp. Physiol., 39; 111119. Holden, A. L. 1969 Receptive properties of retinal cells and tectal cells in the pigeon. J. Physiol., 200;2P4P. Holden, A. L., and T. P. S. Powell 1972 The functional organization of the isthmo-optic nucleus in the pigeon. J. Physiol., 223: 4 1 9 4 4 7 . Honrubia, F. M., and J. H. Elliott 1968 m e r e n t innervation of the retina. I. Morphologic study of the human retina. Arch. Ophthal., 80: 98103. Hughes, C. P., and A. L. Pearlman 1974 Single unit receptive fields and the cellular layers of the pigeon optic tectum. Brain Res., 80: 365377. Huizinga, E., and P. van der Meulen 1951 Vestibular rotatory and optokinetic reactions in the pigeon. Ann. Otol. Rhinol. Laryngol., 60: 92 7-94 7.
Karten, H. J., and W. Hodos 1967 A Stereotaxic Atlas of the Brain of the Pigeon (Columba livia.) Baltimore, The Johns Hopkins Press. Levick, W. R., and B. G. Cleland 1974 Selectivity of microelectrodes in recording from cat retinal ganglion cells. J. Neurophysiol., 37: 1387-1393. Maturana, H. R. 1962 Functional organization of the pigeon retina. Proc. 3rd Int. Union of Physiological Sciences. Excerpta Medica Found., Amsterdam. Congr. Ser., 49: 170-178. Maturana, H. R., and S. Frenk 1963 Directional movement and horizontal edge detectors in the pigeon retina. Science, 142: 977-979. 1965 Synaptic connections of the centrifugal fibers in the pigeon retina. Science, 150:3 5 9 3 6 1 . Miles, F. A. 1972a Centrifugal control of the avian retina. I. Receptive field properties of retinal ganglion cells. Brain Res., 48: 65-92. 197213 Centrifugal control of the avian retina. 11. Receptive field properties of cells in the isthmo-optic nucleus. Brain Res., 48; 93-113. McGill, J. I., T. P. S. Powell and W. M. Cowan 1966a The retinal representation upon the optic tectum and isthmo-optic nucleus in the pigeon. J. Anat., 100: 5-33. 1966b The organization of the projection of the centrifugal fibres to the retina in the pigeon. J. Anat., 100: 35-49. Ogden, T. E. 1968 On the function of efferent retinal fibres. In: Structure and Function of lnhibitory Neuronal Mechanisms. C. von Euler, S. Skoglund and U. Soderberg, eds. Pergamon Press, Oxford, pp. 89-109. Oyster, C. W. 1968 The analysis of image motion by the rabbit retina. J. Physiol., 199: 613635. Pearlman, A. L., and C. P. Hughes 1973 Functional role of efferents to the retina. Trans. Am. Neurol. Assoc., 98; 9-12. 1976 Functional role of efferents to the avian retina. 11. Effects of reversible cooling of the isthmo-optic nucleus. J. a m p . Neur., 166: 123-1 32. Polyak, S. 1957 The Vertebrate Visual Svstem. The University of Chicago Press, Chicago, pp. 25 C 2 5 2 . Stone, J. 1973 Sampling properties of microelectrodes assessed in the cat retina. J. Neurophysiol., 36; 1071-1079. Stone, J., and K.-P. Hoffman 1972 Very slowconducting ganglion cells in the cat’s retina: a major new -functional type? Brain Res., 43: 610-616. Ter Braak, J. W. G. 1936 Untersuchungen uber optokinetischen nystagmus. Arch. Neerl. Physiol., 21 ; 309-376. Wallenberg, A. 1898 Das mediale Opticusbiindel der Taube. Neurol. Zbl., 17: 5 3 2 5 3 7 . Witkovsky, P. 1971 Synapses made by myelinated fibers running to teleost and elasmobranch retinas. J. Comp. Neur., 142: 205-222. Wyatt, H. J., and N. W. Daw 1975 Directionally sensitive ganglion cells in the rabbit retina: specificity for stimulus direction, size and speed. J. Neurophysiol., 38: 613-626. Yazulla, S. 1974 Intraretinal differentiation in the synaptic organization of the inner plexiform layer of the pigeon retina. J. Comp. Neur., 153: 309-324,
ABSTRACT Receptive fields of retinal ganglion cells were analyzed during extracellular microelectrode recordings in the optic tract of the lightly anesthetized pigeon. Four major types of receptive field can be distinguished among the 359 fibers studied. Twenty-five percent of the receptive fields are relatively simple, responding at on and at off to stationary spots of light in the central region. All of the receptive fields have inhibitory surrounds of varying strength that do not produce a response when illuminated alone, but antagonize responses from the central region. Motion sensitive units comprise 15% of the recorded population; they are similar to the o n d f center type except that responses to stationary stimuli are absent or very weak while responses to moving stimuli are vigorous. Directionally selective units also have the basic features of on-off, inhibitory surround cells, but respond to moving stimuli well from the preferred direction and not at all from the null direction. Directional cells have a broad range of null directions; in about one-third of the units the range becomes broader when the stimulus involves both center and surround of the receptive field, thus enhancing directional selectivity. Directionally selective units are common, comprising 38% of the units studied. Cells unresponsive to stimuli moving from anterior in the visual field are much more common than other types, while cells unresponsive to stimuli from posterior in the field are rare. A few units (11%) respond only at on or at off to stationary stimuli in their receptive field centers; they also have antagonistic but unresponsive receptive field surrounds. The area of the visual field sampled is uniform in regard to the relative numbers of the four major receptive field types. Centrifugal fibers to the retina were first described by Cajal (1889) in the avian retina; the bird retina remains the only well substantiated example of a vertebrate retina receiving efferent fibers (Ogden, '68; Cowan, '70). Although efferent fibers have been described in the retinas of the fish (Witkovsky, '71), the dog (Cajal, 1894), the chimpanzee (Polyak, '57), the cat and monkey (Brooke et al., '65), and the human (Honrubia and Elliott, '68), a great deal of controversy continues to surround the question of the exsistence of efferent fibers in these species. The nucleus of origin of the centrifugal fibers in the bird has been clearly demonstrated; no such demonstration is available for other vertebrates. Efferent fibers to the retina in the bird arise in the isthmo-optic nucleus, a distinct cell mass in the caudal midbrain, (WallenJ. COMP. NEUR.,166: 111-122
burg, 1898; Cowan and Powell, '63) and form the final link in an anatomical closed loop. Each point on the retina projects to a restricted area of the tectum and the projection from tectum to isthmo-optic nucleus is also organized in a distinct retinotopic manner. The isthmo-optic nucleus in turn projects back to the retina in a retinotopic fashion, each quadrant of the nucleus sending fibers to the same quadrant of the retina that provides its input via the tectum (McGill et al., '66a,b). The terminations of the efferent fibers 1 This investigation was supported by NIH Research Grant R01-EY-00621from the National Eye Institute. A preliminary report of this work was published earlier in abstract form (Pearlman and Hughes, '73). 2 Gordon R. and Thelma B . Coates Scholar in Neurology, Washington University. 3 Supported by NIH Special Fellowship F11 NS 02437 from the National Institute of Neurological Diseases and Stroke.
111
112
ALAN L. PEARLMAN A N D CHARLES P. HUGHES
to the retina have also been carefully studied in the bird. Cajal (1889) and Dogiel (1895) demonstrated that these fibers terminate on amacrine cells along the inner aspect of the inner nuclear layer. Maturana and Frenk ('65) extended these observations and described two distinct patterns of efferent terminations: (1) a divergent type in which the fibers branch and contact several different cells and (2) a convergent type, in which the fibers also branch but after the branches separate they ultimately converge to form a pericellular nest near the soma of one or a few cells. There are an estimated 100,000 centrifugal axon terminals in the pigeon retina; since there are about 12,000 fibers in the isthmo-optic tract each isthmo-optic tract fiber gives rise to eight to ten terminals (Cowan, '70). The efferent terminals are quite distinct; they form the largest synaptic endings in the junction of inner plexiform and inner nuclear layers, and contain tight aggregates of synaptic vesicles and many mitochondria. These endings are found on the proximal portion of the main processes of the amacrine cells, and occasionally on the cell soma (Dowling and Cowan, '66). In the present study a neurophysiological analysis of this anatomically well described system has been carried out in two parts. The first is a study of the receptive field properties of individual pigeon retinal ganglion cells; it extends the observations of earlier reports (Maturana, '62; Maturana and Frenk, '63; Holden, '69) and is in essential agreement with the recent work of Miles ('72) in the chick. The second part (Pearlman and Hughes, '76) is concerned with the effects of reversible removal of efferent influences emanating from the isthmo-optic nucleus on the physiologic properties of retinal ganglion cells.
of small receptive fields over 1-2 hours. Initially the animals were ventilated with a respiratory pump (Harvard Apparatus) and the percentage of COz in the expired air measured with an infrared COz meter (Beckman Instruments). In later experiments the unidirectional airflow system of bird respiration (Burger and Lorenz, '60) was substituted, consider ably reducing brain pulsations. A mixture of nitrous oxide (40 % ), oxygen (55 % ), and COB(5 % ) was humidified and passed in a steady stream (1,000 cc/mm) into the trachea and out through a hole in a major abdominal air sac. The gas mixture and flow rate were determined by serial measurements of arterial p02, pCOz, and pH on two animals over several hours. Body temperature was monitored by a thermistor deep in the pectoral muscle, and maintained by circulating warm water in a heating pad (Gorman-Rupp). The EKG was recorded at varying intervals on a polygraph. The bird's eye was refracted at 30" with a slit retinoscope, and fitted to within kO.5 diopters with a contact lens. The bird's head was placed in a stereotaxic head holder built to conform to the co-ordinates of the Karten and Hodos ('67) atlas for the pigeon. The head holder was designed so that it blocked only a small portion of the most posterior visual fields. Electrode carriers (Kopf) were mounted on a bar parallel to the ear bars behind the bird's head, so that they were not in the field of view. In early experiments fibers were recorded in the optic chiasm or tract by stereotaxic placement of tungsten or platinum microelectrodes. Electrode placements were subsequently checked in serial frozen or paraffin sections. In later experiments the left cerebral hemisphere was removed acutely by suction to permit direct visualization of the optic tract, thereby greatly enhancing the yield in a given exMETHODS periment. There were no detectable differAdult White Carneaux pigeons (Columba ences in the units recorded in the two livia, Palmetto Pigeon Plant, Sumter , S.C.) preparations. were anesthetized with halothane for surExtracellular action potentials of single gery and then maintained with nitrous optic tract fibers were recorded convenoxide-oxygenmixtures. They were paralyzed tionally, displayed on an oscilloscope, (Tekwith intramuscular curare (7.5 mg initially tronix 565) and monitored through an and 1.5-3 mg every three hours thereafter) audio amplifier and loud speaker. Ampliand artificially respired. With curarization, fied action potentials were led into a winno eye movement could be detected by the dow discriminator that produced a brief method of repeatedly plotting the borders rectangular pulse for each spike falling
AVIAN RETINAL GANGLION CELL RECEPTIVE FIELDS
within the variable upper and lower limits of the window. Action potentials, window limits and window output were monitored on an oscilloscope (Tektronix 564B) and checked frequently to be certain that only the action potentials of the unit under study were producing a pulse at the window output. The brief rectangular pulses provided the input to a LINC computer (Computer Systems Laboratory, Washington University) programmed to produce post-stimulus time histograms. Histogram collection was initiated when the stimulus beam moved across a photo-electric sensor mounted on the tangent screen near the receptive field. Histograms were photographed from an oscilloscope face, and also stored on magnetic tape for later analysis or for display on an x-y plotter. Receptive fields were mapped on the tangent screen with a projector (Leitz Prado) mounted on a tripod and moved by hand. More precise moving stimuli were provided by another projector whose beam was directed at a front-silvered mirror mounted on a pen-motor (Brush-Clevite). The pen motor was mounted in turn on a mechanical swivel that allowed rotations of the motor through 360 *, providing moving stimuli from any direction. The mirror motor was driven by a trapezoid generator with variable ramp slope, ramp duration, and repetition rate. An electronic shutter on the projector was opened and closed at appropriate times in the stimulus sweep on signal from the computer. RESULTS
The receptive field properties of 359 retinal ganglion cells were characterized during microelectrode recordings from single units in the optic tract or chiasm of 69 pigeons. Since the types of receptive field identified are at least in part a function of the stimuli employed in searching for units, it is important to specify the search procedures in some detail. A distinct multiunit response was present when a flashlight beam was shown across the retina. This was useful for indicating a functioning electrode, but since so many units responded it was of no value in indicating the presence of cells sufficiently isolated for proper study. A similar but smaller multi-unit response was often evident when bars of light, dark bars, or more complex
113
objects moved across a specific part of the visual field. Once such an area was located, the electrode was advanced slowly while the area was being stimulated by light and dark bars of varying sizes, both stationary and moving, until a single unit was clearly isolated above the background. Spontaneous activity was not a prominent feature of retinal ganglion cells in our preparation, although roughly 50% had some spontaneous firing, and a small percentage fired vigorously when unstimulated. The presence or absence of spontaneous activity could not be correlated with any of the specific receptive field types outlined below. The receptive field of each isolated unit was studied with stationary squares or rectangles of varying size and orientation turned on and off, with light and dark bars of varying size moved through the receptive field from each of eight directions at 45" intervals, and colored stimuli of the same type produced by placing red, green, and blue Wratten filters in the projector beam. Receptive fields defined with these stimuli can be distributed into four major classes, with several other types occurring less frequently. h - o f f center, inhibitory surround units In the recorded population of optic tract axons, 25 % have relatively simple receptive fields with central regions that respond transiently to small spots of light when turned on and when turned off. The central region is always surrounded by an antagonistic zone that inhibits the center response when the stimulus extends into it, but does not produce a response when illuminated by itself. The inhibitory surround region is quite variable in the strength of its effect from one unit to another; a stimulus that extends only a short distance into the surround may produce profound inhibition in one cell, while a stimulus encompassing the entire inhibitory surround may produce only slight inhibition in another. A typical on-off center, inhibitory surround unit is displayed in figure 1. The receptive field is shown on the left with the responsive central region outlined by a solid line; the f signs indicate responses at on and at off to a small stationary test spot. The open circles indicate no response to a small test spot. A bar of light (hatched
114
ALAN L. PEARLMAN AND CHARLES P. HUGHES
Fig. 1 On-off center, inhibitory surround unit. Receptive field on the left; on responses to small stationary spots shown as + , Off responses -, and no response as 0. The 1 calibration mark refers to both the receptive field and the bar of stimulus light, shown as hatched bars to the right, moving in direction indicated by arrow. Post-stimulus time histograms at bottom of figure show responses to leading and trailing edges of the stimulus bar. Each histogram is the sum of 15 sweeps of the bar across the field at 12.6"/sec,one sweep every 4.8 seconds.
rectangle) moving across the receptive field produces a transient response as the leading edge crosses the field, and another when the trailing edge crosses. The responses are indicated in the post-stimulus time histograms beneath each bar. The marked decrease in the response when the stimulus extends outside the receptive field center illustrates the effect of a strong inhibitory surround. Also evident is the fact that the on (or leading edge) response and off (or trailing edge) response are not always of equal amplitude. This feature is quite variable from cell to cell, with some cells responding more strongly at on, some at off, and some nearly equally to both. When the stimulus is reduced in intensity nearly to threshold for a given unit, the stronger of the two responses usually occurs exclusively. Most of the features described thus far for the on-off center, inhibitory surround cells are also present in the next two types of cell to be described, and therefore in about 80% of the cells studied; these two cell types differ only in their special features in response to moving stimuli. Motion sensitive units
Although all units respond to some degree to moving stimuli, we have somewhat arbitrarily divided off a group of units (15%) that respond far better to moving stimuli than to any others. The receptive
field maps of these units look identical to those of the on-off, inhibitory surround types described above, except that the responses to stationary stimuli are weak or occasionally entirely absent, while the responses to moving stimuli are vigorous. Responses to both leading and trailing edges of a moving bar are the rule, but again with variability in relative amplitude of the two responses from unit to unit. A small spot moving within the receptive field center is frequently a very effective stimulus, as is a small spot moving back and forth through the field. The orientation of the moving edge is unimportant. These units respond to a moving edge no matter in what direction it is moving, thus distinguishing them from the directionally selective cells to be described next. Directionally selective units
The largest single class (38%) of units encountered in this study are those exhibiting directional selectivity in their responses to moving stimuli. An example is shown in figure 2,where the receptive field is displayed in the center, and the poststimulus time histograms generated by a bar of light moving across it are shown around the circumference. Each histogram is displayed alongside an arrow representing the direction of movement of the bar; the bar was a rectangle whose short side was about the same size (2.6")as the short
AVIAN RETINAL GANGLION CELL RECEPTIVE FIELDS
0 O(-9O
0
0
115
0
-
0
0
L 1 0
liPi7 b
Fig. 2 Directionally selective unit.Receptive field in the center of the figure; on responses to small stationary spots shown as off as -, and no response as a circle; l o calibration refers to receptive field; 5-spike4.5 sec calibration refers to post stimulus time histograms elicited by 15 sweeps of a 2.6" X 6.6' bar moving parallel to its long axis in the direction indicated by adjacent arrow at 16.4O/sec, one sweep every five seconds.
+,
side of the field center. The long side of the rectangle was long (6.6') relative to the field center and was oriented parallel to the direction of movement in each case. Bar speed was constant (16.4'/sec), as was the repetition rate (one sweep every 5 seconds). The superior aspect of the bird's visual field is represented at the top of the figure, and the posterior aspect to the right; the receptive field was in the right eye.
For this unit stimuli moving from posterior-inferior are virtually ineffective, and those moving from anterior-superior are quite effective. A quantitative representation of directional selectivity for the same unit is shown in the polar plot of figure 3. Each point on a solid line axis represents the average number of action potentials produced by a sweep of the stimulus bar through the re-
116
ALAN L. PEARLMAN AND CHARLES P. HUGHES Superior
!
\
\ Posterior
----------
Small Bar Medium Bar Large Bar -I-...-------
12 spikes /sweep Inferior
Fig. 3 Polar plot of directional unit shown in figure 2. Points on a given axis are the average number of action potentials produced by 15 sweeps of a stimulus of the indicated size through the field center along that axis. Small bar: 0.66 X 6 . 6 O ; medium bar: 2.6O X 6.6": large bar: 6 . 7 " x 6.6O. Superior, posterior, etc. refer to the visual field of the bird's right eye.
ceptive field along that axis. For example, points on the superior axis represent the number of spikes produced by a stimulus sweep from superior to inferior. The null and preferred directions are again evident, but i t is also evident that the range of directions giving no response is related to the size of the stimulus bar. A small bar, smaller than the receptive field center, produces a good response when moved through the field along the 45" axis from posterior-superior to anterior-inferior; a bar of the same size as the field center (medium bar, fig. 3 ) is about half as effective from the same direction, and a large bar, involving the field surround, is virtually ineffective. The directional selectivity of the unit is thus enhanced for large stimuli, since the cell responds to large stimuli over a narrower range of directions than it does to small stimuli. About one-third of the directional cells in our sample had enhanced directiond selectivity to large stim-
uli. This phenomenon is also present in the receptive fields of chick retinal ganglion cells (Miles, '72), and has been carefully studied in the directional ganglion cells of the rabbit (Daw and Wyatt, '74; Wyatt and Daw, '75). When the entire group of directionally selective cells is analysed, it is apparent that the null directions of the receptive fields are not evenly distributed. More than one-third of directionally selective units were not responsive to stimuli moving from anterior in the visual field, while only one unit had its null direction from posterior in the field. Figure 4 is a polar plot of the number of units with null directions from each of the eight directions tested, with null direction defined as that direction of movement of a small bar that produces the least response. (In the cell illustrated in figures 2 and 3, the null direction is from posterior-inferior). The observed distribution of null directions is significantly dif-
AVIAN RETINAL GANGLION CELL RECEPTIVE FIELDS
ferent than would be expected if null directions were distributed randomly (chisquare test, p < 0.001).
SUPERIOR
On or off center, inhibitory surround units A relatively small number (1 1% ) of units respond exclusively at light on or light off to small or large stationary spots. Like all other receptive fields encountered, they have an inhibitory surround that does not produce a response when stimulated alone. None were more sensitive to moving then to stationary spots, and none were directionally selective.
Miscellaneous units (a) Color units Only 3% of the enitre group of units studied responded differentially to wavelength. The action potentials of these cells were often quite small and the recording time usually less than average, thus none were characterized carefully with spectral sensitivities or with colored adapting lights. The usual situation was to encounter a unit that responded rather poorly to white but showed a better response to a particular color (usually red and occasionally blue) suggesting a n opponent color mechanism. None of the units had surround responses to any wavelength. The small numbers of cells encountered and the relatively short periods of time available for study made more quantitative observations impossible. (b) "Dimming" units Only two cells were encountered in the population that responded to a decrease in diffuse illumination and could not be made to respond to other stimuli. The receptive fields were large and not easily localized; both units were first detected by the presence of sustained activity that increased greatly when the background illumination was decreased. Neither had detectable inhibitory surrounds, and neither were sensitive to movement or direction of movement. Receptive field positions The visual field positions of the four major types of receptive field are plotted in figure 5. Because of the difficulty in locating the pigeon's fovea ophthalmoscopically, we centered our co-ordinate system
117
INFERIOR
Fig. 4 A polar plot of the number of directionally selective cells with null directions from each of the eight directions tested. A cell with its minimum response to a small stimulus moving from superior to inferior in the visual field would be placed in the group with null directions f ~ o msuperior, and be plotted on the axis labelled superior. The horizontal axis corresponds to the bird's horizon when the head is positioned with the long axis of the beak 30° downward from horizontal (see text).
on a line passing through the center of the pupil of the right eye, perpendicular to the birds midline plane. During experiments, the receptive fields were plotted with the head in the position specified by the stereotaxic atlas for the pigeon (Karten and Hodos, '67), with the line between mouth bar and ear bars at an angle of 45" to the horizontal. This position puts the long axis of the beak a t about 75" downward from the horizontal. In order to compare our data with other studies of the pigeon visual system, the visual field coordinates of figures 4 and 5 have been transformed to correspond to a head position that places the beak at an angle of 30 downward from the horizontal (Hamdi and Whitteridge, '54; Bilge, '71; Clarke and W hitteridge, per son a1 communication). This position would be obtained in a stereotaxic instrument that had mouth bar and ear bars level, and is close to the normal head position of a walking or flying pigeon (Duijm, '51; Gray, '53). A transformation of visual field co-ordinates based on head position is possible because there is little if a n y static counter-rotation of the eye with changes in head position (Benjamins and Huizinga, '27). With the beak at 30" downward from the horizontal, the O
10 :13 d nS
AVIAN RETINAL GANGLION CELL RECEPTIVE FIELDS
long axis of the pecten forms an angel of 60" to the horizontal, and serves as a useful reference. The pigeon retina has two specialized regions, the fovea and the red field, that contain a relatively large number of cells in the inner nuclear layer (Galifret, '68). We have superimposed Galifret's outline of the fovea and the red field on our plot of receptive field positions in figure 5. The superimposition was accomplished by matching the magnifications in Galifret's ('68) figure 6 and our figure 5, then orienting the pecten in his figure at 60" to our horizontal axis. We placed the center of the fovea (Galifret, '68: fig. 6) at 15" anterior to our vertical axis and 5" above our horizontal axis, corresponding approximately to the position given by Barlow and Ostwald ('72) and Clarke and Whitteridge (personal communication). The dashed line in figure 5 encloses the projection of the dorsal temporal retinal zone that contains a relatively large number of cells in the inner nuclear layer (more than 900 cells in an area (50 X 50 p , from Galifret, '68: fig. 6). This region corresponds roughly to the region of the retina called the red field because of its high concentration of red oil droplets in the receptors. The placement of these outlines on our co-ordinate system must be taken as approximate; they are nevertheless useful in indicating the retinal regions that we have sampled. Each point in figure 5 represents the position of a single optic tract receptive field in the visual field of the right eye. On-off, inhibitory surround units are shown as filled circles, motion sensitive units as open circles, directional units as filled triangles, and on or off center units as open triangles. The sample is biased towards cells above the horizontal and behind the vertical meridian, presumably because of the position of the usual electrode placement in the optic tract. Each receptive field type was also plotted on an individual visual field map. No clustering of a particular type in a particular part of the retina was evident. Each type occurs in about the same proportions in the nasal and temporal halves of the field, and in each of the four field quadrants. Since the number of units from the fovea and the red field is relatively small, we are unable
119
to make meaningful comparisons between these functionally important regions and the rest of the retina. DISCUSSION
Two major studies of receptive field characteristics of avian retinal ganglion cells have been published (Maturana, '62; Maturana and Frenk, '63; Miles, '72a). The present study is in substantial agreement with the more recent of the two (Miles, '72a), published as the present work was nearing completion; the studies are thus quite independent but without major disagreement. The preparation used by Miles was the decerebrate chick, 1-2 weeks after hatching. In that preparation he found little or no spontaneous activity among retinal ganglion cells studied with micropipettes; spontaneous activity was present in about 50% of the units in our study, and varied greatly from cell to cell. The other major procedural difference between the two studies was that Miles sectioned the isthmo-optic tract prior to study of retinal ganglion cells; we did not carry out tract section. As will become clear in the next paper, which deals with the effects of reversible removal of the efferents (Pearlman and Hughes, '76), basic receptive field properties are not altered by interrupting the isthmo-optic tract. The receptive field types described in Miles' study and in ours are remarkably similar despite these differences in preparations. There are, however, several quantitative differences between the two studies worthy of comment. Miles places 56% of his 159 units in the on-off, non-directional category, at first glance considerably more than the 25% of the 359 units in our study. He does not have a separate category for motion sensitive units, but rather states that most non-directional units were motion sensitive. When the units in our motion sensitive category (15% ) are added, the figures become more nearly equal. A greater numerical difference arises in the consideration of directional cells: Miles described 1 2 % , whereas we find 38%. Miles also found a group of units that respond only at on or at off to stationary spots; the percentages are again rather different, however, in that 32% of his units were on
120
ALAN L. PEARLMAN A N D CHARLES P. HUGHES
or off center, whereas only 11% of our detectors” have features making them sample was in this category. Miles’ non- compatible with our on-off center inhibitory directional off center units are very much surround units and motion sensitive units. like the units that we classify as “dimming” His “directional moving edge detectors” units, in that they have large receptive are the same as our directional units. fields, no inhibitory surround, and dis- Maturana also mentions “luminosity decharge during dimming of ambient light. tectors” and includes in this group cells He also describes two off-center units that that “respond to color”; these are not deare directional; we did not find any such scribed further in his published work. We found a disproportionately large numunits. Color coded retinal ganglion cells occurred in our population, but were not ber of directional cells with null directions well characterized; Miles apparently did from anterior in the birds visual field, and very few with null directions from postenot find any. The differences between the findings of rior. A similar non-uniform distribution of Miles in the chick retinal ganglion cells, null-preferred axes is present in Holden’s and our findings in the pigeon, point out (‘69) preliminary report of a small number some of the problems inherent in receptive of pigeon retinal and tectal cells, in Frost field analysis. In the absence of more com- and Thomsen’s (’72) study of tectal units, pelling explanations, most quantitative dif- and in two studies of receptive fields of the ferences can probably be accounted for by isthmo-optic nucleus (Holden and Powell, differences in sampling. There are an es- ’72; Miles, ’72b). Our own study of pigeon timated 1,000,000 fibers (Cowan, ’70) in tectum (Hughes and Pearlman, ’74) sugthe pigeon’s optic nerve; taking the case gested that a similar distribution exists for of directional cells, the difference between directional cells recorded with metal elec38% of a sample of 359 and 12% of a trodes, but not for those recorded with sample of 159 is quite likely meaningless. glass pipettes, again raising the problem Since the two samples were obtained in of sample selection. If an artifact of samdifferent manners (glass pipettes in the ple selection has given rise to the non-uniretina vs. metal microelectrodes in the form distribution of directionalities that optic tract), there is abundant opportunity we find among retinal ganglion cells, it is to record from slightly different populations an artifact of some interest; sample selecdepending on the vagaries of electrode tion in this case would presumably be reconfiguration (Stone, ’73; Levick and Cle- lated to some basic property of the cells land, ’74), cell and fiber geometry, and under study, such as axon diameter. The rabbit retina also has a large numelectrode position. Such difficulties probably also account for the absence of color- ber of directional retinal ganglion cells coded ganglion cells in some studies (Ma- that respond best to stimuli moving from turana, ’62; Miles, ’72a) and the paucity posterior to anterior in the visual field. in ours. Donner (‘53) described a number The preferred directions of rabbit on-off of retinal ganglion cells with narrow spec- directional cells cluster into four groups tral sensitivities, but gave no indication along the vertical and horizontal axes of of an opponent-color process. The diffi- the visual field, with about 40% preferring culty in finding color opponent cells in the stimuli moving from posterior to anterior retina of an animal known to have color (Oyster, ’68; Daw and Wyatt, ’74), and vision could relate to a problem in sam- only 20% responding best to stimuli movpling; perhaps color-coded retinal gan- ing posteriorly. Directional cells of the on glion cells are simply among the smaller type cluster into three groups, none of which respond to stimuli moving posteriorcells, and thus not easily recorded. ‘ Maturana (‘62) and Maturana and ly (Oyster, ’68). There is a very interesting Frenk (‘63) described six classes of retinal parallel between the physiologic observaganglion cells. They found a small number tions in rabbit and pigeon, and the studies of orientation specific units that they called of optokinetic eye and head movements in “verticality and horizontality detectors.” these animals. Optokinetic eye movements These were not present in Miles’ (‘72a) in the rabbit show directional preference study, in Holden’s (’69) brief account of when the stimuli are presented to only one his work in pigeon, nor in ours. Maturana’s eye. Monocular stimuli moving forward in “general edge detectors” and “convex edge the visual field are much better than those
AVIAN RETINAL GANGLION CELL RECEPTIVE FIELDS
moving backward in eliciting optokinetic nystagmus (Collewijn, '69, Ter Braak, '36). Optokinetic head movement in the pigeon is also elicited preferentially by anteriorly moving monocular stimuli (Huizinga and van der Meulin, '51; Frost and Thomsen, '72; Corbalis and Luthe, '73). It has frequently been pointed out that the ganglion cell receptive field patterns of birds and frogs are more complex than those of "higher" vertebrates like monkey and cat. Dowling ('68) and Dubin ('70) have suggested that this increased complexity is related to the large number of amacrine synapses that are interposed between bipolar and ganglion cells in pigeon and frog retinas, and that are not as prevalent in cat or monkey. Our study of the physiology of retinal ganglion cells conh s the distinction between the more complex receptive fields of the pigeon on the one hand, and the apparently simple center-surround arrangement of the cat on the other. Although this distinction has become less sharp with the description of receptive field types in the cat that are quite like those of the pigeon (Stone and Hoffmann, '72; Cleland and Levick, '74), receptive field properties such as directional selectivity are common in the pigeon but only rarely encountered in the cat (Cleland and Levick, '74), so that the distinction continues to hold on a quantitative if not a qualitative basis. There are also distinct anatomical differences between various regions of the pigeon retina. The fovea and the region of the dorsal nasal retina called the red field contain greater numbers of cells in the ganglion cell layer (Binggeli and Paule, '69) and in the inner nuclear layer (Galifret, '68). The inner plexiform layer of the red field has a more complex synaptic arrangement than the rest of the retina, as judged by a high density of amacrine synapses and a high ratio of amacrine to bipolar synapses (Yazulla, '74). By these criteria the fovea and parafovea are less complex than other retinal regions, despite the large number of cells in ganglion cell and inner nuclear layers (Yazulla, '74). Yazulla has put forward the hypothesis that since the distribution between synaptic complexity in the inner plexiform layer of lower and higher vertebrates correlates reasonably well with the differences in receptive field complexity in these two groups,
12 1
there might also be differences in receptive field complexity in different regions of the pigeon retina to correspond to the anatomical differences. Although there are differences between nasal and temporal yellow fields in the anatomical complexity of the inner plexiform layer (Yazulla, '74), we found no difference in the relative numbers of the various receptive field types in these two regions. A more meaningful test of the hypothesis would be to compare the red field and the fovea with each other and with the rest of the retina, since these are the regions that are anatomically most distinct. Unfortunately, our sample does not contain sufficient numbers of cells from fovea or red field to make these comparisons. The question would best be approached by directly sampling ganglion cells from these regions within the retina rather than in the optic tract. ACKNOWLEDGMENTS
We are grateful to Nigel W. Daw and Harry J. Wyatt for many helpful suggestions during the course of these studies. Dr. Daw also took part in several of the initial experiments of this series. Gerald C. Johns of the Computer Systems Laboratory, Washington University, designed the electronic equipment used in the stimulus apparatus, and wrote programs for the LINC computer. Georgia Harper and Robert White provided valuable technical assistance. LITERATURE CITED Barlow, H . B., and T. J . Ostwald 1972 Pecten of the pigeon's eye as an inter-ocular eye shade. Nature (New Biol.), 236: 88-90, Benjamins, C. E., and E. Huizinga 1927 Untersuchungen uber die Funktion des Vestibularapparatus bei der Taube. Pflugers Archiv, 217: 105-123. Bilge, M. 1971 Electrophysiological investigations on the pigeon's optic tectum. Quart. J. Exp. Physiol., 56: 242-249. Binggeli, R. L., and W. J. Paule 1969 The pigeon retina: quantitative aspects of the optic nerve and ganglion cell layer. J. Comp. Neur., 137: 1-18. Brooke, R. N. L., J. C. Downer and T. P. S . Powell 1965 Centrifugal fibers to the retina in monkey and cat. Nature, 207: 1365-1367. Burger, R. E., and F. W. Loren2 1960 Artificial respiration in bird by unidirectional air flow. Poultry Sci., 39: 236-237. Cajal, S. Ramon y 1889 Sur la morphologie et les connexions des elements de la retine des oiseaux. Anat. Anz., 4: 111-121. 1894 Die Retina der Wirbelthiere. Wiesbaden: Bergmann.
122
ALAN L. PEARLMAN AND CHARLES P. HUGHES
Cleland, B. G., and W. R. Levick 1974 Properties of rarely encountered types of ganglion cells in the cat’s retina and an overall classification. J. Physiol., 240: 4 5 7 4 9 2 . Collewijn, H. 1969 Optokinetic eye movements in the rabbit: input-output relations. Vision Res., 9: 117-132. Corballis, M. C., and L. Luthe 1973 Perception of lateral movement by monocularly viewing pigeons. Per. Psy., 14: 4 1 4 4 . Cowan, W. M., and T. P. S. Powell 1963 Centrifugal fibers i n the avian visual system. Roc. Roy. SOC.(Ser. B), 158: 232-259. Cowan, W. M. 1970 Centrifugal fibers to the avian retina. Brit. Med. Bull., 26: 112-118. Daw, N. W., and H. J. Wyatt 1974 Raising rabbits in a moving visual environment: an attempt to modify directional sensitivity in the retina. J. Physiol., 240: 3 0 9 3 3 0 . Dogiel, A. S. 1895 Die retina der vogel. Arch. mikr. Anat., 44; 622-648. Donner, K. 0. 1953 The spectral sensitivity of the pigeon’s retinal elements. J. Physiol., 122: 524537. Dowling, J. E. 1968 Synaptic organization of the frog retina: An electron microscopic analysis comparing the retinas of frogs and primates. R O C .ROY.SOC.(B), 170: 205-228. Dowling, J. E., and W. M. Cowan 1966 A n electron microscope study of normal and degenerating centrifugal fiber terminals in the pigeon retina. Z. Zellforsch., 71: 14-28. Dubin, M. W. 1970 The inner plexiform layer of the vertebrate retina: A quantitative and comparitive electron microscope analysis. J. Comp. Neur., 140: 4 7 9 4 0 6 . Duijm, M. 1951 On the head position in birds and its relation to some anatomical features. I. Proceedings koninklijke Nederlands. Akudamie van wetenschoppen, 54; 202-211. Frost, B., and D. Thomson 1972 Directional asymmetry of cells in the pigeon tectum. Paper presented to Canadian Psychological Association Meeting, Montreal, June. Galifret, Y. 1968 Les diverse airesfonctionnelles de la retine du Pigeon. Zeitschrift fur Zellforschung, 85: 5 3 5 5 4 5 . Gray, J. 1953 How Animals Move. Cambridge University Press. Hamdi, F. A,, and D. Whitteridge 1954 The representation of the retina o n the optic tectum of the pigeon. Quart. J. Exp. Physiol., 39; 111119. Holden, A. L. 1969 Receptive properties of retinal cells and tectal cells in the pigeon. J. Physiol., 200;2P4P. Holden, A. L., and T. P. S. Powell 1972 The functional organization of the isthmo-optic nucleus in the pigeon. J. Physiol., 223: 4 1 9 4 4 7 . Honrubia, F. M., and J. H. Elliott 1968 m e r e n t innervation of the retina. I. Morphologic study of the human retina. Arch. Ophthal., 80: 98103. Hughes, C. P., and A. L. Pearlman 1974 Single unit receptive fields and the cellular layers of the pigeon optic tectum. Brain Res., 80: 365377. Huizinga, E., and P. van der Meulen 1951 Vestibular rotatory and optokinetic reactions in the pigeon. Ann. Otol. Rhinol. Laryngol., 60: 92 7-94 7.
Karten, H. J., and W. Hodos 1967 A Stereotaxic Atlas of the Brain of the Pigeon (Columba livia.) Baltimore, The Johns Hopkins Press. Levick, W. R., and B. G. Cleland 1974 Selectivity of microelectrodes in recording from cat retinal ganglion cells. J. Neurophysiol., 37: 1387-1393. Maturana, H. R. 1962 Functional organization of the pigeon retina. Proc. 3rd Int. Union of Physiological Sciences. Excerpta Medica Found., Amsterdam. Congr. Ser., 49: 170-178. Maturana, H. R., and S. Frenk 1963 Directional movement and horizontal edge detectors in the pigeon retina. Science, 142: 977-979. 1965 Synaptic connections of the centrifugal fibers in the pigeon retina. Science, 150:3 5 9 3 6 1 . Miles, F. A. 1972a Centrifugal control of the avian retina. I. Receptive field properties of retinal ganglion cells. Brain Res., 48: 65-92. 197213 Centrifugal control of the avian retina. 11. Receptive field properties of cells in the isthmo-optic nucleus. Brain Res., 48; 93-113. McGill, J. I., T. P. S. Powell and W. M. Cowan 1966a The retinal representation upon the optic tectum and isthmo-optic nucleus in the pigeon. J. Anat., 100: 5-33. 1966b The organization of the projection of the centrifugal fibres to the retina in the pigeon. J. Anat., 100: 35-49. Ogden, T. E. 1968 On the function of efferent retinal fibres. In: Structure and Function of lnhibitory Neuronal Mechanisms. C. von Euler, S. Skoglund and U. Soderberg, eds. Pergamon Press, Oxford, pp. 89-109. Oyster, C. W. 1968 The analysis of image motion by the rabbit retina. J. Physiol., 199: 613635. Pearlman, A. L., and C. P. Hughes 1973 Functional role of efferents to the retina. Trans. Am. Neurol. Assoc., 98; 9-12. 1976 Functional role of efferents to the avian retina. 11. Effects of reversible cooling of the isthmo-optic nucleus. J. a m p . Neur., 166: 123-1 32. Polyak, S. 1957 The Vertebrate Visual Svstem. The University of Chicago Press, Chicago, pp. 25 C 2 5 2 . Stone, J. 1973 Sampling properties of microelectrodes assessed in the cat retina. J. Neurophysiol., 36; 1071-1079. Stone, J., and K.-P. Hoffman 1972 Very slowconducting ganglion cells in the cat’s retina: a major new -functional type? Brain Res., 43: 610-616. Ter Braak, J. W. G. 1936 Untersuchungen uber optokinetischen nystagmus. Arch. Neerl. Physiol., 21 ; 309-376. Wallenberg, A. 1898 Das mediale Opticusbiindel der Taube. Neurol. Zbl., 17: 5 3 2 5 3 7 . Witkovsky, P. 1971 Synapses made by myelinated fibers running to teleost and elasmobranch retinas. J. Comp. Neur., 142: 205-222. Wyatt, H. J., and N. W. Daw 1975 Directionally sensitive ganglion cells in the rabbit retina: specificity for stimulus direction, size and speed. J. Neurophysiol., 38: 613-626. Yazulla, S. 1974 Intraretinal differentiation in the synaptic organization of the inner plexiform layer of the pigeon retina. J. Comp. Neur., 153: 309-324,
