Directionally
Sensitive Ganglion
Cells in the
Rabbit Retina: Specificity for Stimulus Direction,
Size, and Speed
HARRY J. WYATT AND NIGEL W. DAW Physiology Department, Washington University
DIRECTIONALLY SENSITIVE cells have been described in the visual systemsof many vertebrates, in the retina of some species (l-5, 13, 15-17, 19, 20, 27, 30, 31, 33), and in the cortex of others (10-E). These cells respond most vigorously to a stimulus moved in one direction (called the preferred direction), give no response for stimuli moved in the opposite direction (called the null direction), and give intermediate responses for stimuli moved at an angle to the preferrednull axis. Directionally sensitive ganglion cells in the rabbit retina have been studied in more detail than any others (2, 3), and the mechanism of directional sensitivity is believed to be lateral inhibition, which is asymmetric, extending only in one direction along the preferred-null axis. In the course of some work on the effects of raising rabbits inside rotating striped drums (6), we noticed that rabbit retinal ganglion cells are quite specific for the size of moving stimuli. This size specificity may be quite lax in the preferred direction of motion (a stimulus of any size is quite effective), but tends to be stringently selective in favor of small stimuli moving perpendicularly to the preferred direction: a corollary of this is that many directionally sensitive units show more specificity in their directionality for large stimuli than for small. A similar result has been described for the directionally sensitive ganglion cells in the bird retina (22; A. L. Pearlman and C. P. Hughes, unpublished observations). We hypothesized that the lateral inhibition responsible for directional sensitivity might be the same as the lateral inhibition
Received for publication
September 16, 1974.
Medical
School, St. Louis, Missouri
63110
responsible for size specificity. This paper describes a seriesof experiments designed to test this hypothesis, and also to relate the extent of the lateral inhibition to the velocity sensitivity of the cell. We were able to measure the-extent of inhibition acting on a small area within the receptive field by moving a grating continually through the small area, while another spot moved through the field nearby. Data from this type of experiment are correlated with data from experiments with two briefly flashed stimuli and from experiments with stimuli of different sizes moved in different directions at different speeds.The results provide evidence for the following findings: I) The receptive fields of directionally sensitive units contain inhibitory in fluences that act in all directions except- directly in the preferred direction. 2) Inhibition within the receptive-field center and inhibition from the surround are part of a single inhibitory network. 3) The velocity sensitivity of directionally sensitive units may be predicted from the results of experiments using pairs of flashed stimuli. METHODS
The methods were similar to those described in our earlier paper (6). Briefly, mature pig mented rabbits were anesthetized, paralyzed, and respirated; the extraocular muscles were cut and the eyeball sutured to a stainless steel ring, and tungsten-glass electrodes were inserted through the sclera near the limbus to record from ganglion cells and their axons. Preliminary analysis techniques, such as receptive-field plotting, have already been described (6). Here we emphasize the stimulus conditions used for analyzing the details of the receptive fields of the directionally sensitive cells. Stimuli were gen-
ownloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (192.236.036.029) on September 7, 201 Copyright © 1975 American Physiological Society. All rights reserved.
H. J. WYATT
614
AND
erally 250 cd/m2 projected onto a background illumination of about 10 cd/m? Moving
stimuli
Moving stimuli were projected onto a vertical screen using a slide projector, the beam of which was reflected from a mirror mounted on a galvanometer pen motor near the projector objective. The pen motor was driven by a function generator which produced voltages which were trapezoidal functions of time. The sweep speed and amplitude were adjustable. “Localized”
moving
stimuli
To excite a unit continuously at a single locus of the receptive field, stripes were moved across that area of receptive field, while a mask shielded the rest of the receptive field from illumination. A Lucite disk with radial black and clear stripes was mounted on a variable-speed motor, and the edge of the rotating disk was positioned at the focal plane of a slide projector. The mask was a small circular hole fixed next to the disk near its edge to block off all except a small area of the striped region of the disk. The projected stimulus was thus a small fixed circular region of the tangent screen, across which edges could be moved in one direction at an adjustable speed. Typically, the circular region might subtend 0.5O at the rabbit’s eye, and each stripe might subtend about 0.8O in width. By adjusting the speed, a rhythmic discharge of the unit was usually obtainable. Since the motor was not synchronized with the data-analysis time window, accumulation of several response histograms tended to smear out the rhythmicity of the excitation, and the effect of a second inhibitory stimulus could be measured. Flashed
stimuli
Two thin bars of light were positioned within a receptive field with their length perpendicular to the preferred-null axis and some separation between bars along the preferred-null axis. The bars were made as thin as was consistent with obtaining a reliable response from the unit to a brief flash of either bar. Typically, for onoff directional units, bars O.l” or less thick with exposure time of ?(jo s could be used. For on directional units, reliable responses could only be obtained with thicker bars, about so thick, and with longer exposure times, typically 1/4 s. Each projector was equipped with an electromechanical shutter (Ilex Optical Co.). The shutters were calibrated with a pin diode (United Detector Technology). The timing for shutter release was obtained in early experiments from
N. W. DAW
a series of waveform and pulse generators (Tektronix, Inc.), and in later experiments from a real-time clock in the computer. Data
processing
and storage
Signals were fed into a window, i.e., a pulse generator, which produced a standard square pulse whenever the peak of an action potential fell between two adjustable potential settings. A classic LINC computer used these pulses to compile response histograms, display them, and store them on magnetic tape during an experiment. In generating a response histogram for a moving stimulus, the start of each histogram was signaled to the computer when the lead edge of the stimulus crossed a light sensor placed on the screen. When a response histogram was collected for the paired flashes, the start of the histogram was coincident with the opening of the first shutter in the pair or preceded it by 500 ms. The distance between the bars was set, their histograms were accumulated for nine time intervals (54,280 ms in powers of two) in both preferred and null sequences, then for each flash alone, and for the two bars flashed simultaneously. To minimize errors due to drift in the excitability of the unit under examination, the computer was programmed to present each of these 22 temporal sequences once, then each a second time, and so on until the desired number of presentations had been accumulated. To produce smoothed records, as in Figs. 1 and 2, a simple smoothing program was applied to the recorded experimental data. The contents of a bin were repIaced by the average of the contents of eight bins (the bin in question, the three to its left, and the four to its right). For Fig. 1, this process was repeated 4 times. Smoothed records are presented as the outline of the resultant histogram, without filling in. Sample
of units
We have quantitatively analyzed the properties of 50 well-isolated directionally sensitive ganglion cells in detail. We did not perform all of the experiments described here on any one unit, but each result was found to be typical of several units, (We observed a few atypical units, but have not included data from these.) RESULTS
The directionally semitive ganglion cells in the rabbit retina respond to a flashing spot of light. Most of them give an on-off response (on-off directionally sensitive cells) and, a few give only an on-response (on directionally sensitive cells). The on direc-
ownloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (192.236.036.029) on September 7, 201 Copyright © 1975 American Physiological Society. All rights reserved.
RABBIT
DIRECTIONALLY
SENSITIVE
GANGLION
CELLS
Glci
only the surround is stimulated. There are also inhibitory influences within the center of the receptive field. In the course of our experiments, it soon became apparent that the inhibitory influences in these receptive fields are very extensive in both the space and time domains. Lateral a”6” 4O2” -
20 spikes set
c per 0+
0-
lAO/sec
2O4O 6O\I,,,l,,,,r,,,,,,,,,,,f 10’ 8” 6’ 4’ t
, , 6
2O 0
, , , , 4 2
2’
, , , 0 2
4O 6” ,
,, 4
8’
10’ 12O , ,f 6 8 set
FIG. 1. Inhibition of one area from other points within the receptive field. Exciting spot at point near X, shown by circle drawn to scale. Tracks of inhibiting spot are marked by records of effect of inhibiting spot: horizontal axis marks both position in visual field and time of transit of inhibiting spot, related to each other by speed of movement of inhibiting spot (1.40/s). Rectangle with response histograms across it outlines area responding to a small moving spot: receptive field as mapped with both a small flashed spot (+ and -), and a small moving spot (rectangle of solid lines) is shown at top. Dots outline arca inside which the inhibitory spot reduces the activity below 7.5 spikes/s. Inhibiting spot shown at right drawn to scale. Unsmoothed and smoothed record from path through x shown at bottom.
tionally sensitive cells also respond to much slower speeds of movement (23). The area of retina which yields a response to a stationary flash of light may be defined as the center of the receptive field. This corresponds approximately, but not exactly, to the area over which responses are obtained from small moving spots of light (Figs. 1, 3, and 5). The response of the cell may also be reduced or inhibited if light is shone on surrounding areas (the surround of the receptive field) at the same time as the center is illuminated. No response is obtained if
extent
of inhibition
The inhibition acting on a single point within the receptive field was found to originate from a wide area around that point. The spatial extent of this inhibition was measured by using a localized moving stimulus (see METHODS) to produce a steady excitatory influence within a confined portion of the receptive field, and moving a second spot through the receptive field in the null direction. The second spot generated no excitation by itself, but reduced the excitation from the localized moving stimulus by an amount which depended on the position of both the moving spot and the localized moving stimulus. The results of such an experiment are shown in Fig. 1 for a localized moving stimulus at point x in the center of the receptive field of the cell. The response histograms have been smoothed as described in METHODS (original and smoothed versions of the record going through point x are shown at the bottom of the figure). The smoothed records have been place>1 across the receptive field so that the base line of each record lies along the path actually taken by the inhibitory spot to produce that record. The receptive field is drawn to scale so that the firing rate indicated at each point on each record is the firing rate which occurred when the inhibitory spot was at that point in the receptive field. A contour of equal inhibition has been drawn in as a dotted line: this simply connects together the portions inside which the smoothed frequency was less than 7.5 spikes/s. (As the inhibitory spot left the area from which it influenced the excitatory stimulus, the unit of Fig. 1 showed a postinhibitory rebound in firing rate. To make certain that this was not an excitatory component, the inhibitory spot was sometimes blanked out just prior to the elevation in firing: a similar rebound occurred even when this was done.) The results show that the area inhibiting the excitation of point
ownloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (192.236.036.029) on September 7, 201 Copyright © 1975 American Physiological Society. All rights reserved.
616
H. J. WYATT
[ 20 spikes /-- ..,-/ “‘-‘~-.
Sensitive Ganglion
Cells in the
Rabbit Retina: Specificity for Stimulus Direction,
Size, and Speed
HARRY J. WYATT AND NIGEL W. DAW Physiology Department, Washington University
DIRECTIONALLY SENSITIVE cells have been described in the visual systemsof many vertebrates, in the retina of some species (l-5, 13, 15-17, 19, 20, 27, 30, 31, 33), and in the cortex of others (10-E). These cells respond most vigorously to a stimulus moved in one direction (called the preferred direction), give no response for stimuli moved in the opposite direction (called the null direction), and give intermediate responses for stimuli moved at an angle to the preferrednull axis. Directionally sensitive ganglion cells in the rabbit retina have been studied in more detail than any others (2, 3), and the mechanism of directional sensitivity is believed to be lateral inhibition, which is asymmetric, extending only in one direction along the preferred-null axis. In the course of some work on the effects of raising rabbits inside rotating striped drums (6), we noticed that rabbit retinal ganglion cells are quite specific for the size of moving stimuli. This size specificity may be quite lax in the preferred direction of motion (a stimulus of any size is quite effective), but tends to be stringently selective in favor of small stimuli moving perpendicularly to the preferred direction: a corollary of this is that many directionally sensitive units show more specificity in their directionality for large stimuli than for small. A similar result has been described for the directionally sensitive ganglion cells in the bird retina (22; A. L. Pearlman and C. P. Hughes, unpublished observations). We hypothesized that the lateral inhibition responsible for directional sensitivity might be the same as the lateral inhibition
Received for publication
September 16, 1974.
Medical
School, St. Louis, Missouri
63110
responsible for size specificity. This paper describes a seriesof experiments designed to test this hypothesis, and also to relate the extent of the lateral inhibition to the velocity sensitivity of the cell. We were able to measure the-extent of inhibition acting on a small area within the receptive field by moving a grating continually through the small area, while another spot moved through the field nearby. Data from this type of experiment are correlated with data from experiments with two briefly flashed stimuli and from experiments with stimuli of different sizes moved in different directions at different speeds.The results provide evidence for the following findings: I) The receptive fields of directionally sensitive units contain inhibitory in fluences that act in all directions except- directly in the preferred direction. 2) Inhibition within the receptive-field center and inhibition from the surround are part of a single inhibitory network. 3) The velocity sensitivity of directionally sensitive units may be predicted from the results of experiments using pairs of flashed stimuli. METHODS
The methods were similar to those described in our earlier paper (6). Briefly, mature pig mented rabbits were anesthetized, paralyzed, and respirated; the extraocular muscles were cut and the eyeball sutured to a stainless steel ring, and tungsten-glass electrodes were inserted through the sclera near the limbus to record from ganglion cells and their axons. Preliminary analysis techniques, such as receptive-field plotting, have already been described (6). Here we emphasize the stimulus conditions used for analyzing the details of the receptive fields of the directionally sensitive cells. Stimuli were gen-
ownloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (192.236.036.029) on September 7, 201 Copyright © 1975 American Physiological Society. All rights reserved.
H. J. WYATT
614
AND
erally 250 cd/m2 projected onto a background illumination of about 10 cd/m? Moving
stimuli
Moving stimuli were projected onto a vertical screen using a slide projector, the beam of which was reflected from a mirror mounted on a galvanometer pen motor near the projector objective. The pen motor was driven by a function generator which produced voltages which were trapezoidal functions of time. The sweep speed and amplitude were adjustable. “Localized”
moving
stimuli
To excite a unit continuously at a single locus of the receptive field, stripes were moved across that area of receptive field, while a mask shielded the rest of the receptive field from illumination. A Lucite disk with radial black and clear stripes was mounted on a variable-speed motor, and the edge of the rotating disk was positioned at the focal plane of a slide projector. The mask was a small circular hole fixed next to the disk near its edge to block off all except a small area of the striped region of the disk. The projected stimulus was thus a small fixed circular region of the tangent screen, across which edges could be moved in one direction at an adjustable speed. Typically, the circular region might subtend 0.5O at the rabbit’s eye, and each stripe might subtend about 0.8O in width. By adjusting the speed, a rhythmic discharge of the unit was usually obtainable. Since the motor was not synchronized with the data-analysis time window, accumulation of several response histograms tended to smear out the rhythmicity of the excitation, and the effect of a second inhibitory stimulus could be measured. Flashed
stimuli
Two thin bars of light were positioned within a receptive field with their length perpendicular to the preferred-null axis and some separation between bars along the preferred-null axis. The bars were made as thin as was consistent with obtaining a reliable response from the unit to a brief flash of either bar. Typically, for onoff directional units, bars O.l” or less thick with exposure time of ?(jo s could be used. For on directional units, reliable responses could only be obtained with thicker bars, about so thick, and with longer exposure times, typically 1/4 s. Each projector was equipped with an electromechanical shutter (Ilex Optical Co.). The shutters were calibrated with a pin diode (United Detector Technology). The timing for shutter release was obtained in early experiments from
N. W. DAW
a series of waveform and pulse generators (Tektronix, Inc.), and in later experiments from a real-time clock in the computer. Data
processing
and storage
Signals were fed into a window, i.e., a pulse generator, which produced a standard square pulse whenever the peak of an action potential fell between two adjustable potential settings. A classic LINC computer used these pulses to compile response histograms, display them, and store them on magnetic tape during an experiment. In generating a response histogram for a moving stimulus, the start of each histogram was signaled to the computer when the lead edge of the stimulus crossed a light sensor placed on the screen. When a response histogram was collected for the paired flashes, the start of the histogram was coincident with the opening of the first shutter in the pair or preceded it by 500 ms. The distance between the bars was set, their histograms were accumulated for nine time intervals (54,280 ms in powers of two) in both preferred and null sequences, then for each flash alone, and for the two bars flashed simultaneously. To minimize errors due to drift in the excitability of the unit under examination, the computer was programmed to present each of these 22 temporal sequences once, then each a second time, and so on until the desired number of presentations had been accumulated. To produce smoothed records, as in Figs. 1 and 2, a simple smoothing program was applied to the recorded experimental data. The contents of a bin were repIaced by the average of the contents of eight bins (the bin in question, the three to its left, and the four to its right). For Fig. 1, this process was repeated 4 times. Smoothed records are presented as the outline of the resultant histogram, without filling in. Sample
of units
We have quantitatively analyzed the properties of 50 well-isolated directionally sensitive ganglion cells in detail. We did not perform all of the experiments described here on any one unit, but each result was found to be typical of several units, (We observed a few atypical units, but have not included data from these.) RESULTS
The directionally semitive ganglion cells in the rabbit retina respond to a flashing spot of light. Most of them give an on-off response (on-off directionally sensitive cells) and, a few give only an on-response (on directionally sensitive cells). The on direc-
ownloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (192.236.036.029) on September 7, 201 Copyright © 1975 American Physiological Society. All rights reserved.
RABBIT
DIRECTIONALLY
SENSITIVE
GANGLION
CELLS
Glci
only the surround is stimulated. There are also inhibitory influences within the center of the receptive field. In the course of our experiments, it soon became apparent that the inhibitory influences in these receptive fields are very extensive in both the space and time domains. Lateral a”6” 4O2” -
20 spikes set
c per 0+
0-
lAO/sec
2O4O 6O\I,,,l,,,,r,,,,,,,,,,,f 10’ 8” 6’ 4’ t
, , 6
2O 0
, , , , 4 2
2’
, , , 0 2
4O 6” ,
,, 4
8’
10’ 12O , ,f 6 8 set
FIG. 1. Inhibition of one area from other points within the receptive field. Exciting spot at point near X, shown by circle drawn to scale. Tracks of inhibiting spot are marked by records of effect of inhibiting spot: horizontal axis marks both position in visual field and time of transit of inhibiting spot, related to each other by speed of movement of inhibiting spot (1.40/s). Rectangle with response histograms across it outlines area responding to a small moving spot: receptive field as mapped with both a small flashed spot (+ and -), and a small moving spot (rectangle of solid lines) is shown at top. Dots outline arca inside which the inhibitory spot reduces the activity below 7.5 spikes/s. Inhibiting spot shown at right drawn to scale. Unsmoothed and smoothed record from path through x shown at bottom.
tionally sensitive cells also respond to much slower speeds of movement (23). The area of retina which yields a response to a stationary flash of light may be defined as the center of the receptive field. This corresponds approximately, but not exactly, to the area over which responses are obtained from small moving spots of light (Figs. 1, 3, and 5). The response of the cell may also be reduced or inhibited if light is shone on surrounding areas (the surround of the receptive field) at the same time as the center is illuminated. No response is obtained if
extent
of inhibition
The inhibition acting on a single point within the receptive field was found to originate from a wide area around that point. The spatial extent of this inhibition was measured by using a localized moving stimulus (see METHODS) to produce a steady excitatory influence within a confined portion of the receptive field, and moving a second spot through the receptive field in the null direction. The second spot generated no excitation by itself, but reduced the excitation from the localized moving stimulus by an amount which depended on the position of both the moving spot and the localized moving stimulus. The results of such an experiment are shown in Fig. 1 for a localized moving stimulus at point x in the center of the receptive field of the cell. The response histograms have been smoothed as described in METHODS (original and smoothed versions of the record going through point x are shown at the bottom of the figure). The smoothed records have been place>1 across the receptive field so that the base line of each record lies along the path actually taken by the inhibitory spot to produce that record. The receptive field is drawn to scale so that the firing rate indicated at each point on each record is the firing rate which occurred when the inhibitory spot was at that point in the receptive field. A contour of equal inhibition has been drawn in as a dotted line: this simply connects together the portions inside which the smoothed frequency was less than 7.5 spikes/s. (As the inhibitory spot left the area from which it influenced the excitatory stimulus, the unit of Fig. 1 showed a postinhibitory rebound in firing rate. To make certain that this was not an excitatory component, the inhibitory spot was sometimes blanked out just prior to the elevation in firing: a similar rebound occurred even when this was done.) The results show that the area inhibiting the excitation of point
ownloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (192.236.036.029) on September 7, 201 Copyright © 1975 American Physiological Society. All rights reserved.
616
H. J. WYATT
[ 20 spikes /-- ..,-/ “‘-‘~-.
