Visual Neurosciencehttp://journals.cambridge.org/VNS

Additional services for Visual Neuroscience:

Email alerts: Click hereSubscriptions: Click hereCommercial reprints: Click hereTerms of use : Click here

Orientation and Direction Tuning of Goldfish Ganglion Cells

Joseph Bilotta and Israel Abramov

Visual Neuroscience / Volume 2 / Issue 01 / January 1989, pp 3 ­ 13DOI: 10.1017/S0952523800004260, Published online: 02 June 2009

Link to this article: http://journals.cambridge.org/abstract_S0952523800004260

How to cite this article:Joseph Bilotta and Israel Abramov (1989). Orientation and Direction Tuning of Goldfish Ganglion Cells. Visual Neuroscience,2, pp 3­13 doi:10.1017/S0952523800004260

Request Permissions : Click here

Downloaded from http://journals.cambridge.org/VNS, IP address: 139.184.30.131 on 04 Sep 2012

Visual Neuroscience (1989), 2, 3-13. Printed in the USA.Copyright © 1989 Cambridge University Press 0952-5238/89 $5.00 + .00

Orientation and direction tuningof goldfish ganglion cells

JOSEPH BILOTTA AND ISRAEL ABRAMOVVisual Research Laboratory, Department of Psychology, Brooklyn College of CUNY, Brooklyn, New York

(RECEIVED October 16, 1987; ACCEPTED APRIL 20, 1988)

Abstract

Orientation and direction tuning were examined in goldfish ganglion cells by drifting sinusoidal gratingsacross the receptive field of the cell. Each ganglion cell was first classified as X-, Y- or W-like based on itsresponses to a contrast-reversal grating positioned at various spatial phases of the cell's receptive field.Sinusoidal gratings were drifted at different orientations and directions across the receptive field of the cell;spatial frequency and contrast of the grating were also varied. It was found that some X-like cells respondedsimilarly to all orientations and directions, indicating that these cells had circular and symmetrical fields.Other X-like cells showed a preference for certain orientations at high spatial frequencies suggesting thatthese cells possess an elliptical center mechanism (since only the center mechanism is sensitive to high spatialfrequencies). In virtually all cases, X-like cells were not directionally tuned. All but one Y-like cell displayedorientation tuning but, as with X-like cells, orientation tuning appeared only at high spatial frequencies. Asubstantial portion of these Y-like cells also showed a direction preference. This preference was dependenton spatial frequency but in a manner different from orientation tuning, suggesting that these twophenomena result from different mechanisms. All W-like cells possessed orientation and direction tuning,both of which depended on the spatial frequency of the stimulus. These results support past work whichsuggests that the center and surround components of retinal ganglion cell receptive fields are not necessarilycircular or concentric, and that they may actually consist of smaller subareas.

Keywords: Goldfish ganglion cell', Orientation tuning, Direction tuning

Introduction

The "Difference of Gaussians" receptive field (DOG) modelproposed by Rodieck (1965) assumes a concentric and symmet-ric center and surround organization. With this arrangement,an individual ganglion cell would be incapable of providing anyorientation information. That is, a bar of light or a gratingdrifted across the receptive field of a ganglion cell would yieldthe same response regardless of the orientation or direction atwhich the stimulus crossed the field. However, careful inves-tigation of cat ganglion cell receptive fields has revealed thatthis is not the case (Rodieck & Stone, 1965). For example, whenthe center mechanisms of cat ganglion cells were mapped withsmall spots of light, it was found that the center was not cir-cular but elliptical (Hammond, 1974).

Levick and Thibos (1982) drifted sinusoidal gratings at dif-ferent orientations across the receptive fields of cat ganglion

This work is based in part on a dissertation submitted by J. Bilottain partial fulfillment of the requirements for a Ph.D. Degree from theCity University of New York, NY, USA.

Reprint requests to: Joseph Bilotta, Department of Psychology, 134Wesley Hall, Vanderbilt University, Nashville, TN 37240, USA.

cells and found that the majority of cells possessed an orienta-tion bias. However, the bias was found only when high spatialfrequency gratings were presented. At low spatial frequencies,the cells behaved as if they were circularly and concentricallyorganized, whereas at high spatial frequencies they clearly re-sponded as if the receptive fields were elliptical. These findingssupport the notion that cat ganglion cell centers are ellipticalsince the center component dominates the response at high spa-tial frequencies.

Soodak et al. (1985, 1987) have investigated orientation tun-ing in the cat retina and the lateral geniculate nucleus (LGN).They confirmed the findings of Levick and Thibos that gan-glion cells are orientation tuned at high spatial frequenciesbecause of an elliptical center mechanism. They also discoveredthat the cell's response to different orientations changed as afunction of spatial frequency. At low to moderate frequencies,there were two peaks in response which were at orientations 180deg apart; at very high spatial frequencies, there were four ormore peaks. Soodak et al. (1985, 1987) have suggested that thischange in the orientation tuning curve occurs because the centercomponent is not a single, uniform mechanism but consists ofsmaller subareas. The four-peaked orientation tuning curve

J. Bilotta and I. Abramov

found at high spatial frequencies can be best explained by theexistence of two adjacent subareas in the center (Soodak, 1986).Thus, at moderate spatial frequencies, the center will appearelliptical, and only at high spatial frequencies will the subareasbecome apparent. These findings were robust for both X- andY-cells.

Although Soodak et al. (1985, 1987) found that most cellswere clearly orientation tuned, there were a few cases in whichthey also displayed a slight direction bias. Unfortunately, dueto the nature of their recording technique, few W-cells wereexamined; W-cells are the most likely candidates for directiontuning in the retina (see Rodieck, 1979). There is some evidencethat direction tuning exists in the cat retina. Dawis et al. (1984)found that the receptive fields of most X-cells in the cat retinaare somewhat asymmetric. Thus, a grating drifting across thereceptive field in one direction can produce a different responsethan if the same grating were drifted in the opposite direction.

Turning to the goldfish, there is currently little evidence thatganglion cells are orientation or direction biased. Despite theexistence of orientation and direction tuned cells in highervisual centers, such as the optic tectum (Cronly-Dillon, 1964;Jacobson & Gaze, 1964; Wartzok & Marks, 1973; Riemslag &Schellart, 1978), there have been only a few examples of direc-tion or orientation tuned cells in recordings from optic nerve(Daw & Beauchamp, 1972; Riemslag & Schellart, 1978).

This lack of tuning at the ganglion cell level is quite puz-zling, since there is both anatomical and physiological evidencethat, at the very least, orientation tuning should be found atthis level. For example, the dendritic spread of teleost bipolarcells clearly shows asymmetries in the form of elliptical fields(Ishida et al., 1980). Similar findings have been reported in carpganglion cells (Kock & Reuter, 1978).

Levine and Zimmerman (1985) have investigated the re-sponses of the subareas of the center and surround portions ofthe goldfish ganglion cell. By examining the ON and OFF re-sponses to small spots of light, they were able to map the re-sponse characteristics of the subareas. They found that thesubareas were not uniform in their responses within the centerand surround mechanisms. In fact, many of the cells examinedhad receptive field properties that were much more complicatedthan the DOG model would predict; i.e. they were generallyelongated and asymmetrical. Elliptical fields and asymmetrieswithin the fields would suggest that goldfish ganglion cells arecapable of providing orientation and direction information.

In this study, orientation and direction tuning were exam-ined in goldfish ganglion cells using an approach similar to thatof Levick and Thibos (1982) and others in the cat retina. Iforientation tuning in the goldfish retina is a function of stim-ulus spatial frequency, then it is possible that the stimuli usedin past work on the goldfish (spots and bars of light) could notreveal tuning. Evidence suggests that the goldfish ganglion cellreceptive field contains irregular subareas which can only betested adequately with stimuli of high spatial frequencies.

An interesting question regarding orientation and directiontuning concerns the properties of W-like cells. Although thesecells are the best candidates for this type of tuning, they havenot been adequately assessed in the past. The reason for thisscarcity of W-like cells, in general, is that they are difficult toisolate and maintain in cat preparations. This is not a problemfor our goldfish preparation, where maintaining stable isolationof W-like cells is quite routine (see Bilotta, 1987; Bilotta &Abramov, 1985).

Materials and methods

Preparation and recording

Common goldfish (Carassius auratus), 10-15 cm in length,were housed in an aquarium where the water temperature wasmaintained at about 21°C; illumination was on a light/darkcycle of 12 h each. Two to four hours prior to surgery, a fishwas placed in a dark-adaptation tank. The animal was thensacrificed by decapitation, the eye enucleated and hemisected.The retina was removed from the eyecup and placed, receptorside up, on a glass plate located within an isolation chamber.The chamber was maintained at a temperature of 17°C andmoist oxygen (100%) was passed through the chamber at a rateof 75 ml/min. (See Abramov & Levine, 1972; Mackintosh etal., 1987, for more details.)

Extracellular recordings from single ganglion cells weremade with glass-insulated platinum-iridium microelectrodes(Wolbarsht & Wagner, 1963). The electrode was lowered intothe retina from the receptor side; a platinum-iridium indiffer-ent electrode was placed on the edge of the retina. Ganglion cellresponses were amplified, filtered, and recorded, together withcoded stimulus information, by a computer for later analysis.Summary statistics and an on-line Fourier decomposition of theresponse to each stimulus were used to guide the experimenter.

Optical stimulator

The optical stimulator consisted of a high-resolution oscillo-scope (CRT) (Tektronix; Model 606, P31 phosphor) whichdisplayed the output of an electronic visual stimulator (seeMilkman et al., 1978). This stimulator was capable of gener-ating sinusoidal gratings whose spatial frequency, orientation,contrast, and temporal frequency were independently variable.It was also capable of generating contrast-reversal gratingswhose spatial position (phase) could be manipulated, as well asgratings drifting in any direction.

Each pattern was modulated around some mean luminance;the contrast of the stimulus was defined as: (maximum lumi-nance — minimum luminance)/(maximum luminance + mini-mum luminance). The CRT display was projected onto theretina by a high-quality camera lens; the display was restrictedto a 7.5-mm diameter circle on the retina and produced a reti-nal illuminance of 0.2 lm/m2. (See Bilotta, 1987, for moredetails.)

Procedures

Once a cell was isolated, the CRT display was turned on andthe retina was allowed to adapt to the illumination. At thispoint, contrast-reversal gratings, presented at various spatialphases, were used to determine whether the cell possessed aspatial null point. To provide an estimate of the cell's spontane-ous rate, a contrast-reversal grating of zero contrast was pre-sented with phase marks at a specified temporal rate. Aftereach stimulus was presented, the computer performed an on-line Fourier decomposition of the averaged response, and thefirst and second harmonics were displayed.

Once enough information was obtained to classify the cellby its spatial summation properties (see below), sinusoidal grat-ings of different spatial frequencies (cycles per millimeter on theretina (cy/mm)) were drifted across the cell's receptive field at

Orientation tuning

a constant temporal rate (cycles per second across the field(Hz)). To examine the orientation and direction tuning of eachcell, the drifting gratings were presented at different orienta-tions in 45-deg steps from 0 to 135 deg. For each orientation,the grating was drifted in each of the two possible directions;that is, drifting gratings of 0 and 180 deg had the same orien-tation, but moved in opposite directions.

RESULTS

Spatial summation classification

The spatial summation properties of each cell were determinedby examining its response to a contrast-reversal grating at dif-ferent spatial phases with respect to the cell's receptive field.The cell's responses to each stimulus cycle were superimposedto give the average response per cycle. The cycle was dividedinto discrete time bins and the average number of spikes per binwere converted into spike rates; a discrete Fourier transformyielded the spectrum of the response. The criteria and responseproperties used to determine the spatial summation class ofthese cells were the same rigorous classification scheme used oncat neurons (see Hochstein & Shapley, 1976a).

A cell was classified as X-like if all of the following crite-ria were met: (1) The cell possessed a null point; that is, therewas a spatial position at which there was no response to acontrast-reversal grating. (2) When the grating was positionedaway from the null point, the response was modulated at thesame frequency as the contrast-reversal stimulus. (3) The ampli-tude of the fundamental component of the response was a si-nusoidal function of spatial phase.

A cell was classified as Y-like if it responded at double thestimulus frequency at all spatial phases and did not possess anull point. This criterion had to be met at high, but not nec-essarily at low, spatial frequencies. Cells that could not be clas-sified into the above categories were labelled W-like cells. Manyof these W-like cells appeared X-like at most spatial positionsof the contrast-reversal grating, but at spatial positions near thenull point, the modulation of the response changed from thefundamental component to the second harmonic componentand there was no true null. This classification scheme has beenquite useful in categorizing ganglion cells in nonmammalianvertebrates (e.g. eel; Shapley & Gordon, 1978) including gold-fish (Bilotta, 1987; Bilotta & Abramov, 1985, 1988).

ro

so.

so

X-like cell <cioc)

Amplitude of Fundamental Component

Drilling grating, 4 Hz, 40% contrast. N: 20

SO

(a) 0.19 cy/mmSponUnvoua rala: 0

SO 120 ISO 240 0 0 3 0 60 90A1S0 210 240 270

(b) 0.76 cy/mm

80 120 180 240 O 6 SO SO 904160 210 240 270

30

20

V — ^

60 120 180 240 O 6 30 SO 90A1S0 210 240 270

Orientation (Degrees)

Fig. 1. Orientation tuning of an X-like cell. The response measure wasthe amplitude of the fundamental component. The responses of the cellto a grating drifting at 4 Hz at various orientations and at several spa-tial frequencies (a, b, and c) are shown. The stimulus contrast in eachcase was 40%. The spatial frequency of the grating (cy/mm) is givenat the bottom of each figure. "N" refers to the number of stimuluspresentations; the cell identification code is shown in parentheses. Thiscell had no spontaneous rate. To examine direction tuning, the orien-tation axis in each figure (a, b, and c) was redrawn such that orienta-tions 180 deg out-of-phase to one another were superimposed. Theseredrawn figures are shown in (d), (e), and (f).

Orientation and direction tuning

X-like cellsSince X-like cells are linear, the response measure used to

examine orientation and direction tuning was the amplitude ofthe fundamental component. Four out of the ten X-like cellstested for orientation and direction bias displayed orientationtuning. Typical results from an orientation biased X-like cellare shown in Fig. 1. The drift rate of the grating (40% contrast)was 4 Hz. The ordinate is the amplitude of the fundamentalcomponent of the response. The spontaneous rate of this cellwas zero spikes/s. As can be seen in this figure, the degree oforientation tuning depended upon the spatial frequency of thestimulus grating. At low spatial frequencies (Fig. la), there wasrelatively little orientation tuning, but as the spatial frequencyof the stimulus was increased, orientation tuning increased(Figs, lb.lc).

Since drifting gratings of 0 and 180 deg have the same orien-tation but move in opposite directions, the responses of the cellto these two stimuli were superimposed on the orientation axisto test for direction tuning. This was done for all values 180 degout-of-phase. The replotted values of the data in Figs, la, lb,and lc are shown in Figs. Id, le, and If, respectively. In Figs, leand If, the two curves superimpose, indicating no directionbias; that is, the cell responded similarly to gratings at the sameorientation but moving in opposite directions. From Figs, leand 1 f, it can be seen that the cell was orientation tuned sincethe functions were not straight horizontal lines. However, theresponses in Fig. Id show a slight direction bias at low spatialfrequencies since the responses to gratings 180 deg apart did notsuperimpose on one another. This was the only X-like cell iso-lated that displayed any indication of direction tuning.

J. Bilotta and I. Abramov

ao

O 60<oCO

•**.COCDJ£"Q.

2 40(DCOcoQ .COCD11 20

X-like cell (ciooAmplitude of Fundamental ComponentDrifting grating, 4 Hz, 40% contrast. N: 20

X-like cell <C52B>Amplitude of Fundamental Component

• • Pr*f«rr«d orUnt. (0°)

O—O Non-pr.l.rr.d Orl.nl. (9O»I

0.19 0.38 0.76 1.52 3.05 6.10

Spatial Frequency (cy/mm)

Fig. 2. Spatial resolution of an X-like cell at the preferred and non-preferred orientations. The response measure was the amplitude of thefundamental component. The stimuli consisted of a grating drifting at4 Hz at 40% contrast at the preferred (closed squares) and the non-preferred (open hexagons) orientations. "N" refers to the number ofstimulus presentations; the cell identification code is shown in paren-theses. The spontaneous rate of the cell was zero.

Figure 2 compares the spatial resolution of the cell in Fig. 1for gratings at the preferred and nonpreferred orientations. Atthe lower spatial frequencies the responses were similar, but asspatial frequency increased, the responses at the two orienta-tions differed. These findings suggest that only the mechanismresponsible for responses at high spatial frequencies (i.e. thecenter component of the receptive field) was not circular sincethis is where the orientation differences occurred.

The remaining six out of ten X-like cells displayed little orno orientation tuning. Figure 3 shows one such cell at severalspatial frequencies. Once again the response measure was theamplitude of the fundamental component; the drift rate was 2Hz and the contrast was 40%. The broken line represents theamplitude of the fundamental component when only meanluminance was present (i.e. the cell's spontaneous rate). As canbe seen, there was little difference in the response across orien-tations, even at high spatial frequencies. However, at the high-est spatial frequency (1.52 cy/mm), there did appear to be anirregularity in the function which is discussed below.

Y-like cellsSince Y-like cell responses to a drifting grating contain both

linear and nonlinear components, the maximum response minusthe minimum response (i.e. the peak-to-peak amplitude) wasused to examine the overall output of the cell. This measure isquite reliable in examining the nonlinear properties of goldfishganglion cells (Bilotta, 1987; Bilotta & Abramov, 1985, 1988).For comparison, the fundamental component of the cell's re-sponse was also derived. All but one of the 23 Y-like cells testeddisplayed orientation tuning. However, as in X-like cells, thedegree of orientation tuning depended upon the spatial fre-

u<DCO

CO

'5.2 30

CDCO

O 20aCOCD

40-,

30-

3 0

20

10-

Drifting

- ^

(a) 0.38

grating.

• «~-~

cy/mm

2 Hz 40% contrast. N:

v^*—*~~^

20

60 120 180 240 300

(b) 0.76 cy/mm

60 120 180 240 300

(c) 1.52 cy/mm

0 60 120 180 240 300

Orientation (Degrees)

Fig. 3. Orientation tuning of an X-like cell with a circular receptivefield. The response measure was the amplitude of the fundamentalcomponent. The responses to a grating drifting at 2 Hz at various orien-tations, and at several spatial frequencies (a, b, and c), are shown. Thecontrast of each stimulus was 40%. The spatial frequency of the gratingis given at the bottom of each figure. "N" refers to the number of stim-ulus presentations; the cell identification code is shown in parentheses.The dotted line in (a) represents the response amplitude of the funda-mental component to a stimulus of zero contrast.

quency of the stimulus. Figure 4 shows the orientation tuningof a Y-like cell and its dependence on spatial frequency. Theresponse measure was max-min, the drift rate of the stimuluswas 4 Hz, and the contrast for each stimulus is shown withineach figure. The max-min response with no stimulus contrastis represented by the dashed line in Figs. 4d and 4e.

At low spatial frequencies the orientation tuning was weak(Fig. 4a), but as the spatial frequency of the stimulus increased,so did the amount of orientation tuning (Figs. 4b, 4c). At evenhigher spatial frequencies (Figs. 4d, 4e), several preferred orien-tations were seen, and in some cases, the orientation of max-imal response differed from the maximum at lower spatialfrequencies. For example, at 0.76 cy/mm (Fig. 4b), the pre-ferred orientation was roughly 180 deg, while the nonpreferredorientation was at 90 deg. The shape of the function resembleda "W," which is what would be expected if the receptive fieldmechanism was elliptical. The same holds true for 1.52 cy/mm(Fig. 4c). However, at 3.05 cy/mm and 6.10 cy/mm (Figs. 4d,4e, respectively), the preferred orientation was not at 180 degbut at 90 deg; an orientation of 180 deg at these high spatialfrequencies actually produced the minimum response.

These changes in the orientation function at the higher spa-tial frequencies were not due to the nonlinear subunits foundin Y-like cells (Hochstein & Shapley, 19766) because they alsooccurred when the response measure was the fundamental com-

Orientation tuning

Orientation Tuning

Y-like cell <c«»>

Y-like Cell (C52A) Max-Min ResponseDrifting grating. 4 Hz. 25% contrast. N: 56

Drilling grating. * Hi. N: 40

fr '•(•) 0.38 cy/mm. 13% contrail

tO ttO ISO 140 300

(b) 0.76 cy/imn. 13% contrail

O «0 ISO ISO 140 300

(d) 3.05 cy

0 90 130 ISO 240 100 0 tO 120 110 340 300

Orientation (Degrees)

Fig. 4. Orientation tuning of a Y-like cell. The response measure wasthe maximum response minus the minimum response. The responses ofthe cell to a grating drifting at 4 Hz at various orientations, and atseveral spatial frequencies (a, b, c, d, and e) are shown. The spatial fre-quency and contrast of the grating are given at the bottom of each fig-ure. "N" refers to the number of stimulus presentations; the cellidentification code is shown in parentheses. The dashed line representsthe max-min response to a grating of zero contrast.

ponent; the nonlinear subunits of Y-cells are not seen in thefundamental component response measure. There was alsosome evidence for this unusual orientation tuning in X-likecells. Referring back to the X-like cell in Fig. 3, the slight fluc-tuations found in the function across orientation in Fig. 3ccould be due to the same phenomenon.

Figure 5 shows another Y-like cell which verifies that orien-tation tuning changes with spatial frequency and is maximal atintermediate frequencies. The response was max-min, the driftrate was 4 Hz, and the contrast of the grating was 25%. Thespontaneous rate is indicated by the dashed line. At low andhigh spatial frequencies, responses at the preferred and non-preferred orientations were similar, but at intermediate spatialfrequencies the responses at the two orientations became dis-parate.

Eight out of 15 Y-like cells tested displayed direction tun-ing (i.e. responses to stimuli 180 deg apart were not similar).As with orientation tuning, direction tuning depended on thespatial frequency of the stimulus. However, the dependence

IOO n

aCO

«

a.

20

O — O Non-pratcrrod orlont. (270")

0.19 0.38 0.76 1.52 3.05 6.10Spatial Frequency (cy/mm)

Fig. 5. Spatial resolution of a Y-like cell at the preferred and non-preferred orientations. The response measure was the maximum re-sponse minus the minimum response. The stimuli consisted of a gratingdrifting at 4 Hz at 25% contrast at the preferred (closed squares) andthe nonpreferred (open hexagons) orientations. "N" refers to the num-ber of stimulus presentations; the cell identification code is shown inparentheses. The dashed line represents the max-min response to a grat-ing of zero contrast.

on spatial frequency for direction tuning was not the sameas for orientation tuning. For example, the Y-like cell inFig. 6 clearly showed a direction preference at low spatialfrequencies (Fig. 6a), but not at higher spatial frequencies(Fig. 6b). The response measure was max-min, the drift ratewas 4 Hz, and the contrast was 40%. The spontaneous rate isshown in Fig. 6b by the dashed line, and orientations 180 degapart were superimposed on the abscissa. Note that in this cellthere was orientation tuning at both spatial frequencies (sincethe functions were not horizontal straight lines) but there wasno direction bias at the higher spatial frequency. This findingimplies that orientation and direction tuning are two distinctphenomena even though they both depend on the spatial fre-quency of the stimulus.

The distinction between orientation and direction tuning isexplored further in Fig. 7, which displays results from a Y-likecell for both the max-min response and the fundamental com-ponent. The stimulus was a drifting grating at 4 Hz and waspresented at various contrasts. The responses to a zero contraststimulus for the max-min response (Figs. 7c, 7d) and the fun-damental component (Figs. 7f, 7g, 7h) are shown by the dashedlines. Once again, gratings whose orientations were 180 degapart were superimposed on the abscissa to examine directiontuning.

At low spatial frequencies (Figs. 7a, 7e), there was little ori-entation tuning but at intermediate spatial frequencies (Figs. 7b,7f), both orientation and direction tuning became apparent. Athigher spatial frequencies (Figs. 7c, 7g), only orientation tun-

J. Bilotta and I. Abramov

ISO-

'S ISOa>CO

"5CD

£ noomoa.«

9 0

70-

Y - I I k e Cell (C46B) Max-Min ResponseDrifting grating, 4 Hz, 40% contrast. N: 40

(a) 0.38 cy/mm

e o ao «o 90i1IO 210 240 270

120 ISOS00 3»0

9 0

70

SO'

3 0 '

•

1 0 •

AA •/ V/ V/ V

/ V/ V/ \\J \\\\\\

\\V\"̂

1 1.52 cy/mm

O 6 30 6*0 90 120 ISOA1S0 210 240 270 300 330

Orientation (Degrees)

Fig. 6. Orientation and direction tuning of a Y-like cell. The response measure was the maximum response minus the mini-mum response. The responses of the cell to a grating drifting at 4 Hz at various orientations, at two spatial frequencies (a andb), are shown. The contrast of each grating was 40%. The spatial frequency of the grating is given at the right of each fig-ure. "N" refers to the number of stimulus presentations; the cell identification code is shown in parentheses. Orientation val-ues 180 deg out-of-phase are superimposed on the abscissa. The dashed line represents the max-min response to a grating ofzero contrast. Note that at low spatial frequencies (0.38 cy/mm) the cell is both orientation and direction biased (a), but athigh spatial frequencies (1.52 cy/mm) the cell is only orientation tuned (b).

ing was seen; at the highest spatial frequencies (Figs. 7d, 7h),there was no direction tuning but possibly the unusual orien-tation tuning seen in other cells. However, the responses weretoo close to the spontaneous rate to be sure. Since the findingshold for both the max-min and the fundamental componentresponse measures, orientation and direction tuning are prob-ably not due to the nonlinear subunits found in Y-like cells'receptive fields.

W-like cellsSince W-like cells are also nonlinear, the max-min response

was used to test for orientation and direction tuning. All 15 W-like cells displayed orientation and direction tuning, and bothtypes of tuning were dependent on the spatial frequency of thegrating. Figures 8 and 9 show typical responses of W-like cellsto drifting gratings at different orientations. In both figures(Fig. 9 shows two cells), max-min is the response measure andthe responses to gratings 180 deg apart are superimposed. Thedrift rate was 2 Hz for Fig. 8, and 4 Hz for Fig. 9. The con-trasts are indicated in each figure and the spontaneous rates arerepresented by the dashed line or indicated in the figure. Al-though W-like cells always displayed orientation tuning, therewas very little consistency in the magnitude of the orientation

tuning across spatial frequencies; there was no regular patternas seen in X- and Y-like cells. Direction tuning in W-like cellsalso displayed no consistent pattern across spatial frequencies.However, a spatial frequency could always be found where thecell displayed no direction bias and that was usually a high spa-tial frequency. For example, the cell shown in Figs. 8a and 8bshowed both orientation and direction tuning, but at a higherspatial frequency there was only orientation tuning (Fig. 8c).Similar results are shown in Fig. 9; at low spatial frequencies,the cell had both types of tuning (Figs. 9a, 9d, 9e), but athigher spatial frequencies, there was only orientation tuning(Figs. 9b, 9c, 9f). Finally, there was no indication that W-likecells possessed any of the unusual tuning at very high spatialfrequencies found in Y- and possibly in X-like cells. Onceagain, the W-like cells' characteristics appear to be differentfrom X- and Y-like cells (Bilotta & Abramov, 1985).

Discussion

To be consistent with the DOG model, ganglion cell receptivefields should be composed of concentric, circular center andsurround mechanisms. We have shown that many ganglion cellsin goldfish retina display orientation tuning and that some of

Orientation tuning

Y-IIke Cell (C51A) Drilling grating. 4 Hi. N: 40

Mai-MIn Response Amplitude of Fundamental Component

•O-i

• 0

• (b) 1.52 cy/mm, 13% contrast

(c) 3.0S cy/mm, 40% contrast

(d) 6.10 cy/mm, 95% contrast

120300

Orientation (Degrees)

Fig. 7. Spatial frequency dependence of orientation and direction tuning of a Y-like cell. The response measures are the max-imum response minus the minimum response (a, b, c, and d) and the amplitude of the fundamental component (e, f, g, andh). The responses of the cell to a grating drifting at 4 Hz at various orientations at several spatial frequencies are shown. Thespatial frequency and contrast of the grating are given at the bottom of the max-min response figures. The arrows indicatethat the paired figures represent different response measures to the same stimulus (e.g., (a) and (e) result from the same stimulus)."N" refers to the number of stimulus presentations; the cell identification code is shown in parentheses. The dashed lines representthe response of the cell to a grating of zero contrast. Orientation values 180 deg out-of-phase are superimposed on the abscissa.This figure demonstrates that orientation and direction tuning are found in both the max-min and the fundamental compo-nent response measures.

10 J. Bilotta and I. Abramov

W-I Ike cell (C08A) Max-Min Response

Drifting grating, 2 Hz, 25% contrast. N: 10

100 160

09

Q.CO

oa

140

120

100

(•) 0.19 cy/mm

120

100

80

60

40

(b) 0.76 cy/mm

O 0 SO 60 90A1«0 210 240 270

(c) 1.52 cy/mm

O 0 30 60 90A 180 210 240 270

Orientation (Degrees)

e oA 180

30210

eo240

90270

Fig. 8. Orientation and direction tuning of a W-like cell. The response measure was the maximum response minus the mini-mum response. The response of the cell to a grating drifting at 2 Hz at various orientations, at several spatial frequencies (a,b, and c) are shown. The contrast of each grating was 25%. The spatial frequency of the grating is given at the bottom of eachfigure. "N" refers to the number of stimulus presentations; the cell identification code is shown in parentheses. The dashedline in (a) represents the max-min response to a grating of zero contrast. Orientation values 180 deg out-of-phase were superim-posed on the abscissa.

them display direction tuning as well. These findings are indirect contrast to the predictions of the DOG model. However,with a few minor modifications of the model, these results canbe explained.

The orientation tuning of many cells suggests that theirreceptive fields are not circular but elliptical. This is supportedby the fact that the orientation producing the maximumresponse is orthogonal to the orientation of minimum response.Since orientation tuning occurs primarily at high spatial fre-quencies where the center component dominates the response,the center must be elliptical. This finding is consistent withcomparable work with cat ganglion cells (Levick & Thibos,1982; Soodak et al., 1985, 1987). The goldfish ganglion cellreceptive fields also appear to be organized into smallersubareas as found in the cat (Soodak, 1986). This is suggestedby the finding that orientation tuning changes drastically atvery high spatial frequencies. As in the cat, these subareas arenot the nonlinear subunits found only in Y-like cells, sinceorientation tuning occurs in X-like cells, and is also prominentin the fundamental response component of the Y-like cells'responses. These results are supported by the work of Levineand Zimmerman (1985). Not only do they show that the cen-ter portion of many goldfish ganglion cells is not circular, butthey also find that the center portion consists of smallersubareas with different ON/OFF response patterns. Thus, astimulus of appropriately high spatial frequency could activate

these subareas and produce a response pattern different froma stimulus consisting of a lower spatial frequency in which theresponses of these areas are merely averaged.

Another modification of the DOG model must be intro-duced to account for direction tuning. Dawis et al. (1984) haveproduced such a model referred to as the "Modified Differenceof Gaussian" model. The modification is that the center andsurround components need not be concentric. By positioningthe center component slightly off-axis with respect to the sur-round, it is possible to produce directionally biased cells, evenin cells with linear spatial summation (X-cells). From thismodel, direction tuning can be explained by differences in thephase relations between the center and surround responses.Because of the asymmetry, a stimulus moving in one directioncould create a situation in which the center and surroundresponses are in-phase (producing a maximal response) whilethe same stimulus moving in the opposite direction will causethe center and surround responses to be out-of-phase (produc-ing a minimum response).

There are two findings which support this model for gold-fish ganglion cells. One finding is from the work of Levine andZimmerman (1985) who found that the majority of goldfishganglion cell receptive fields possessed some form of inhomo-geneity; for example, surround component responses that werestronger on one side of the field, or subareas with differentresponse patterns located adjacent to one another (R. Zimmer-

Orientation tuning 11

W H H f COMa Max-Min Rcspons* Drilling grating, 4 Hi. N: 40

110

100

• 0

(C49B)

; t (a) 0.38 cy/mm, 13% contrast

• 0t40

100 1

100

(C4SB)

' : (d) 0 38 cy/mm. 25% contrast

O 04 ISO

•oS40

120SOO

u•

•w

a

•coa

C

110 '

100

so

so

1«\\\\\\\\\\

1

: (b) 1.62

\

\

cy/mm,

/ /

/ /

/

13% contrast

• 041*0

•0140

120SOO

140

110

100

1.52 cy/mm, 40% contrast

O 0Aiao

00240

110SOO

•0

40

10

(c) 3.06 cy/mm, 25% contrast

441*0

00140 ,s§

100'

ao

• 0

JJ

: (I)

/ /

/

3.05 cy/mm.

^ ^

\

\

\

••«• rat*: IS

95% contrast

\ »\

O 04iao

oo140

Orientation (Degreos)

Fig. 9. Orientation and direction tuning of two W-like cells. The response measure was the maximum response minus the min-imum response. The responses of two cells to a grating drifting at 4 Hz at several spatial frequencies are shown. The spatialfrequency and contrast of each grating are given at the bottom of each figure. Orientation values 180 deg out-of-phase aresuperimposed on the abscissa. "N" refers to the number of stimulus presentations; the cell identification codes are shown inparentheses. The max-min response of the first cell (a, b, and c) to a grating of zero contrast is represented by the dashed linein (c). The response of the second cell (d, e, and f) to a zero contrast grating was 15 spikes/s.

12 J. Bilotta and I. Abramov

man, pers. comm.). Any of these, with the appropriate stimuli,could produce a directionally biased response.

The second piece of evidence which supports this modelstems from the fact that direction tuning may be a function ofthe center/surround interaction. When the stimulus activatesonly one receptive field component, there can be no interactionand thus, no direction tuning. A consistent finding in this studyis that for all cells that displayed direction tuning, the biasoccurred only at the lower frequencies. For example, in Fig. 6at the low spatial frequency, the cell displayed both orientationand direction tuning; however, at the high spatial frequency,the direction bias literally "disappeared" even though orienta-tion tuning was still apparent. This was true for all cells, includ-ing W-like cells. Note that this does not apply to orientationtuning which further emphasizes that orientation and directionbiases are two independent phenomena produced by two dif-ferent mechanisms. Orientation tuning appears to depend onthe shape of the summing area within a given mechanism (e.g.center), whereas direction biases result from the interaction orphase relationships between mechanisms. However, this modelcannot entirely explain the dramatic direction tuning found inmany cells, especially since the overwhelming majority of direc-tionally biased cells were nonlinear in their spatial summationproperties (i.e. Y- and W-like). An alternative, and perhapscoexistent, explanation stems from work on the rabbit retinain which direction selectivity is presumed to be the result of asubarea or portion of the receptive field which when stimulatedfirst inhibits the response of an adjacent area (the null direc-tion), but not if stimulated after the adjacent subarea (the pre-ferred direction) (see Barlow & Levick, 1965).

It is somewhat puzzling why orientation and direction tun-ing have not been reported in other studies of goldfish ganglioncells. The same is true for cat ganglion cells —some studiesreport orientation tuning (Levick & Thibos, 1982; Soodak etal., 1985) whereas others show no evidence and suggest that thereceptive fields are circular and concentric (e.g. Kuffler, 1953).One possibility is that the studies which found no orientationtuning used stimuli that were not adequate to examine the typeof specialized tuning found in both the goldfish and cat gan-glion cells. It is interesting, and most likely significant, that themajority of studies which report orientation tuning used sinu-soidal gratings of various spatial frequencies as stimuli. Theother studies examined orientation and direction tuning withspots or bars of light which passed through the receptive field.Based on the findings of this study, it is quite clear that the tun-ing characteristics depended on the stimulus parameters. Sincea bar of light consists of a variety of spatial frequencies, it ispossible that any orientation or direction tuning will be ob-scured or averaged out with this stimulus.

To summarize the important aspects of goldfish ganglioncells, we have shown in our earlier work (Bilotta & Abramov,1985, 1988), that the majority of X- and Y-like cells possess acenter and surround organization. However, these receptivefield components are not necessarily circular or concentric.Some cells may possess an elliptical center and a circular sur-round. It is also possible for a cell to possess a slightly ellipti-cal surround as well. The receptive field centers are notnecessarily concentric within the surround field. Finally, somecells may have a center mechanism which is slightly displacedfrom a completely concentric organization. Although a modi-fied DOG model can adequately describe the behavior of a gan-glion cell under most circumstances, one should also be aware

that this model is not complete. The center and possibly thesurround components can be further subdivided into smallersubareas. For the most part, the responses of these subareasremain obscured and averaged into oblivion; however, with theappropriate stimuli (stimuli of high spatial frequency) theseareas can alter the response pattern of the cell.

Acknowledgments

This work was supported by grants from the N1H Eye Institute, No.EY01697 and from the PSC-BHE Research Award Program of the CityUniversity of New York, No. 11188, No. 665192, and No. 666344. J.Bilotta was supported by NIH Eye Institute Grant No. EY-06088 duringthe final preparation of this manuscript. We thank Jeffrey S. Farbmanfor his help in the data collection and analysis, Drs. James Gordon andMaureen K. Powers, and Paul J. DeMarco for their critical evaluationand comments of this manuscript.

References

ABRAMOV, I. & LEVINE, M.W. (1972). The effects of carbon dioxide onthe excised goldfish retina. Vision Research 12, 1881-1895.

BARLOW, H.B. & LEVICK, W.R. (1965). The mechanism of direction-ally selective units in rabbit's retina. Journal of Physiology 178,477-504.

BILOTTA, J. (1987). Spatial and spectral properties of the goldfish ret-ina. Ph.D. Dissertation, City University of New York.

BILOTTA, J. & ABRAMOV, I. (1985). Spatial properties of goldfish gan-glion cells. Investigative Ophthalmology and Visual Science (Suppl.)26, 117.

BILOTTA, J. & ABRAMOV, I. (1988). Spatial properties of goldfish gan-glion cells. (Submitted for publication).

CRONLY-DILLON, J.R. (1964). Units sensitive to direction of movementin goldfish optic tectum. Nature 203, 214-215.

DAW, N.W. & BEAUCHAMP, R.D. (1972). Unusual units in the goldfishoptic nerve. Vision Research 12, 1849-1856.

DAWIS, S., SHAPLEY, R.M., KAPLAN, E. & TRANCHINA, D. (1984). Thereceptive field organization of X-cells in the cat: spatiotemporalcoupling and asymmetry. Vision Research 24, 549-564.

HAMMOND, P. (1974). Cat retinal ganglion cells: size and shape ofreceptive field centres. Journal of Physiology 242, 99-118.

HOCHSTEIN, S. & SHAPLEY, R.M. (1976a). Quantitative analysis of reti-nal ganglion cell classifications. Journal of Physiology 262, 237-264.

HOCHSTEIN, S. & SHAPLEY, R.M. (19766). Linear and nonlinear spa-tial subunits in Y cat retinal ganglion cells. Journal of Physiology262, 265-284.

ISHIDA, A.T., STELL, W.K. & LIGHTFOOT, D.O. (1980). Rod and coneinputs to bipolar cells in goldfish retina. Journal of ComparativeNeurology 191, 315-335.

JACOBSON, M. & GAZE, R.M. (1964). Types of visual response from sin-gle units in the optic tectum and optic nerve of the goldfish. Quar-terly Journal of Experimental Physiology 49, 199-209.

KOCK, J.-H. & REUTER, T. (1978). Retinal ganglion cells in the cruciancarp (Carassius carassius). II. Overlap, shape, and tangential orien-tation of dendritic trees. Journal of Comparative Neurology 179,549-568.

KUFFLER, S.W. (1953). Discharge patterns and functional organizationof mammalian retina. Journal of Neurophysiology 16, 37-68.

LEVICK, W.R. & THIBOS, L.N. (1982). Analysis of orientation bias incat retina. Journal of Physiology 329, 243-261.

LEVINE, M.W. & ZIMMERMAN, R.P. (1985). Mechanisms contributingto the receptive fields of ganglion cells in the retinae of fish. Inves-tigative Ophthalmology and Visual Science (Suppl.) 26, 263.

MACKINTOSH, R.M., BILOTTA, J. & ABRAMOV, I. (1987). Contributionsof short-wavelength cones to goldfish ganglion cells. Journal ofComparative Physiology A 161, 85-94.

MILKMAN, N., SHAPLEY, R. & SCHICK, G. (1978). A microcomputer-based visual stimulator. Behavior Research Methods and Instrumen-tation 10, 539-545.

REMSLAG, F.C.C. & SCHELLART, N.A.M. (1978). Evoked potentials andspike responses to moving stimuli in the optic tectum of goldfish.Journal of Comparative Physiology A 128, 13-20.

Orientation tuning 13

RODIECK, R.W. (1965). Quantitative analysis of cat retinal ganglion cellresponse to visual stimuli. Vision Research 5, 583-601.

RODIECK, R.W. (1979). Visual pathways. Annual Review of Neurosci-ence 2, 193-225.

RODIECK, R.W. & STONE, J. (1965). Analysis of receptive fields of catretinal ganglion cells. Journal of Neurophysiology 28, 833-849.

SHAPLEY, R.M. & GORDON, J. (1978). The eel retina: ganglion cellclasses and spatial mechanisms. Journal of General Physiology 71,139-155.

SOODAK, R.E. (1986). Two-dimensional modeling of visual receptivefields using Gaussian subunits. Proceedings of the National Acad-emy of Sciences (USA) 83, 9259-9263.

SOODAK, R.E., SHAPLEY, R.M. & KAPLAN, E. (1985). Unusual orien-tation tuning in the LGN and perigeniculate nucleus of the cat.Investigative Ophthalmology and Visual Science (Suppl.) 26, 264.

SOODAK, R.E., SHAPLEY, R.M. & KAPLAN, E. (1987). Linear mechanismof orientation tuning in the retina and lateral geniculate nucleus ofthe cat. Journal of Neurophysiology 58, 267-275.

WARTZOK, D. & MARKS, W.B. (1973). Directionally selective visualunits recorded in optic tectum of the goldfish. Journal of Neu-rophysiology 36, 588-604.

WOLBARSHT, M.L. & WAGNER, H.G. (1963). Glass-insulated platinummicroelectrodes: design and fabrication. In Medical Electronics, ed.BOSTEM, H., pp. 510-515. Liege: University of Liege Press.