Receptive field dimensions of lateral geniculate cells in the common marmoset (Callithrix jacchus)

Vision Res., Vol. 37, No. 16, pp. 2171-2181, 1997

Pergamon

PII: S0042-6989(97)00041-2

© 1997 Elsevier Science Ltd. All rights reserved Printed in Great Britain 0042-6989/97 $17.00 + 0.00

Receptive Field Dimensions of Lateral Geniculate Cells in the Common Marmoset (Callithrix

jacchus) JAN KREMERS,*'~ STEFAN WEISS* Received 28 March 1996; in revised form 2 August 1996; in final forrn 28 January 1997

We measured the spatial receptive field dimensions of cells in the lateral geniculate nucleus (LGN) of the common marmoset (Callithrixjacchus) using a bipartite field stimulus in which the two halves of the field were modulated identically but in counterphase. Horizontal and vertical edges between the two fields were positioned at different locations in the receptive field. By assuming that centers and surrounds have gaussian profiles, we were able to obtain a satisfactory mathematical description of the data. Receptive field centers were about a factor 1.6 larger than those of macaque LGN cells, in accordance with the smaller marmoset eye. There was a limited correspondence with dendritic tree dimensions of marmoset retinal ganglion cells. We further found that center and surround gaussians were not always concentric, and that the centers of some cells were elongated. This might allow some direction or orientation biases in LGN cells. © 1997 Elsevier Science Ltd.

Marmoset

Lateral geniculate nucleus

Spatial Receptive field

INTRODUCTION Dimensions of receptive field centers and surrounds in retinal ganglion cells and cells in the lateral geniculate nucleus (LGN) are closely related to spatial processing of visual information in the retina. The size of the receptive field centers is, for instance, inversely correlated to the spatial resolution of the cell (Peichl & W~issle, 1979). Most receptive field measurements in LGN cells and retinal ganglion cells were exhibited on macaques or cats. For an overview of receptive field structure in primates, we refer to a recent review by Lee (1996). To date, only few electrophysiological data are available on receptive field dimensions in New World monkeys such as the common marmoset. Recently, the functional architecture of the visual cortex was studied in marmosets (Sengpiel et al., 1996; Rosa & Schmid, 1995). To obtain more insight in the spatial processing in the retina and the LGN of New World monkeys, we measured spatial receptive field properties of marmoset LGN cells. Knowledge on receptive field structures of LGN cells in New World monkeys is of interest because it addresses possible differences in the neuronal organization of the retina between Old- and New World monkeys. A cause for neuronal differences might be the genetically based *Department of Experimental Ophthalmology, University of T~bingen Eye Hospital, Rtintgenweg 11, 72076 Tiibingen, Germany. "~To whom all correspondence should be addressed [Tel +7071 • 2985031; Fax +7071 295777; Email jan.kremers@uni-tuebin gen.de].

Bipartitefield stimulus

polymorphism in color vision of most New World monkeys. On the X-chromosome, they are presumed to have only one gene, with three different alleles, which expresses a visual pigment absorbing in the middle and long wavelength (L/M) range. As a result, all males and the homozygotic females are dichromats, whereas the heterozygotic females have trichromatic color vision (Tovee et al., 1992; Tovee, 1994; Travis et al., 1988; Jacobs et al., 1987; Jacobs et al., 1993; Mollon et al., 1984). Further, unexpectedly strong rod input to marmoset LGN cells has been observed (Yeh et al., 1995; Weiss et al., 1995; Kremers et al., 1997). Color vision is not polymorphic in Old World monkeys (Tovee, 1994). Shapley & Perry (1986) propose that the small receptive fields of macaque P-cells are related to their chromatic opponent response properties which are required for color vision. If this is true, then one might expect differences in receptive field dimensions between Old- and New World monkeys. However, Wilder et al. (1996) found that the spatial anatomical organization of retinal ganglion cells and photoreceptors is very similar in marmosets and macaques, concluding that the functional properties will be similar as well. They argue that spatial rather than color vision determines the spatial organization. In that case we expect receptive fields sizes which are similar to macaques. Direct measurements of spatial receptive field dimensions will help to resolve the controversy. The comparison between the spatial properties of Oldand New World monkeys will indicate to what extent knowledge about the retina of New World monkeys can

2171

2172

J. KREMERS and S. WEISS

be used for understanding the human visual system. Especially because of the large number of dichromats in platyrrhines, they might be good animal models for the human dichromatic system. We further wanted to know to what extent the receptive fields of primate LGN cells can be considered radially symmetric. Deviation from that symmetry might cause some orientation selectivity in the cells. Finally, the isolation of rod and cone activities provide information about how the rod and cone signals are processed in the retina. Parts of the results have been previously presented in abstract form (Kremers et al., 1995b,c).

METHODS

Animal preparation The results are based on measurements from 10 adult (six males and four females) dichromatic common marmosets (Callithrix jacchus, 300-400 g). The animal experiments were conducted in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC). The animals were initially anasthetized by an intramuscular injection of 1530mg/kg ketamine hydrochloride (Ketanest ®) and 3.5mg/kg Xylazin hydrochloride with 1.5mg/kg Methyl-4-hydroxybenzoate (0.15 ml/kg Rompun ® 2% solution). Additional doses of ketamine hydrochloride were administered if necessary. After tracheotomy the animals were ventilated through a tracheal canula with 70%/30% N20/O2 with 0.2-0.8% Enflurane (Ethrane®; 0.4-0.8% during surgery; 0.2-0.4% during measurements). Eye movements were suppressed by intravenous administration of Gallamin triethiodide (Flaxedil®; 5 mg/kg/hr) dissolved in Ringer, together with glucose and Solu Decortin ® (infusion rate: 0.6 ml/hr). A rectal probe connected to a thermal blanket was used to maintain a rectal temperature of 37.2°C. EEG and EKG were continuously monitored to check depth of anesthesia. Pupils were dilated with atropine sulfate (1%) and neosynephrine (5%). Contact lenses (radius of curvature: 3.5 nun; diameter: 5 mm), with appropriate correction (determined with a slit skiascope) to focus the eyes on the stimulator at 1.14 m distance, protected the eye against desiccation. A 2 m m diameter artificial pupil was positioned in front of the animals' eyes. A craniotomy was made and tungsten in glass microelectrodes were lowered into the lateral geniculate nucleus for extracellular recordings. The stereotaxic location of the craniotomy was based on an atlas of the marmoset brain (Stephan et al., 1980). Occasionally, small lesions were made to confirm the recording site after histological processing of the brains. An experiment typically lasted between 24 and 60 hr. At the end of the experiments the animals were euthanized with an overdose of sodium pentobarbital (Nembutal®), blood samples were taken for genetic analysis to determine which L/M photopigment was expected to be present

(Hunt et al., 1993; Williams et al., 1992), and after perfusion with paraformaldehyde the brain was removed for histological processing.

Visual stimulation The stimuli were generated on a BARCO Calibrator monitor (100Hz refresh rate) driven by a VSG 2/2 graphics card (Cambridge Research Systems) at a distance of 114 cm. The spectral output of the guns was measured with a Spectrascan spectroradiometer (Photo Research). The luminances were recalibrated at regular intervals using the internal calibration of the BARCO monitor and checked with an UDT luminance detector and a International Light ILl700 Radiometer. The VSG card automatically performed a gamma correction on the calibrations and gave the appropriate output to control the monitor guns. The stimuli consisted of a bipartite field with the hemifields modulated sinusoidally at 4 Hz in counterphase but otherwise identically. The common border of the two hemi-fields was positioned at different locations in the receptive field of the cells. Horizontal and vertical edges were used, to study receptive field dimensions in two directions (orthogonal to the edge). In each direction a full measurement consisted of two full-field stimulations and 29 bipartite field stimulations with different edge positions. The total range of different edge positions scanned was either 96 or 192 rain arc. The maximal spatial resolution of edge position was one pixel (0.4 mm or 72 sec of arc at 114 cm distance). One stimulus condition with the edge close to the receptive field's center was measured at the beginning and the middle of the run to check for eye movements. Since near the receptive field center a change in edge location of only 1 pixel resulted in a marked change in cell response, we are confident that small eye movements were detected. We further measured the full field stimulations at the middle and the end of a run. Measurements in which small eye movements, or large response changes to full field stimulation occurred (more than 50%), were discarded. All receptive fields were measured with a luminance modulation on both hemi-fields with 75% Michelson contrast. We used this high contrast in order to obtain reliable measurements at all edge positions, i.e., also for edge positions near the center of the receptive field where response amplitudes are relatively small. The choice of contrast, however, might influence the results (see Discussion). The luminance modulation was obtained by in-phase modulation of the red and the green monitor phosphors each having a time-averaged luminance of 5 cd/m 2. Taking into account the smaller eye of the marmoset we calculated that the retinal illumination was equivalent to about 4.9 times the human retinal illumination with the same stimuli. Therefore, the retinal illuminance was equivalent to about 150 human trolands. In a different study we found that cells often received a combined rod and cone input at the used luminances (Kremers et al., 1997), although the cone input will

RECEPTIVE FIELDS IN THE MARMOSET LGN

dominate. We therefore employed other types of modulation to obtain selectively rod or cone input. To obtain selective cone stimulation, we used a mean luminance of 40cd/m 2 with an additional 15 cd]m 2 unmodulated background from the blue phosphor. With this stimulus, we are confident that rod input was insignificant (Weiss et al., 1995). On some cells we used a 40 cd/m 2 mean luminance with a modulation of the red and green phosphors which was calculated to be a silent substitution condition for the rods (12% green and 43% counterphase red). The signal to noise ratio of responses to this latter stimulus, however, was slightly worse. The spatial extent of rod input was measured as follows: we determined electrophysiologically which L/M-photopigment was present (Weiss et al., 1995). (Briefly, the red and the green guns of the monitor were modulated in counterphase. The Michelson contrast of the red gun was kept constant, whereas green contrast was varied. The responses of the cells were minimally at green contrasts where the present photopigment was silently substituted. The condition at which the minimum appears was characteristic for the present photopigment type. After the experiment, the genetic analysis confirmed independently our electrophysiological determination.) We then used a modulation which was silent for that cone at a 2 cd/m 2 mean luminance. To increase rod signals we used counterphase modulation of the red and blue phosphors. We could use the blue gun, since the cells which were included in this study did not show any signs of S-cone input. The silent substitution conditions were for the 543 nm pigment a 100% red and 32% blue modulation, for the 556 nm pigment a 100% red and 53% blue modulation, for the 563 nm pigment a 100% red and 67% blue modulation. We cannot exclude the possibility of residual cone input under these conditions, but the contribution of the rod signal will have been significantly increased relative to the luminance modulation condition. We also employed moving grating stimuli on some cells, to compare the results with those obtained with the bipartite field stimuli and to study the influence of contrast on the receptive field dimensions. Since moving gratings have been used more often in other studies, these measurements will facilitate comparison with literature data. We used horizontal gratings which drifted in the vertical direction. Between 6 and 13 different spatial frequencies were used. The drift velocity was inversely related to spatial frequency, so that the temporal frequency was 4 Hz for all spatial frequencies. At each spatial frequency the responses of the cells were measured at 917 different contrasts between 0 and 100%.

Recording procedure Spike occurrences were recorded with a resolution of 0.5 msec and stored on a CED 1401 on-line computer (Cambridge Electronic Design Ltd.). In order to avoid onset artifacts, the CED did not start to acquire data before at least one period of the stimulus was presented. These recordings were sent to the host computer, stored and PSTHs were created. The cell responses were

2173

measured for 6 sec at each position of the edge. A pulse of the VSG graphics card was used to trigger the CED 1401 computer, enabling the synchronizing between stimulus and spike recordings. RESULTS

Bipartite field data We measured responses of 45 P- and 28 M-cells. Cells with S-cone input were encountered very rarely and were not measured in this way. Figure 1 displays original responses of a P-cell to bipartite field stimuli with a horizontal and a vertical edge each at three different locations. With the edge centered on the receptive field the cell shows a small frequency doubled response. The nonlinearity index of this cell (as defined below) was in the normal range for P-cells. This cell clearly responds when the edge is located 8-10 min of arc away from the center. With further increasing distance between edge and receptive field center, the responses decrease, eventually reaching full field response when the edge is outside the receptive field. The responses were Fourier analysed and amplitudes and phases of the first and second harmonic components were plotted against edge position. An example is shown in Fig. 2. The first harmonic has a very sharp minimum at a certain edge location. A displacement of the edge location by 1.2 min of arc causes a significant response increase. The responses reach a maximum with the edge

2°°/ 240 rain of a r c - - ~ - - @ ---~ . . . . . . . . . .

0m,

,oo,

r

~ 0 0 ~

.... y\

[

,0011

1

'

100 ~

100

0

100

0 52.8 rainof arc

0 8.4 rninof arc

0 rainof are

FIGURE 1. Responses of a parvocellular cell to the bipartite field stimulus. The two hemi-fields of the stimulus were modulated (4 Hz, 75% contrast, luminance modulation) in counterphase. The responses (peristimulus time histograms with 10 msec binwidth; ordinate is spike rate in spike/sec) are shown for six different locations of the border between the hemi-fields. Three responses are displayed for horizontal edges, and three for vertical edges. When the edge is located on the receptive field center, then only a small frequency doubled response can be observed. The cell responds vigorously when the edge is located on the border of the center. Responses decrease for locations further away from the center and approach the response to full field modulation when the edge is located outside the receptive field.

2174

J. KREMERS and S. WEISS

vertical edges

horizontal edges A

"~ 200

•

200

E ~- 150

C

150

c

o 100 ("1

100

50

50

1st Harmonic

o 0

I

I

-60 -40 -20

I

2nd Harmonic

tl

0

II I

20

•

0

40

60

-60 -40 -20

180

0

I

I

20

40

180

B

60

D /

t~ t-

0 (1) cO (1)

rr

-180

I

I

-60 -40 -20

I

I

I

0

20

40

-180 60

I

I

-60 -40 -20

X-Location (min of arc)

I

I

I

0

20

40

- -

1st Harmonic

----

2nd Harmonic

60

Y-Location (min of arc)

FIGURE 2. First and second harmonic component amplitudes and phases of the response of a parvocellular cell as a function of edge location. (A) and (B) display the response amplitude and phases, respectively for vertically oriented edges. (C) and (D) show the results for horizontal edges. The horizontal lines in the amplitude plots indicate the cells response to full field modulation measured at different instances. The first harmonic amplitudes go through a sharp minimum, at which the response phase shifts about 180 deg. Also, the second harmonic component is minimal at this location, indicating the spatial linearity of the cell. The drawn curves are the results of a fit with the mathematical function described in the text. The fitting parameters for the vertical edges were: Ac: 184 imp/sec; As: 145 imp/sec; ac: 2.4 min of arc; ors: 19.2 min of arc; #c: - 1.2 rain of arc; #s: 4.8 min of arc. The fitting parameters for the horizontal edges were: A¢: 226 imp/sec; As: 184 imp/sec; a¢: 8.4 min of arc; as: 18 min of arc; #c: 4.8 min of arc; #s: 6 min of arc. Note the asymmetric profile for the horizontal edges (C) which can be modeled by the different midpoints (p) of the center and the surround. Note further that the minimum for the horizontal edges is narrower than for the vertical edges. This can be modeled by a larger center radius for the vertical edges (ac = 8.4 min of arc) than for the horizontal edges (ac = 2.4 min of arc).

is about 15-20 min of arc away from the center, and then decrease again for edge locations further away from the center. The horizontal lines indicate the full field response. The responses with the edge in the outermost positions were still larger than those to the full field stimulation, indicating that the size of the cell's surround is larger than the measuring range. The first harmonic phase remains constant with the exception of a sudden 180 deg change at the edge position where the response is minimal. The second harmonic component is for all edge positions smaller than the first harmonic component with about a constant ratio, indicating that the second

Rc(x) = ac × abs

{I

in macaques and cats at 4 Hz (Smith et al., 1992; Dawis et al., 1984). We made use of the fact that the line spread function along each direction of an elliptical twodimensional gaussian is itself again a gaussian (Soodak, 1986). As with the data for macaque retinal ganglion cells (Kremers et al., 1995a) the center's response to one hemifield was modeled by the spatial integral of the line spread function over the extent of the hemi-field. Similarly, the response to the other hemi-field is an integral, over the rest of the line spread function. Since both hemi-fields are modulated in counterphase we subtracted both integrals, assuming spatial linearity:

Guc,~c(#) " d# -

--0 (3

harmonic component is probably involved in response shaping rather than spatial nonlinearities. Spatially, more nonlinear cells had second harmonics components which exceeded the first harmonics when the edge was located on the receptive field center. To describe the first harmonic amplitude data, we assumed that centers and surrounds had gaussian responsivity profiles, and that center and surround provided opponent input to the cell, which is the case

Xx

Guc,oc(#) " d#

/

(1)

where: Guc,~c(#) = 1 / ~ × e x p { - ( / ~ - #c)2/2a~2} is the ganssian line spread function of the center. Equation (1) can be reduced to:

Rc(x)=Acxabs{erf(~)}

(2)

Rc (x) is the center response depending on edge location (x); ac, Ac and #c are the standard deviation, the strength and the position of the center, respectively, and erfis the

RECEPTIVE FIELDS IN THE MARMOSET LGN 25

o

M-cells vertical direction

o

M-cells horizontal direction

•

P-cells vertical direction

•

P-cells horizontal direction

P-Cells ]

•

20

['-

24

o

/

o

i

c 0

.~ 20

._N 10

c~ o

o

I"

•

•

C

•

•

"~ 10-

cl

•

~ 12"0

o

oo

•

•

e ~

...~.9 ....

o o

......

Il

,'J

o

....... • ....

I

• /,', o

•

I

•

,

/

f

• ~o I

•

~t •

•

•

i li li

i i"

j vi'/ •

•

,~-•

I I

m/

•

0- ' / J

,

"1-

o

I/

I

/

•

/~ •

~ 22 c

/, •

////

"O

o

////~

/

c 0

26

~ la

I II1"

•

iii

28

o

2175

i i

/.~.-

i II

/I //

-...:-

i iil

• o

~

8-

o

i

w

I

I

I

10

15

20

25

Vertical direction

Hill i

l i l l

no

2

I 5

•

I

I

I

I

I

I

5

10

15

20

25

30

35

Retinal eccentricity (deg) FIGURE 3. Center radii of LGN cells as a function of retinal eccentricity. The center radii vary considerably at each retinal eccentricity. Further, the data for P- and M-cells overlap. The linear regression through the data points show that there is no difference between centers of M- and P-cells at low retinal eccentricity. At larger eccentricities P-cell centers are slightly smaller than M-cell centers. Data for a given cell appear twice in the plot when the radii were measured for both horizontal and vertical edges.

expression for the error function. Similarly the surround response can be calculated.

Rs(x) = As × abs{erf(\ x~rsX/~/} - #s~

(3)

The resultant center and surround responses were subtracted from each other to get the cell response (Rcell(X) = Rc(x ) - R s ( x ) ) . W i t h eccentric centers and surrounds, there might be some positions of the edge where the surround might reinforce the center. We, however, neglected this effect, since it occurred near the center of receptive field where the responses were very small. The resultant function was fitted through the data points using least sum of squares criteria. In the fitting procedure there were six free parameters: the amplitude (A), standard deviation (a) and the positions (/~) of the center and the surround gaussians. The drawn curve in Fig. 2 is a fit with this function, showing that it was possible to get good descriptions of the data. The resultant values of the free parameters are given in the legends. Because of the large number of free parameters it was not possible to find the best fit by scanning all possible combinations of parameters. We used two different algorithms and different starting parameters to obtain information on the reliability of the fitting parameters. The goodness of fit (expressed as the sum of distances between data and fit) was very similar for both algorithms. It proved that the resultant fits were relatively robust for standard deviation (SD) (ac) and position (/~¢) of the

FIGURE 4. Measured center radii in horizontal and vertical directions plotted against each other. Cells with circular centers would yield data points along the diagonal. The dashed lines indicate tolerance limits: data points outside the area between the dashed lines come from cells with elongated centers. The fraction of cells with elongated centers is presumably larger than the plot suggests because cells with the longer axis of the elliptical center not exactly along the horizontal or vertical axis might still give points lying between the tolerance limits.

center gaussian. The amplitude of the center (Ac) and surround (As) as well as the SD (o-s) and the positions (/xs) of the surround gaussian were less well constrained. The variability in the surround data was partly due to the small range of edge locations, which sometimes insufficiently sampled the surround. Therefore, we were conservative in using these parameters for further consideration. The surround SD (as) and ratios of center and surround amplitudes were only used when the data showed that the majority of the surround was sampled, when the whole range of data points was described satisfactorily, and when the subtraction of center and surround amplitude was within 20% of the response to full field modulation. Figure 3 shows the center radius (ac) plotted against retinal eccentricity. There is a considerable amount of scatter in the data. Since we mainly obtained cells with eccentricities between 5 and 30 deg, we restrict the comparison with other data to this range. It can be seen that there is no clear distinction between center radii of M- and P-cells. In 28 P-cells and 22 M-cells, the center radii were measured in both horizontal and vertical directions. In Fig. 4 the radii in both directions are plotted against each other. From multiple measurements and using different fitting procedures and starting parameters for the fits, we estimated that the maximal variance in center radius was about a factor of 0.4. We additionally assumed an absolute possible error in the center radii of 2.4 min of arc (2 monitor pixels). The dashed lines indicate this variance. Data points outside these limits indicate genuine differences in center radius between both directions. Although most cells have centers with similar radii in both directions, there are some with elongated

2176

J. K R E M E R S and S. WEISS 0

M-cells vertical direction

[]

M-cells horizontal direction

•

P-cells vertical direction

•

P-cells horizontaldirection

120

O c-

[]

100 0

m

80 0 0

"0 "13 c

~

£

60

o

• "o

~

40

o

o ii•

0

it

•

.% •

0

o

Ilia D •

20

•

o~

•

.>_ •0 rr

[]

0[]

•

•

•

I

I

I

I

I

I

5

10

15

20

25

30

35

Retinal eccentricity (deg)

FIGURE 5. Surround radii of P- and M-cells as a function of retinal eccentricity. Surround radii can vary greatly. The surrounds of M-cells can be very large.

centers. The radii in one direction were between approx. 1 and 3 times larger than in the other direction. The proportion of cells with elongated centers might be larger than the plot in Fig. 4 suggests, since cells with obliquely oriented centers might still give data points near the diagonal. The data for surround radii are shown in Fig. 5. We were restrictive in including surround data. The requirement that the surround must have been largely sampled 0 [] • •

25

A

///

E 2o tO

E ~

2

B

iii I I

~ 15

/11 O / /// /

"13 0

E 10

"0

£

.~ s

5

0

0

I

I

I

I

5

10

15

20

m

25

/

////// / / IIII ////~

iiII /•DIll III / O/ /7"

eliiI I/ I

Q

~

III/

c 0

I/II; 0~

10

c 0 o

25

.__q20

,/ • O~ • 11

"0 0

M-cells vertical direction M-cells horizontal direction P-cells vertical direction P-ceils horizontal direction

/

//'/

t'-

by the different edge locations, does introduce some bias towards cells with smaller surrounds. The data indicate surround diameters are variable and can be very large. The ratio of center to surround strength varied between 1 and 3.5 for both P- and M-cells. The values for P-cells are comparable with the center to surround weightings found in macaque retinal ganglion cells belonging to the parvocellular pathway (Smith et al., 1992). We did not find a relationship between center amplitude and radius, nor between surround amplitude and size. This stimulus was suitable to investigate the spatial linearity of cells. We therefore calculated an index analogous to the nonlinearity index introduced by Hochstein & Shapley (1976) for sine wave gratings. We defined the nonlinearity index as the second harmonic component at the edge location where the first harmonic component was minimal (thus at the receptive field's center) divided by the maximal first harmonic component (when the edge is located near the centersurround border). The P-cell responses shown in Fig. 1 displayed a small frequency-doubled response when the edge was centered on the receptive field. The mean nonlinearity index of this cell was 0.085. The nonlinearity indices were 0.0731 _ 0.0424 (mean+SD) for 45 Pcells and 0.1352 +__0.0944 for 26 M-cells. Although there is considerable overlap in nonlinearity index values between cell classes, M-cells are significantly spatially more nonlinear than P-cells (assuming a gauss•an distribution, Student's t-test, P < 0.0003). So far we have considered measurements with a mean luminance of about 150 td. At this intensity level, the cells obtain input from both rods and cones (Weiss et al., 1995). We now present data from experiments in which we isolated the cone or rod input. Figure 6(A) shows the center radii measured with cone isolating stimuli plotted

0

0

5

I

I

I

10

15

20

25

luminance modulation (min of arc)

FIGURE 6. Center radii measured with cone isolating (A) and rod isolating (B) stimuli plotted against the radii measured and with luminance modulation around a m e a n of 20 cd/m 2, where probably predominantly (but not exclusively) cones are involved. The same tolerance limits as for Fig. 4 were used. The data suggest that center sizes are similar for rod- and cone-mediated cell responses.

RECEPTIVE FIELDS IN THE MARMOSET LGN • •

P-CellScontrast gain P-cells 75% contrast

o o

M-cells contrast gain M-cells 75% contrast

20 E

8

LO t~

,~ 15 C3

• •

o o

o o

•

o

10

§ 0~00 0•

1DO £3

g

o

I

I

I

5

10

15

20

Center radius with responses at 10% contrast

FIGURE 7. Center sizes with drifting gratings using different criteria. The center sizes when using contrast gains and responses at 75% contrast are plotted against the center sizes when the responses at 10% contrast are used. The data lie around the diagonal, indicating that contrast has no influence on center size.

against center radii (same cell and same direction) with the usual stimulus. In Fig. 6(B) center radii with rod isolation are plotted against the data with the usual stimulus. We used the same tolerance limits as used in Fig. 4 to indicate significant differences in center sizes. We were not able to detect any systematic size differences. We therefore conclude that center dimensions were relatively similar for rod and cone input.

Moving gratings The bipartite field stimulus previously was not used very often. To enable a direct comparison of marmoset data with those of Old World monkeys we measured the sizes of 10 P-cells and seven M-cells using vertically drifting gratings. The responses (again defined as the first harmonic component of the Fourier analysis) were analysed in two different ways. First, the response amplitudes at a fixed contrast as a function of spatial frequencies were fitted with the following function:

R = C.(KcTr~exp(-(rcfTr)2) - KsTr~exp(-(rsfTr)2)),

(4) where R is the cell's response amplitude, C is contrast, Kc and Ks are peak sensitivities (in imp.sec-l.%contrast-Ldeg -2) and rc and rs are radii of center and surround, respectively. The radii in this expression are the distances from the midpoint to the point where the sensitivity has fallen to 1/e of the maximal value. We divided these radii by ~ to make them directly comparable to values of a obtained from

2177

the bipartite field experiments (a gaussian sensitivity profile with a maximum of, say, A at position 0 looks like: R = A. exp(-x2/2cr2) which drops to a level of A/e when x2/(2a 2) = 1 and therefore x = x/2" ~r). Note that there are only four free parameters in this procedure, instead of six with the bipartite field data. But contrary to the bipartite field data, the fits do not yield any information on the location of the center and surround. It is possible to extract the locations of center and surround from moving gratings, but that would require more than four free parameters (Dawis et al., 1984). We used responses at a relatively low contrast (approx. 10%) which is identical to Croner and Kaplan's (1995) analysis. Additionally, we used the responses at 75% contrast to study the influence of contrast. For the second analysis, we fitted a Michaelis-Menten function (R(C) = R(O) + R,,~x.C/(C + b); R(O) is the response at zero percent contrast; R,,~ is the maximal response of the cell; b is the contrast at half maximal response) through the response vs contrast data at each spatial frequency (Naka & Rushton, 1966). The cells' contrast gain, defined as the initial slope of this function (Kaplan & Shapley, 1986) was plotted as a function of spatial frequency. Through these data a similar function as in Eq. (4) was fitted: C was deleted and the contrast gain (=Rm~/b) was used instead of response R. Theoretically, contrast gains should give similar results as the responses at low contrasts, because the MichaelisMenten function is approximately linear at low contrasts. However, by fitting the Michaelis-Menten function the estimation of the contrast gain makes use of the responses at all contrasts. Figure 7 gives the center radii using the 75% contrast stimuli and the contrast gains as a function of the center radius measured with the 10% contrast stimuli. The center sizes do not differ from each other, since all the data points lie around the diagonal. These data indicate that the stimulus contrast does not influence the center size. Figure 8 displays the center sizes of 10 P-cells and seven M-cells, measured with moving gratings, as a function of retinal eccentricity. The scale of the axes is the same as used in Fig. 3, so that the data are directly comparable. The data confirm our previous result that Pand M-cell centers have very similar sizes at the measured retinal eccentricities. The regression line for the M-cell was rc= 5.0 + 0.38x (where x is retinal eccentricity) with the gratings and rc = 5.6 + 0.34x with the bipartite field stimulus. For P-cells the linear regression gave rc = 6.0 + 0.29x with the drifting gratings and re = 5.7 + 0.21x with the bipartite field stimulus. These regressions should be regarded with some scepticism because of the relatively small cell sample in the grating experiments and because of the data scatter, but they show that the two methods yield very similar results. For seven P-cells and four M-cells, we measured the center size in vertical direction using both the bipartite field data and the drifting gratings. The ratio between both center radii varied between 0.6 and 2.0, which is a

2178

J. KREMERS and S. WEISS 28

"6 "6 E

E E

26

•

P-cells

24

O

M-cells

• •

Q~

'0 •

22

centers surrounds

•

10 eJ"

20

"6 ._E E

18 16 14

O

'% •

E 12

•

E

•..•.

%.

10 8

il

6

g o

~........ 0

i

•

II

4

a-

2 0 0

I

l

I

I

I

I

5

10

15

20

25

30

0.1 35

Retinal eccentricity (deg)

10

100

Radius (rain of arc)

FIGURE 8. Center sizes measured with drifting gratings as a function of retinal eccentricity. Although only a limited amount of cells is measured, there are no differences between the center sizes of P- and M-cells at the given eccentricities. This is in accordance with the data from the bipartite field experiments (Fig. 3).

FIGURE 9. Center (dots) and surround (squares) sensitivities as a function of their radii. Data of P- and M-cells are used. There is a linear relation between both when plotted double logarithmically. The slope of the regression line was -2.37.

further indication that both methods gave similar center sizes. Figure 9 shows the peak center and surround sensitivities Kc and Ks [see Eq. (4)] as a function of their radius, for all cells tested with drifting gratings. Since only the responses at 75% contrast yielded reliable surround data we used these measurements only. The sensitivity clearly decreases with center and surround radius. A linear regression through the (logarithmic) data revealed a slope of -2.27, which is very similar to previously reported slopes (Croner & Kaplan, 1995; Irvin et al., 1993).

Yolton (1970) gave center sizes of squirrel monkey LGN cells. Although they did not distinguish between P- and M-cells, the center sizes they found formed a relatively homogeneous group, suggesting only minor differences. We calculated the expected spatial resolution from the linear regression through the center data obtained with the bipartite field stimulus, assuming that the cells' spatial resolution is reached when 1.8 cycles fit into the receptive field center (Peichl & W~issle, 1979). Because Peichl & W/issle (1979) defined center gaussian radius at the position where responsivity had fallen to 1/e of the maximal value, we multiplied our center radii with x/r2 (see above). Figure 10 shows the expected resolution of marmoset LGN cells together with the data for macaques as described by Crook et al. (1988), whose data agree well with others (Blakemore & Vital-Durand, 1986; Derrington & Lennie, 1984). The resolution of marmoset cells is about a factor of between 1.3 and 2 lower than that of macaque ganglion cells. This difference in resolution can be explained because the marmoset eye is about a factor of 1.6 smaller than the macaque eye, and can anatomically be considered as a scaled-down macaque eye (Troilo et al., 19931. A simple correction for eye size is possible because marmoset and macaque ganglion cells have similar dendritic tree sizes (Goodchild et al., 1996). We will discuss the quantitative comparison between physiological and anatomical data in a later section. de Monasterio & Gouras (1975), Croner & Kaplan (1995) and Irvin et al. (1993) found that M-cell centers are larger than P-cell centers at all eccentricities. Direct comparison of our and Croner and Kaplan's data, after normalizing for eye size, shows that they find smaller Pcell centers than we do around 5 deg eccentricity. Further, we find smaller M-cell center sizes around 25 deg eccentricity. The results of our measurements with drifting gratings exclude the possibility that stimulus

DISCUSSION

The objectives of our experiments were the comparison between receptive fields of Old and New World monkeys, the radial symmetry of receptive field in primate LGN cells, and the spatial extent of rod and cone input. We will discuss now to what extent the results of our measurements contribute to these objectives. Comparison of receptive field sizes We found that receptive field centers of P- and M-cells in the marmoset LGN had similar radii. Only at higher eccentricities, is there a tendency for M-cells to have larger centers. Further, the radii increased with increasing retinal eccentricity. These measurements were performed with two different stimulus types (counterphase modulating bipartite fields and moving gratings), the results of which agreed with each other. In accordance with our data, others (Derrington & Lennie, 1984; Blakemore & Vital-Durand, 1986; Crook et al., 1988; Kremers et al., 1995a; Kaplan & Shapley, 1982; Lee et al., 1993) have found that P- and M-cells have similar center sizes or spatial resolutions. Jacobs &

RECEPTIVE FIELDS IN THE MARMOSET LGN

- -

\

10

developed as an adaptation to color vision. Color vision seems to have added surprisingly few additional constraints to the neural wiring in the retina. Recently, Wilder et al. (1996) came to the same conclusion on the basis of the topography of retinal ganglion cells and photoreceptors in the marmoset retina.

\~OOOke~l.,1988) ----

\

M-cells P-cells

2179

"6

Comparison between physiology and anatomy e,

.2 O

11¢

10

20

30

Retinal eccentricity (deg) FIGURE 10. Calculated spatial resolution of marmoset LGN cells (calculated from the regression line through the center data as shown in Fig. 3) as a function of retinal eccentricity. Also shown is the measured spatial resolution in macaque LGN cells (Crook et al., 1988). Expected spatial resolution of marmoset LGN cells are given within the eccentricity range where we encountered receptive fields. The spatial resolution of marmoset cells is between 1.3 and 2 times smaller. This factor seems to be relatively independent of retinal eccentricity.

contrast or stimulus type has major influences on the center size. Because of the large variability of center sizes at certain retinal eccentricities, a large cell sample is needed to study the influence of all variables more closely. Other spatial properties of marmoset LGN cells are also observed in other primates: as in macaque LGN (Kaplan & Shapley, 1982), marmoset M-cells are spatially more nonlinear than P-cells. Further, in the drifting grating experiments we found that the peak sensitivity term (K) as a function of size in a double logarithmic plot gave a linear relationship with a slope of about - 2 . Thus, the peak sensitivity is about proportional to size -2 This was also found by Croner & Kaplan (1995) and Irvin et al. (1993). In the bipartite field experiments we found that the amplitude does not vary with center size. A size-independent amplitude term is in agreement with a size-dependent sensitivity term, since sensitivity K is defined as the sensitivity per unit area, whereas the amplitude term A is a response term for the whole receptive field. We conclude that marmoset LGN cells have spatial properties that are very similar to those of Old World monkeys. The main difference in center size and thus in spatial resolution is caused by a scaling down of the marmoset eye. This conclusion implies that the retinal wiring is probably the same in Old and New World monkeys, and must have evolutionary largely originated in the period before the two groups separated (about 40 to 55 million years ago). Since many New World monkeys are dichromats and the common ancestor of the two groups, therefore, probably was at least an incomplete trichromat, many retinal structures probably have not

We compared our receptive center radii with anatomical data on primate ganglion cells. As mentioned above, Goodchild et al. (1996) found that the dendritic tree diameters of marmoset and macaque retinal ganglion cells are very similar, which therefore explains the differences in angular ganglion cell diameters and in receptive field dimensions. The data also allow a direct comparison between dendritic tree (Goodchild et al., 1996) and receptive field center radii (dendritic tree diameters were kindly provided by Dr P. R. Martin). Figure 11(A) displays Mcell receptive field center radii and A-cell (homologous with parasol cells) dendritic tree radii (defined as half the dendritic tree diameter divided by x/2 to account for our use of standard deviation as a definition of receptive field radius, rather than the position where the center gaussian has decreased to 1/e times the maximum) as a function of retinal eccentricity. With few exceptions the M-cells' centers are smaller than the parasol cells' dendritic trees. For macaques, Lee (1996) also came to the conclusion that the larger receptive field centers of foveal M-cells were within the range of the measured dendritic tree diameters of foveal parasol cells (Grtinert et al., 1993). Figure 11(B) shows the receptive field radii of P-cells together with the dendritic tree radii of B-cells, which are probably homologous with macaque midget cells (Goodchild et al., 1996). Only the center radii of P-cells at eccentricities beyond about 20 deg and the smallest centers of the more central P-cells lie within the range of the dendritic tree sizes. We conclude that, similarly to macaques (Lee, 1996), there is a limited correspondence between receptive field center sizes of LGN cells and retinal ganglion cell dendritic tree dimensions. Receptive field anisotropies The measurements reveal two different kinds of receptive field anisotropies. The asymmetries in the response profiles can be explained by non-concentric centers and surrounds. Secondly, centers do not have the same diameters in all directions. It is unlikely that astigmatic optical distortions have influenced the data because we found elongated and non-elongated centers in cells with input from the same eye and at similar retinal locations. We conclude therefore, that a proportion of the cells have an elongated receptive field center. This proportion is probably larger than Fig. 4 suggests since cells with oblique elliptical centers (say along a diagonal) will give data points within the tolerance limits. Hammond (1974) found that cat ganglion cells can have elongated centers. Subsequent studies indeed

2180

J. KREMERS and S. WEISS o

Parasoll cells dendritic tree radius

o

Midget cells dendritic tree radius

•

M-cells' centers vertical direction

•

P-cells vertical direction

•

M-cells' centers horizontal direction

•

P-cells horizontal direction

30

30

A

B

25

"6

o8

o

°°oo"~ o o

0

20

o

~, •

•

"o0O".

t--

o °8°g° • o oo

o

o

20

o

o°°° o'.~°o8 o oo800 o

ffl

25

o

oo~o

o•o

o-

to 0

•

::

o

•

--

iI~o

.

:15 10 rr

•

"--

O o ~ °on

•

.

•

•

.

.

' ".'"

.

.

o

o '~ ~ , o ~ - ~

9 , 0 %O~o%O~ N °o

•

o i*• 0

I

I

I

I

I

5

10

15

20

25

30

Retinal eccentricity(deg)

•I,NI

,l~l~ep8

"~oOV

•

•

o

.o

- "U-o

o

I

I

I

I

I

I

5

10

15

20

25

30

35

Retinal eccentricity(deg)

FIGURE 11. Measured center radii of marmoset LGN cells and dendritic tree radii of their retinal afferents (Goodchild et al., 1996). (A) M-cells and parasol ganglion cells; (B) P-cells and midget ganglion cells.

revealed that retinal ganglion cells and LGN cells in the cat are direction biased (Creutzfeldt & Nothdurft, 1978; Vidyasagar & Urbas, 1982; Soodak et al., 1987; Levick & Thibos, 1980, 1982). We show that marmoset LGN cells have elongated receptive field centers similar to cat LGN cells. An orientation bias of primate LGN cells might be expected. Lee et al. (1979) found an indication for orientation bias in primates. Recently, Smith III et al. (1990) confirmed this. Their more quantitative study showed that the orientation bias indeed might be related to receptive field dimensions. Further experiments should be performed to establish the correlation between receptive field anisotropies and orientation bias in marmosets. Spatial extension o f rod and cone inputs Since we measured receptive field dimensions in dichromatic animals, the vast majority of cells had input from only one cone type (the L/M cone) and from rods (Yeh et al., 1995; Weiss et al., 1995). By using stimuli which selectively stimulated either rods or L/M-cones we were able to obtain receptive field dimensions for both photoreceptor inputs separately. The centers and the surrounds of cones and rods were always of the same sign, confirming previous results using chromatic flash stimuli (Kremers et al., 1997) that rods and cones had non-opponent inputs. Our data suggest that receptive field center dimensions do not change systematically for rod and for cone inputs. Similar results have been found before in the cat retina (Barlow et al., 1957; EnrothCugell et al., 1977). However, there it was reported that at very low retinal illuminances the surround seems to be very insensitive. The finding that center sizes do not change in a systematic manner with the transition from rod to cone input is surprising considering the presumed convergence of rod signals onto retinal ganglion cells through rod bipolars and AII amacrine cells (W~issle & Boycott,

1991). However, it has also been suggested that at retinal illuminances as used in our experiments, the rod signal enters the cone pathway through electrical coupling of rods and cones (Stockman et al., 1995). In that case, similar receptive field dimensions with rod and conemediated signals can be expected. However, there are no direct data conceming the pathways of rod signals which may falsify or verify this hypothesis.

REFERENCES

Barlow, H. B., Fitzhugh, R. & Kuffler, S. W. (1957). Change of organization in the receptive fields of the cat's retina during dark adaptation. Journal of Physiology (London), 137, 338-354. Blakemore, C. B. & Vital-Durand, F. (1986). Organization and postnatal development of the monkey's lateral geniculate nucleus. Journal of Physiology, 380, 453-491. Creutzfeldt, O. D. & Nothdurft, H. C. (1978). Representation of complex visual stimuli in the brain. Naturwissenschafien, 65, 307318. Croner, L. J. & Kaplan, E. (1995). Receptive fields of P and M ganglion cells across the primate retina. Vision Research, 35, 7-24. Crook, J. M., Lange-Malecki, B., Lee, B. B. & Valberg, A. (1988). Visual resolution of macaque retinal ganglion cells. Journal of Physiology, 396, 205-224. Dawis, S. M., Shapley, R., Kaplan, E. & Tranchina, D. (1984). The receptive field organization of X-cells in the cat: spatiotemporal coupling and asymmetry. Vision Research, 24, 549-564. de Monasterio, F. M. & Gouras, P. (1975). Functional properties of ganglion cells of the rhesus monkey retina. Journal of Physiology, 251, 167-195. Derrington, A. M. & Lennie, P. (1984). Spatial and temporal contrast sensitivities of neurones in lateral geniculate nucleus of macaque. Journal of Physiology, 357, 219-240. Enroth-Cugell, C., Hertz, B. G. & Lennie, P. (1977). Convergenceof rod and cone signals in the cat's retina. Journal of Physiology (London), 269, 299-317. Goodchild, A. K., Ghosh, K. K. & Martin, P. R. (1996). A comparison of photoreceptorspatial density and ganglion cell morphologyin the retina of human, macaquemonkey, cat, and the marmoset Callithrix jacchus. Journal of Comparative Neurology, 366, 55-75. Gr(inert,U., Greferath,U., Boycott,B. B. & Wiissle, H. (1993). Parasol (Pa) ganglion-cells of the primate fovea: immunocytochemical

RECEPTIVE FIELDS IN THE MARMOSET LGN staining with

antibodies against GABAA-receptors.

Vision

Research, 33, 1-14. Hammond, P. (1974). Cat retinal ganglion cells: size and shape of receptive field centres. Journal of Physiology, 242, 99-118. Hochstein, S. & Shapley, R. M. (1976). Quantitative analysis of retinal ganglion cell classifications. Journal of Physiology, 262, 237-264. Hunt, D. M., Williams, A. J., Bowmaker, J. K. & Mollon, J. D. (1993). Structure and evolution of polymorphic photopigment gene of the marmoset. Vision Research, 33, 147-154. Irvin, G. E., Casagrande, V. A. & Norton, T. T. (1993). Center/ surround relationships of magnoceUular, parvocellular, and koniocellular relay cells in primate lateral geniculate nucleus. Visual Neuroscience, 10, 363-373. Jacobs, G. H., Neitz, J. & Crognale, M. (1987). Color vision polymorphism and its photopigment basis in a callitrichid monkey (Saguinus fuscicollis). Vision Research, 27, 2089-2100. Jacobs, G. H., Neitz, J. & Neitz, M. (1993). Genetic basis of polymorphism in the color vision of platyrrhine monkeys. Vision Research, 33, 269-274. Jacobs, G. H. & Yolton, R. L. (1970). Center-surround balance in receptive fields of cells in the lateral geniculate nucleus. Vision Research, 10, 1127-1144. Kaplan, E. & Shapley, R. M. (1982). X and Y cells in the lateral geniculate nucleus of macaque monkeys. Journal of Physiology, 330, 125-143. Kaplan, E. & Shapley, R. M. (1986). The primate retina contains two types of ganglion cells with high and low contrast sensitivity. Proceedings of the National Academy of Sciences USA, 83, 27552757. Kremers, J., Lee, B. B., and Yeh, T. (1995a). Receptive field dimensions of macaque retinal ganglion cells. In Drum, B. (Ed.), Colour vision defiencies Xll (pp. 399-405). Dordrecht, Boston, London: Kluwer. Kremers, J., Weiss, S. & Zrenner, E. (1995b) Spatial receptive field organisation of lateral geniculate cells of the common marmoset (abstract). Investigative Ophthalmology and Visual Science, Suppl., 36, 691. Kremers, J., Weiss, S. & Zrenner, E. (1995c) Are primate lateral geniculate cells orientation and direction sensitive? (abstract).

Perception, 24, 5-6. Kremers, J., Weiss, S., Zrenner, E., and Maurer, J. (1997). Spectral responsivity of lateral geniculate cells in the dichromatic common marmoset (Callithrix jacchus). In Drum, B. (Ed.), Colour vision deficiencies XIII (pp. 87-97). Dordrecht, Boston, London: Kluwer. Lee, B. B. (1996). Receptive field structure in the primate retina. Vision Research, 36, 631-644. Lee, B. B., Creutzfeldt, O. D. & Elepfandt, A. (1979). The responses of magno- and parvocellular cells of the monkey's lateral geniculate body to moving stimuli. Experimental Brain Research, 35, 547-557. Lee, B. B., Wehrhahn, C., Westheimer, G. & Kremers, J. (1993). Macaque ganglion cell responses to stimuli that elicit hyperacuity in man: detection of small displacements. Journal of Neuroscience, 13, 1001-1009. Levick, W. R. & Thibos, L. N. (1980). Orientation bias of cat retinal ganglion cells. Nature, 286, 389-390. Levick, W. R. & Thibos, L. N. (1982). Analysis of orientation bias in cat retina. Journal of Physiology, 329, 243-261. Mollon, J. D., Bowmaker, J. K. & Jacobs, G. H. (1984). Variations of colour vision in a new world primate can be explained by polymorphism of retinal photopigments. Procedings of the Royal Society B, 222, 373-399. Naka, K. I. & Rushton, W. A. (1966). S-potentials from colour units in the retina of fish (Cyprinidae). Journal of Physiology (London), 185, 536-555. Peichl, L. & W~issle, H. (1979). Size, scatter and coverage of ganglion cell receptive field centres in the cat retina. Journal of Physiology, 291, 117-141. Rosa, M. G. P. & Schmid, L. M. (1995). Visual areas in the dorsal and medial extrastriate cortices of the marmoset. Journal of Comparative Neurology, 359, 272-299.

2181

Sengpiel, F., Troilo, D., Kind, P. C., Graham, B. & Blakemore, C. (1996). Functional architecture of area 17 in normal and monocularly deprived marmosets (CaUithrix jacchus). Visual Neuroscience, 13, 145-160. Shapley, R. M. & Perry, V. H. (1986), Cat and monkey retinal ganglion cells and their visual functional roles. Trends in Neuroscience, 9, 229-235. Smith III, E. L., Chino, Y. M., Ridder III, W. H., Kitagawa, K. & Langston, A. (1990). Orientation bias of neurons in the lateral geniculate nucleus of macaque monkeys. Visual Neuroscience, 5, 525-545. Smith, V. C., Lee, B. B., Pokomy, J., Martin, P. R. & Valberg, A. (1992). Responses of macaque ganglion cells to the relative phase of heterochromatically modulated lights. Journal of Physiology, 458, 191-221. Soodak, R. E. (1986). Two-dimensional modeling of visual receptive fields using gaussian subunits. Proceedings of the National Academy of Sciences USA, 83, 9259-9263. Soodak, R. E., Shapley, R. M. & Kaplan, E. (1987). Linear mechanism of orientation tuning in the retina and lateral geniculate nucleus of the cat. Journal of Neurophysiology, 58, 267-275. Stephan, H., Baron, G., and Schwerdtfeger, W. K. (1980). The brain of the common marmoset ( Callithrix jacchus). Berlin, Heidelberg, New York: Springer. Stockman, A., Sharpe, L. T., Rtither, K. & Nordby, K. (1995). Two signals in the human rod visual system: a model based om electrophysiological data. Visual Neuroscience, 12, 951-970. Tovee, M. J., Bowmaker, J. K. & Mollon, J. D. (1992). The relationship between cone pigments and behavioural sensitivity in a New World monkey (Callithrixjacchusjacchus). Vision Research, 32, 867-878. Tovee, M. J. (1994). The molecular genetics and evolution of primate colour vision. Trends in Neuroscience, 17, 30-37. Travis, D. S., Bowmaker, J. K. & Mollon, J. D. (1988). Polymorphism of visual pigments in a callitrichid monkey. Vision Research, 28, 481-490. Troilo, D., Howland, H. C. & Judge, S. J. (1993). Visual optics and retinal cone topography in the common marmoset (Callithrix jacchus). Vision Research, 33, 1301-1310. Vidyasagar, T. R. & Urbas, J. V. (1982). Orientation sensitivity of cat LGN neurones with and without inputs from visual cortical areas 17 and 18. Experimental Brain Research, 46, 157-169. W/issle, H. & Boycott, B. B. (1991). Functional architecture of the mammalian retina. Physiological Reviews, 71,447-479. Weiss, S., Kremers, J. & Zrenner, E. (1995). Cone and rod input of parvocellular lateral geniculate cells in the common marmoset (Callithrix jacchus) (abstract). Investigative Ophthalmology and Visual Science (Suppl.), 36, 691. Wilder, H. D., Griinert, U., Lee, B. B. & Martin, P. R. (1996). Topography of ganglion cells and photoreceptors in the retina of a New World monkey: the marmoset (Callithrix jacchus). Visual Neuroscience, 13, 335-352. Williams, A. J., Hunt, D. M., Bowmaker, J. K. & Mollon, J. D. (1992). The polymorphic photopigments of the marmoset: spectral tuning and genetic basis. EMBO Journal, 6, 2039-2045. Yeh, T., Lee, B. B., Kremers, J., Cowing, J. A., Hunt, D. M., Martin, P. R. & Troy, J. B. (1995). Visual responses in the lateral geniculate nucleus of dichromatic and trichromatic marmosets (Callithrix jacchus). Journal of Neuroscience, 15, 7892-7904.

Acknowledgements--We are indebted to Professor Zrenner for his continuous support and contributions. We also wish to thank Professor Barry Lee, Dr Paul Martin and Dr Ted Sharpe and the anonymous referees for their comments on the manuscript, Dr Paul Martin for providing the anatomical data, and Dr Tomoaki Usui, Sabine Meierkord and Eva Burkhardt for technical assistance. The authors were supported by DFG Grant Zr 1/9-2.