The Journal of Neuroscience, December 1993, f3(12): 53344355

The Mosaic of Midget Ganglion Cells in the Human Retina

Dennis M. Dacey
Department of Biological Structure, The University of Washington, Seattle, Washington 98195

To study their detailed morphology, ganglion cells of the a pattern of clusters and holes. We hypothesize that midget
human retina were stained by intracellular tracer injection, cell dendritic trees may contact individual axon terminals of
in an in vitro, whole-mount preparation. This report focuses some midget bipolar cells and avoid contacting others, pro-
on the dendritic morphology and mosaic organization of the viding a basis for the formation of cone-specific connections
major, presumed color-opponent, ganglion cell class, the in the IPL.
midget cells. Midget cells in the central retina were recog- [Key words: primate, in vitro preparation, visual acuity,
nized by their extremely small dendritic trees, -5-l 0 pm in color vision, parvocellular pathway, coverage factor]
diameter. Between 2 and 6 mm eccentricity, midget cells
showed a steep, IO-fold increase in dendritic field size, fol- The central retina of the primate contains a specializedvertical
lowed by a more shallow, three- to fourfold increase in the pathway that links a singleconephotoreceptor through a bipolar
retinal periphery, attaining a maximum diameter of -225 Am. cell to a singleganglion cell. The existence of this private line
Despite large local variation in dendritic field size, midget from photoreceptor to ganglioncell wasfirst inferred from Golgi
cells formed one morphologically distinctive class at all ret- preparations (Polyak, 1941; Boycott and Dowling, 1969; Kolb,
inal eccentricities. 1970)and hasrecently beenconfirmed by electron microscopic
Two midget cell types were distinguished by their dendritic reconstruction of singlecells(Kolb and Dekorver, 1991; Calkins
stratification in either the inner or outer portion of the inner et al., 1992). The ganglion cells in this pathway were termed
plexiform layer (IPL), and presumably correspond to ON- and midget ganglion cells becausetheir dendritic trees were found
OFF-center cells respectively. The mosaic organization of to be only about 5-10 Frn in diameter.
the midget cells was examined by intracellularlyfilling neigh- The midget cellsalmost certainly belong to the classof phys-
boring cells in small patches of retina. For both the inner and iologically identified, color-opponent ganglioncells(for reviews,
outer midget populations, adjacent dendritic trees apposed see Shapley and Perry, 1986; Kaplan et al., 1990). Color-op-
one another but did not overlap, establishing a coverage of ponent cells generally have very small receptive fields (De-
no greater than 1. The two mosaics differed in spatial scale, Monasterio and Gouras, 1973, and midget cellshave the small-
however: the outer midget cells showed smaller dendritic est dendritic fields of all primate ganglioncells. The midget cells
fields and higher cell density than the inner midget cells. An provide the major projection to the parvocellular layers of the
outer:inner cell density ratio of 1.7: 1 was found in the retinal dorsal lateral geniculate nucleus (Leventhal et al., 1981; Perry
periphery. An estimate of total midget cell density suggested et al., 1984) where cells with small, color-opponent receptive
that the proportion of midget cells increases from about 45% fields are recorded (e.g., Wiesel and Hubel, 1966; Creutzfeldt et
of total ganglion cell density in the retinal periphery to about al., 1979; Derrington et al., 1984; for a review, seeLennie and
95% in the central retina. Nyquist frequencies calculated D’Zmura, 1988). Lesions confined to the parvocellular layers,
from midget cell spacing closely match a recent measure of or direct destruction of parvocellular-projecting ganglion cells,
human achromatic spatial acuity (Anderson et al., 1991), compromise color vision, peak visual acuity, and the transmis-
from -6” to 55” eccentricity. sion of high-spatial- and low-temporal-frequency signals(Mer-
Outside the central retina, midget cell dendrites arborized igan and Eskin, 1986; Merigan, 1989; Schiller et al., 1990; Mer-
in clusters within the overall dendritic field. With increasing igan et al., 1991; Lynch et al., 1992).The midget cell population
eccentricity, the dendritic clusters increased in number and is thus implicated as a substrateboth for the spatial resolution
remained small (- 1 O-20 pm diameter) relative to the size limit acrossthe visual field and for color vision (Shapley and
of the dendritic field. Because neighboring midget cell den- Perry, 1986; Lennie et al., 1991).
dritic trees do not overlap, the mosaic as a whole showed Despite the presumedimportance of midget ganglion cellsin
primate vision, fundamental questionsabout their anatomical
organization remain [Link] both human and macaque,
Received Nov. 24, 1992; revised June 16, 1993; accepted June 24, 1993.
I am esueciallv erateful to Rim Allen and the staff of the Lions Eve Bank for
it has been concluded that midget ganglion cells form a single
the timely rettie; of donor eyes for this research and for the contributions of population that showsa characteristic increasein dendritic field
Michael Petersen. Christine Curcio, Rate Mulligan, Helen Sherk, Peter Sterling, size with increasingeccentricity (Perry et al., 1984; Rodieck et
and Heinz Wassle offered helpful comments on various drafts of the manuscript.
Dan Possin helped with histological sections and the photographers of the De-
al., 1985; Watanabe and Rodieck, 1989; Dacey and Petersen,
partment ofOphthalmology, University of Washington provided excellent service. 1992). These studies all agree on a further subdivision of the
This work was supported by U.S. Public Health Service Grants EY 06678 and midget cells into two distinct populations that stratify in either
EYOl730 (Vision Research Corel.
Correspondence should be addressed to Dennis Dacey, Ph.D., The University the inner or outer portion of the inner plexiform layer (IPL) and
of Washington, Department of Biological Structure, SM-20, Seattle, WA 98195. presumably correspond, respectively, to ON- and OFF-center
Copyright 0 1993 Society for Neuroscience 0270-6474/93/135334-22$05.00/O types. However, it has also been reported that human midget
 The Journal of Neuroscience, December 1993, f3(12) 5335

cells remain small (and presumably transmit signals from single retina. Anatomical resolving power, calculated from midget cell
cones) far into the retinal periphery and that larger, midget-like spacing,closely matchesa recent measureof human achromatic
cells are present in central retina (Kolb et al., 1992). Since a spatial acuity (Anderson et al., 1991) from -6” to 55” eccen-
number of distinct color-opponent cell types have been distin- tricity. Over this eccentricity range, midget cell dendritic trees
guished physiologically (DeMonasterio and Gouras, 1973, such arborize into multiple, small clustersof high dendritic density,
larger field midget-like cells could, if present, represent an an- suggestingthat peripheral midget cells make selective contact
atomically and functionally distinct population of parvocellular- with a limited submosaicof midget bipolar cells within their
projecting ganglion cells. dendritic fields.
A second question concerns the relationship of midget cell
density to achromatic spatial acuity. On the one hand, there is Materials and Methods
evidence in macaque that, from peripheral to central retina, In vitro isolated retina. The retinal preparation and the intracellular
midget cells make up a constant fraction of the ganglion cell injection technique were developed to study the morphology of neurons
population, about 80% (Perry et al., 1984). This estimate has in the macaque and other mammalian retinas (Vaney, 1985, 1986;
been used in calculations of cell density and of the spatial res- Dacey, 1988; Watanabe and Rodieck, 1989). These methods were ap-
plied with little modification to human retinas that were dissected from
olution limit of the midget cell mosaic as a function of eccen- eyes obtained 90-120 min after death. Eyes (n = 46; age range, 16-82
tricity (Thibos et al., 1987; Lennie et al., 1990; Merigan and years) were accepted from donors to the Lions Eye Bank at the Uni-
Katz, 1990; Wbsle and Boycott, 199 1). From these calculations versity of Washington. Eyes were taken both from cornea donors and
it has been concluded that the intercell spacing of midget cells from donors whose eyes were medically unsuitable for cornea donation.
may set the limit on spatial acuity for both human and macaque No obvious differences in the data were observed over the range of ages
sampled and this variable was not further studied.
in the retinal periphery. On the other hand, the actual dendritic To isolate the retina, the eye was sectioned with a razor blade at the
overlap and intercell spacing of midget cells have never been comeoscleral junction and the vitreous was drained from the eye cup.
directly observed, and there is indirect evidence to suggest that The eye cup was then placed in a continuously oxygenated tissue-culture
the proportion of midget cells changes dramatically as a function medium (Ames, Sigma) at room temperature, and the retina was dis-
sected free of the sclera and choroid and then freed from the eye cup
of eccentricity (Connolly and Van Essen, 1984; Schein and by cutting the optic nerve head. Isolated retinas were left in the oxy-
DeMonasterio, 1987; Dacey and Petersen, 1992). It has also genated medium for a l-2 hr “recovery” period and subsequently placed
been pointed out that certain measures of human spatial reso- flat, photoreceptor side down, in a superfusion chamber on the stage of
lution may be subject to sampling artifacts that could lead to a light microscope. Each retina was maintained in the chamber for -8-
10 hr with no apparent deterioration in morphology. At the end of an
an overestimation ofacuity limits in peripheral vision (Williams experiment the retina was removed from the chamber and fixed for 2-
and Coletta, 1987; Anderson et al., 1991). Thus, a further as- 3 hr in a phosphate-buffered fixative (2% glutaraldehyde, 2% parafor-
sessment of the potential link between the anatomy of midget maldehyde or 4% paraformaldehyde; 0.1 M, pH 7.4). Intracellular in-
cells and acuity requires an understanding of the human midget jection experiments were typically performed consecutively on both the
cell mosaic, together with a measurement of human spatial res- left and right eyes of a pair. Because experiments could be conducted
on only one retina at a time, the retina from the fellow eye was main-
olution in the periphery that is free of sampling artifacts. tained free floating in oxygenated medium during the course of the first
A final question concerns the neural basis for color opponen- experiment.
cy. Both anatomical and physiological evidence has recently Intracellular injection and histology. Retinal cells were stained in vitro
converged to suggest that cone-specific connections arise at the with the fluorescent dye acridine orange and observed under blue ep-
iscopic illumination. Intracellular injections were made into the fluo-
level of the IPL (Reid and Shapley, 1992). Horizontal cells rescing cells under direct microscopic control with beveled, microcap-
contact all cone types nonselectively (Boycott et al., 1987; Was- illary glass electrodes. Electrodes were filled with a solution of Lucifer
sle et al., 1989a), and do not convey chromatically selective vellow (-2%: Aldrich) in 20 mM. DH 7.0 MOPS buffer (Sigma), and
signals (Dacheux and Raviola, 1990). However, midget bipolar either rhodamine-conjugated horseradish peroxidase (HRP) (:4%; Sig-
ma) or Neurobiotin (-4%; Vector Labs). HRP-filled electrodes were
cells contact single cones even in the retinal periphery and could
beveled to a resistance of -40 MR. Neurobiotin-filled electrodes were
thereby transmit a cone-specific signal to the midget ganglion beveled to a resistance of - 80 MQ or left unbeveled (- 120 MB). Lucifer
cells (Boycott and Wassle, 199 1). For example, red ON-center yellow fluorescence in the electrode and the acridine orange fluorescence
cells could directly contact only the submosaic of midget bipolar of the ganglion cells were observed with the same excitation filter (4 1O-
cells that receive input from red cones, while neighboring green 490 nm; barrier filter, 515 nm), permitting direct observation of the
electrode tip as it penetrated a cell. Lucifer yellow was passed into an
ON-center cells would overlap and selectively connect to the impaled cell with l-2 nA negative current for 30 set to confirm a
midget bipolar cells that receive from green cones (Whsle and successful penetration and to identify the cell type being injected or to
Boycott, 1991). This hypothesis suggests that the shape and determine its depth of stratification in the inner plexiform layer. Rho-
overlap of midget cell dendritic trees could provide evidence as damine-conjugated HRP was subsequently passed into the cell with l-
to whether such selective connections exist. 5 nA positive current for l-3 min and was observed directly with a
green filter (excitation filter, 545 nm longpass; barrier filter, 590 nm).
In the present study I have used intracellular injection tech- Neurobiotin was injected with 0.1-0.5 nA positive current for 30-60
niques in an in vitro preparation of the human retina (Dacey set but, as it was unconjugated to a fluorescent tag, could not be observed
and Petersen,1992)to addressthe above [Link] wasfound passing into the cell.
that midget ganglion cells form only two distinct cell mosaics. For the cells that were injected with rhodamine-conjugated HRP, the
These stratify broadly in either the inner or outer portions of HRP was demonstrated using diaminobenzidine (DAB) as the chro-
mogen. Retinas were incubated in the DAB solution (0.1% in 0.1 M
the IPL, and presumably correspond to ON- and OFF-center phosphate buffer, pH 7.4) for 5 min. H,O, (3.0%) was then added to
types, respectively. Midget cell dendritic trees display nonov- DAB solution (30 ~1 in 30 ml) and agitated for 3-4 min. Retinas were
erlappingterritories with a coveragefor both the inner and outer rinsed in buffer, whole-mounted on gelatin-coated slides, and air dried
mosaicsof no more than 1. Estimates of cell density suggest for a few hours. The tissue was then dehydrated in a graded alcohol
series, cleared in xylene, and coverslipped in DPX. When the retina
that the midget cell population does not make up a constant was mounted in this way there was little shrinkage in the plane of the
proportion acrossthe retina, but rather increasesfrom -45% retina (-2OYo). Radial shrinkage was -60%; the inner plexiform layer,
of total ganglioncellsin retinal periphery to - 95% in the central for example, was reduced from -25-30 pm thickness (measurements
5336 Dacey l Human Midget Ganglion Cells

magnification, 480 x or 960 x ) and an outline of the cell body at higher

magnification (1940 x) were made. A measure ofdendritic field diameter
was acquired for the intracellularly filled ganglion cells by tracing a
convex polygon around the perimeter of the traced dendritic tree. The
area of this polygon was then calculated by entering the outline into a
computer via a graphics tablet. Dendritic field diameter was expressed
as the diameter of a circle with the same area as that of the polygon.
Cell body area was similarly expressed in terms of an equivalent di-
ameter.
Measurements of the depth of stratification in the inner plexiform
layer of the midget cell dendritic trees were obtained in two cases for
overlapping pairs of midget cells; one member of each pair branched
in the inner portion of the IPL and the other branched in the outer
portion. A fifth cell, an outer branching midget cell, was also included.
All cells were located in peripheral, temporal retina, the two cell pairs
at 13 and 14 mm eccentricity and the single cell at 15.5 mm eccentricity.
Tissue pieces containing the cells were cut from reacted retinas prior to
mounting and were embedded in plastic resin (Medcast). Vertical sec-
tions through the dendritic fields were cut at 10 pm with a glass knife
and mounted on slides in serial order. A detailed camera lucida tracing
of each section was made (total magnification, 1940 x), that indicated
the outer and inner borders of the inner plexiform layer and all HRP-
Figure 1. Ganglion cells stained with the fluorescent dye acridine labeled dendritic segments. Each of these tracings was then placed on
orange in an in vitro whole-mount preparation of the isolated human a graphics tablet and overlain with a grid of lines orthogonal to the
retina. Plane of focus is on the ganglion cell layer in the retinal periphery. retinal layers and spaced 5 pm apart. The locations of the inner and
The distinctive soma size range of midget ganglion cells permitted them outer borders of the inner plexiform layer, and of the labeled dendrites
to be reliably targeted for intracellular injection. The small arrowheads that intersected each grid line, were then entered into the computer.
indicate the cell bodies of two midget ganglion cells. The large arrow- Measurements were made on every serial section that contained labeled
heads indicate the cell bodies of the other common ganglion cell class, dendrites. The relative depth of stratification (expressed as a percentage
the parasol cells. The larger cell body of each pair tends to belong to of the total thickness of the inner plexiform layer) was then calculated
the cell type with dendrites that stratify in the inner portion of the IPL, at each point of measurement. A histogram displaying all the depth
the presumed ON-center cells. Scale bar, 25 Wm. measurements was then compiled.
The dendritic trees of the midget cells displayed distinct regions of
taken from in vitro, a&dine-stained retinas) to about - lo- 12 pm after high density or “clusters,” surrounded by regions of low density or
dehydration. “holes.” To quantify the tendency toward dendritic clustering, contour
plots of the midget dendritic tree densities were made using commercial
The injected Neurobiotin was revealed by an HRP reaction product software (Transform, Spyglass Inc.) and a Macintosh computer. Camera
using the Vector ABC protocol (Vector, Elite kit). Retinas fixed in 4% lucida tracings of the branching patterns of dendritic trees were con-
paraformaldehyde were first placed in 0.5% Triton X-100 (in 0.1 M verted to PICT files at 150 dni using a Microtech Grav Scale Scanner.
phosphate buffer, pH 7.4) at room temperature for 3 hr, and then in- Using Transform, the PICT image if the dendritic trek was converted
cubated in buffer containing the Vector avidin-biotin-HRP complex for to an array of 8-bit data values. Before generating a contour plot the
3 hr. The tissue was then rinsed in buffer for 1 hr and standard HRP data set was smoothed to eliminate the sharp edges created by individual
histochemistry was performed using DAB as the substrate, as described dendrites. The data smoothing was accomplished by averaging the value
above for the HRP reaction. T&on-treated retinas were easily damaged at a given point in the data set with those ofits eight immediate neighbors
by dehydration and were therefore whole-mounted in an aqueous so- and then assigning the average value to each of the points. The smooth-
lution of glycerol and polyvinyl alcohol (Heimer and Taylor, 1974). The ing procedure was repeated 15 times for each dendritic tree and a gray
mountant partially dries to a water-retaining, rubbery film that provides scale image of the smoothed data set was generated. Finally, contour
reasonable tissue clarity and permits long-term storage and analysis of lines were overlaid on this image at 30% and 70% of maximal image
the retinas with little or no shrinkage. density for each of the dendritic trees. The areas bounded by the 70%
Conversion to angular eccentricity in degrees. In order to convert contour lines were defined as clusters of high dendritic density and the
distance from the fovea in millimeters to degrees of visual angle, the area of each cluster was measured.
nonlinear conversion of Drasdo and Fowler (1974) was used. In their
schematic human eye, the distance-to-angle conversion is 275 pmlde-
gree in the fovea and decreases to - 135 pm/degree at 90” eccentricity.
The nonlinear relationship between retinal distance and visual angle Results
(Fig. 2 in Drasdo and Fowler, 1974) is well fitted by the second-order An in vitro preparation of the human retina
polynomial equation y = 0.1 + 3.4x + 0.035~~ (R = l.O), where y is
eccentricity in degrees and x is eccentricity in millimeters. Conversion The appearance of cells in the isolated human retina, when
to angular eccentricity in the central retina is complicated by the lateral maintained in oxygenated Ames medium at room temperature,
displacement of ganglion cells from the inner segments of the photo- was very similar to that of cells observed in the isolated macaque
receptors from which they receive input. Therefore, ganglion cell ec- monkey retina maintained in the same way (Dacey, 1989a, 1990;
centricity was converted to inner segment eccentricity for all cells within
3 mm of the fovea. The lengths of photoreceptor axons (Henle fibers) Watanabe and Rodieck, 1989). Acridine orange intensely stained
have been measured directly for the macaque retina (Perry and Cowey, the basophilic material of the cell bodies in the ganglion cell
1988; Schein, 1988), but in the human retina it has been predicted that layer against an unstained background (Fig. 1); fluorescent cy-
Henle fibers are somewhat longer than in the macaque (Curcio and toplasm surrounded an unstained nucleus containing an in-
Allen, 1990). Direct measurements of human Henle fiber lengths using tensely stained nucleolus. With some practice it was possible to
the method of Schein (1988) (C. Curcio, personal communication, un-
published observations) bear out this prediction, and were used to gen- distinguish reliably a number of morphologically distinct gan-
erate the following equation to correct for ganglion cell displacement: glion cell types on the basis of soma size, shape, and subtle
y = -0.020 + 0.46x + 0.20~~ (R = l.O), where y is inner segment details of the acridine fluorescence. Dendritic field diameters
eccentricity in millimeters and x is ganglion cell eccentricity in milli- were determined for a sample of 1225 intracellularly filled ret-
meters.
Data analysis. For each retina the location of every injected ganglion inal ganglion cells, including midget, parasol, and many other
cell relative to the fovea1 center was recorded. Each cell was photo- ganglion cell types (e.g., Dacey, 1993). Cells that were included
graphed, and a simple camera lucida tracing of the dendritic tree (total in this database showed uniform staining that ended abruptly
Figure 2. Intracellularly stained human midget ganglion cells. A, Fovea1 cell, 0.8 mm eccentricity, temporal retina. The four panels show successive
planes of focus from cell body and axon (Ieft) to terminal dendritic arbor (right). B, Near retinal periphery, 2.4 mm eccentricity, temporal retina.
C, Near retinal periphery, 3.2 mm eccentricity, upper retina. Two cells are shown; the upper cell is partly out of the plane of focus. D, Mid retinal
periphery, 7.8 mm eccentricity, temporal retina, E, Retinal periphery, 9 mm eccentricity, upper retina. F, Far retinal periphery, 16 mm eccentricity,
nasal retina. Scale bar, 50 pm (for all panels).
5336 Dacey l Human Midget Ganglion Cells

Figure 3. Identification of midget

ganglion cells by dendritic field diam-
eter. Dendritic field diameter of all in-
tracellularly stained ganglion cells is
plotted as a function of retina1 eccen-
tricity. Midget ganglion cells were iden- 0
tified qualitatively by their distinctive
dendritic morphology. When dendritic
field diameter of all cells was plotted,
the midget cells, as expected, formed a
distinct cluster that showed little over-
lap with the nonmidget ganglion cells.
The few midget and nonmidget cells that
did show overlapping dendritic field di-
ameters (in the retina1 periphery) are
actually located in different retina1
quadrants: the largest midget cells in
the retinal periphery reside in the tem-
poral retina1 quadrant whereas the
smallest nonmidget cells reside in the 0 4 8 12 16 20
nasal retinal quadrant. Eccentricity (mm from the fovea)

250
0
0 temporal, upper and lower quadrants A
0 nasal quadrant 0
a OOO
200 - 0 0%

s
$ 150- 0

.
5
-0 . l*
2
B loo- .O
6 .
5
D .
50 -

16 20
Eccentricity (mm from the fovea)

Figure 4. Dendritic field size as a

function of retina1 eccentricity for hu- ‘I;
man midget ganglion cells. A, Scatter- 5
plot of field diameter (in pm) against 3 40
.E
eccentricity (in mm) from the fovea. !z
Nasal cells (solid circles) am smaller than 5
cells from the temporal, upper or lower
retina1 quadrants. Cells from the central .E 30
0
2 mm show a relatively constant den- $
dritic field size of S-10 pm diameter. g
B, Scatterplot of field diameter con- g20
vetted to minutes of arc plotted against
eccentricity converted to degrees of vi- ii
sual angle. See Materials and Methods
for calculations used to make conver- 10
sion from distance in millimeters to de-
grees of visual angle. Equation for the
curve fit to the data is a third-order
0
polynomial: y = 2.1 + 0.058x + 0.022x* 0 10 20 30 40 50 60 70
- 0.00022x3. Eccentricity (degrees)
 The Journal of Neuroscience, December 1993, f3(12) 5339

at the distal ends of the dendrites and appeared completely 1

midget cells - central retina
stained by the HRP or Neurobiotin injection.

Identification of midget ganglion cells

Intracellularly filled midget ganglion cells, as characterized in
the macaque retina (Watanabe and Rodieck, 1989) and ob-
served in Golgi studies of the human retina (Rodieck et al., 0.5 mm
1985; Kolb et al., 1992) were easily recognized as the dominant
cell group in our sample of injected cells (Fig. 2). With practice
it was possible to recognize the cell bodies of midget cells and
fill them at most eccentricities. Injections of midget cells were
technically difficult, however, in the central 24 mm of retina.
In this region the increased thickness of the inner limiting mem-
brane, optic fiber layer, and ganglion cell layer, together with
the small size of the ganglion cell bodies, made a successful 0.8 mm
electrode penetration difficult and the sample of recovered cells
was consequently lower than in the retinal periphery.
The midget cells had small cell bodies, and dendritic trees
that were extremely small compared to the dendritic trees of all
other morphologically characterized ganglion cells. By this qual-
itative feature we identified 539 midget cells from the total
sample of 1225 filled cells. The midget cells appeared to com-
prise a single morphologically distinct group and, as shown in
previous studies of both human and macaque, increased in den- 0.8 mm
dritic field size with retinal eccentricity (Fig. 2). A scatterplot
of dendritic field diameter for all stained ganglion cells as a
function of retinal eccentricity indicated a clear separation be-
tween those cells classified as midgets and all other cells between
0 and 9 mm eccentricity (Fig. 3). Between 9 and 16 mm eccen-
tricity a few of the largest midget cells (dendritic field diameter
of 180-225 pm) overlapped with the smallest nonmidget cells.
However, even for these few cells there was not a true overlap
because the smallest nonmidget cells were from the nasal retina, 0.9 mm
where cell density is high and ganglion cell dendritic trees are
relatively small, and the largest midget cells were from temporal,
upper and lower retina, where cell density is much lower and
dendritic trees are relatively large. Thus, all of the cells that were
initially identified as midget cells by their appearance also formed
a single, distinct group on the basis of dendritic field diameter
in a given retinal quadrant.

Midget dendritic field size range

The relationship between dendritic field size and retinal eccen-
tricity for the midget cells was not linear. In the central retina,
from -0.25 to 2.0 mm (O-6’) eccentricity, the dendritic tree
was composed of a single, primary dendrite that terminated in
a small arbor between 5 and 10 pm in diameter (Figs. 4, 5). 2.0 mm
Although the arbor size was somewhat variable, there appeared
to be no systematic increase in arbor size over this eccentricity
range. Between 2 and 7 mm (6-25’) eccentricity, the dendritic
tree showed a steep 1O-fold increase in dendritic field size (Fig.
4). Over this range dendritic field diameter exceeded soma di-
- Fovea 20 pm
ameter. The dendritic arbor also became more complex, show-
ing second- and third-order branches that often arborized into
multiple clusters within the dendritic field (Figs. 6, 7). The ten- Figure 5. Dendriticmorphologyof midgetganglioncellswithin 2 mm
dency of the midget cell dendritic tree to form regions of high of the fovea. Midget cellsfrom the centralretina give riseto a single
and low density, and its possible significance, are examined primary dendritethat branchesinto a single,compactarbor of fine
[Link] cell shownat 2 mm eccentricity(bot-
below. Between 7 and 20 mm (25-75”) the dendritic tree showed tom) givesriseto two majordendritesthat terminatein a slightlylarger
a more gradual two- to threefold increase in size while the den- dendritic fieldthan [Link]
dritic tree retained its characteristic morphology (Fig. 8). given for eachcell.
A distinct feature of the midget cell population was the large
scatter in field sizes in the retinal periphery. As observed in the
5340 Dacey * Human Midget Ganglion Cells

midget cells - near periphery

3 2 mm, outer

3.8 mm. Inner

Figure 6. Dendritic morphologyof

midgetganglioncellsfrom the nearret-
inal [Link] 2 to 2.5 mm ec-
centricity the dendritic field size of
midgetganglioncellsbeginsto increase.
The largerdendritic arborsdisplaya
nonuniformdistribution of dendritic
branches within the dendriticfield;ter- 5 4 mm, Inner
minal branchesoften arborize into
multipleclusterswithin the dendritic
field. The dendritic clusterssurround
regionsthat showa very low densityof
dendriticbranches. Retinaleccentricity
andlevelof stratification(inneror out-
er) in the IPL aregiven for eachcell.

macaqueretina (Watanabe and Rodieck, 1989), midget cells in branching dendritic [Link] is illustrated for two neighboring
the nasal retinal quadrant in humans showed smaller field di- midget cells in Figure 9. A cell with a smaller, more densely
ametersthan cells in the temporal, upper or lower quadrants. branched tree was filled adjacent to a cell with a larger tree that
However, even excluding cells from the nasal quadrant, there branchedsparselyin multiple, widely [Link]
was a consistently large variation in dendritic field size at a is presented below that this variation in dendritic tree mor-
given eccentricity. Between 5 and 10 mm from the fovea, the phology is an inherent feature of the human midget cell mosaic.
midget cells showeda 2.8fold rangein dendritic field diameter.
Between 10 and 15 mm this range increasedslightly to about Inner and outer midget cell types
threefold. This large variation in midget dendritic field size at The midget ganglion cells could be divided into two cell types
a given retinal location thus appearedto be a relatively constant on the basisof their depth of stratification in the IPL, as pre-
proportion of the mean dendritic field size. This scatter was viously found in macaque (Polyak, 1941; Perry et al., 1984;
larger than that previously observed for either the midget or Watanabe and Rodieck, 1989). One type stratified in the inner
parasol ganglion cells of the macaque retina (Watanabe and half of the IPL and the other type in its outer half, presumably
Rodieck, 1989). corresponding respectively to ON- and OFF-center cell types
The variation in dendritic field size was a consequenceof (Famiglietti and Kolb, 1976). I will refer to thesetwo types as
variation in the dendritic branching patterns of [Link] largest the inner midget and the outer midget cells, respectively.
cellsat a given retinal eccentricity tended to have more sparsely Midget cells at eccentricities beyond 1.5 mm could be iden-
 The Journal of Neuroscience, December 1993, 13(12) 5341

tified aseither inner or outer cellsin whole-mountsby measuring

midget ceils - mid perioherv
I- - I the distance from the cell body to the level of major dendritic
stratification. After dehydration the IPL measuredabout 1O-12
Km in total thickness; cells that were identified asinner midget
f cells tended to be stratified - 2-5 pm from the parent cell body,
and those identified as outer cells were stratified -7-10 pm
inner cell from the parent cell body. Inner midget cells tended to have
larger dendritic trees than outer midgets. A scatterplot of inner
and outer midget dendritic field diameter asa function of retinal
eccentricity showed that the population of inner midgets was
on average about 30% larger in dendritic field diameter than
the outer midgets(Fig. 1OA). The inner midget cellsalsoshowed
a slightly larger soma size than the outer midget cells (mean f
SD, for inner cells, 18.6 f 2.3 pm; for outer cells, 17.4 * 2.3
pm; p < 0.0001, Student’s t test).
Could the size difference between the inner and outer midget
cells be artifactual? It could be argued that the intracellular
injection technique introduced a bias that favored the recovery
of inner midget cells with large dendritic trees, or that some
experimenter bias occurred when the cells were classified as
either inner or outer types. To addressthese problems, intra-
cellular injections were made into pairs of inner and outer cells
with highly overlapping dendritic [Link] waspossibleto target
overlapping inner/outer midget pairs for injection in the retinal
periphery becausemidget cell bodies that were closely spaced,
with a separation of about 25 pm or less,were invariably an
inner/outer cell pair with highly overlapped dendritic [Link]-
dritic field sizesof clearly identified inner and outer cells could
thus be compared at the samelocation in the sameretina (Fig.
1OC). In 29 out of 30 inner/outer pairs, the inner cell showed

superior retina, 6-7 mm eccentricity 50pm t

field diameter of 50-75 pm. Dendrites continue to show characteristic
Figure 7. Midget ganglion cells from mid-peripheral retina. The cells clustered branching patterns. Most cells that branch in the outer portion
illustrated are from the same retina, separated by about 500 pm at 6- of the IPL display a single primary dendrite that extends vitread before
7 mm eccentricity. At this eccentricity the midget cells reach a dendritic arborizing.

midget cells - far periphery

superior retina - 10 mm eccentricity

Figure 8. Midget ganglion cells from the far peripheral retina. Both cells are from a single retina, located within a few hundred micrometers of
each other at 10 mm eccentricity. Dendritic field diameter, 75-l 50 pm. At this eccentricity and beyond, the clustering patterns of the dendritic
tree are highly variable, ranging from cells with relatively densely branching, small dendritic fields to cells with larger and more sparsely branching
fields.
5342 Dacey l Human Midget Ganglion Cells

outer midget cells

Figure 9. Variable dendritic mor-

phology of neighboring midget cells in
the far retinal periphery. The two cells
shown are neighboring midget cells at
13 mm eccentricity that both stratify at
the same depth in the outer portion of
the IPL. The upper cell has a smaller,
more compact dendritic tree; the lower
cell has a larger tree that is divided into
multiple distinct and widely separated superior retina, 13 mm eccentricity 50 urn
arbors.

a larger dendritic field diameter (Fig. 10B). The increase in _. . , . . , . II ~ , ., ,

diameter ranged from -20% to 60% and averaged -3O%, the Clustereddendritic morpnology of nonfoveal midget Ceh
sameas was observed in the original sampleof single cells. Outside of the central 2 mm, as midget cells increasedsharply
Both the inner and outer midget cells appearedbroadly strat- in dendritic field diameter, a distinct spatial nonuniformity in
ified over about a third of the depth of the IPL. In whole-mount the branching patterns of the dendritic [Link]-
preparationsthe dendritesof an individual cell crossedover one in one dendritic tree, branches tended to form small, dense
another at different depths. Thus, a stratified pattern of branch- clusters,separatedby regionsof very low dendritic density. The
ing was observed in which two or more tiers of dendrites were degreeof clusteringcan beappreciatedby comparingthe branch-
clearly distinguished at distinct focal planes (Fig. 11A). The ing pattern of a midget cell with that of a parasolcell, the other
depth of stratification was measured,as describedin Materials common primate ganglion cell type that projects to the dorsal
and Methods, in serial vertical sectionstaken through five cells lateral geniculate nucleus (Leventhal et al., 1981; Perry et al.,
(two overlapping inner/outer cell pairs and a singleouter midget; 1984). The parasol cell distributes its dendritic branchesuni-
all cells from the retinal periphery). For the inner midget cells, formly throughout the dendritic field, economically filling the
90% of the dendritic measurementsoccupied about 30% of the available spacewith a relatively constant density of processes.
total depth of the IPL, falling between55%and 85%depth (inner By contrast, the midget cell showsobvious dendritic clusters
nuclear layer cell border = 0%, ganglion cell layer border = and equally obvious “holes” within the dendritic field (Fig.
100%) (Fig. 11B). Inner midget cells dendrites did not invade 12A,B).
the inner 10% of the IPL. The outer midget cells extended den- The dendritic clustering appearedto be characteristic of all
dritic branchesto the outer border of the IPL and were stratified nonfoveal midget cells; there was great variation, however, in
more broadly than the inner midgets,extending over about 40% the pattern of the clustering from cell to cell. Somecells showed
of the IPL depth, with 90% ofthe dendritic measurementsfalling few clusters and thus appeared sparsely branched; other cells
between 10% and 47% depth. showeda number of clustersand were more densely branched.
 The Journal of Neuroscience, December 1993, f3(12) 5343

1000

-z
1
5 100
5
n
5!
2
B
=
F 10
$
n = 30 cell pairs
Inner, n = 264
0 outer, n = 239 12 16 20
04 8
eccentriotv
1
0 4 6 12 16 20 0 50 100 150 200 250
eccentricity (mm from the fovea) outer mldget dendrltlc field diameter

midget cell pair

inner cell

r
-
\
25 pm nasal retina - 12 mm eccentricity relative location of dendrltbc fields

Figure 10. Dendritic field size difference between inner (presumed ON-center) and outer (presumed OFF-center) branching midget cells. A,
Scatterplot of midget dendritic field size as a function of retinal eccentricity with cells categorized as inner (solid circles) and outer (open circles)
branching types. Inset plots the inner to outer cell dendritic field diameter ratio and shows that a relatively constant ratio of 1.3 is maintained
across the retina. B, Dendritic field diameter for inner midget cells is plotted against diameter of outer midget cells for 30 midget cell pairs with
overlapping dendritic fields. A line indicating an inner:outer ratio of 1.3 was drawn through the data. C, An example of one inner/outer midget
cell pair used for the plot shown in B. Both cells display are similar overall morphology but the inner cell is larger than the outer cell. The inset
polygons were traced around the perimeter of the dendritic fields and show their actual spatial relationship in the retina.

There also appeared to be variation in the size of the clusters. maximum. The results of such an analysis for one midget and
It thus seemed possible that the midget cells could be further one parasol cell are shown in Figure 12. The midget cell had
divided into distinct types on the basis oftheir clustering pattern. distinct regions of high dendritic density (clusters) and low den-
To test this possibility, the degree of dendritic clustering was dritic density (holes) that clearly reflected the patchy organi-
quantified and compared for 118 well-filled midget cells. Con- zation of the dendritic tree illustrated in the camera lucida trac-
tour plots of dendritic density were generated from digitized ings in Figure 12, A and B. There was also a reasonably close,
images of individual dendritic trees (see Materials and Meth- but not always perfect, match between the subjectively observed
ods). Dendritic clusters were defined as those regions of the number of dendritic clusters (Fig. 12A,B, approximately eight
dendritic tree that had a density equal to or greater than 70% clusters) and the clusters that were arbitrarily defined by the
of the maximum density. Dendritic “holes” were defined as that 70% contour lines (Fig. 12C, nine clusters). By contrast, the
part of the dendritic field that had a density lower than 30% of parasol cell displayed a relatively uniform distribution of den-
5344 Dacey l Human Midget Ganglion Cells

field areas,shown in Figure 140, displayed from three to eight

clusters each.
The size of the dendritic clusterswas not strictly correlated
with the number of clustersper cell or with the diameter of the
dendritic tree. Cluster diameter (Fig. 148) ranged from -7 to
25 pm with a mean of 13 pm (SD = + 3). For small cells, many
clusterswere in the samesize range asthose of larger cells (Fig.
14C). The variation in cluster size and number together pro-
duced a range of morphologiesthat could appear quite distinc-
tive from one another. Sparselybranching cellswith a few widely
spacedclusterswere observed at the sameretinal eccentricities
asdenselybranching cellswith a larger number of closely spaced
clusters.
I- I*--- ---
The overall picture of midget dendritic morphology that
inl
emergesfrom this analysis can be summarized as follows. All
outer cells
cells with dendritic field diameters over 5-10 Mm displayed a
tendency for the dendritic tree to break up into small clustered
domains. These clusterstended to increasein number with in-
creasingdendritic field size, though the number of clusterswas
highly variable. At the sametime the clusterstended to remain
small relative to the overall size of the dendritic tree. Thus, the
ipl largest cells with relatively few clusters appeared sparsely
branched, and the smaller cells with many clustersappearedto
-i be denselybranched. Midget cellscould not be subdivided into
distinct morphological variants by cluster diameter, cluster
number, or both featurescombined.
Midget cell mosaics
Neither dendritic field size nor morphology provided a basis
WI
for subdividing midget cells beyond the well-establishedinner
Number of measurements (IO / division) and outer types. If only thesetwo types are present, then they
should establishtwo independentmosaicsof cell bodies; within
Figure Il. Depth of stratificationof midgetcell dendritic trees.A, each mosaic, all the dendritic trees should stratify at the same
Whole-mountview of an outer midgetcell intracellularlyfilled with depth in the IPL. This arrangementwould be equivalent to what
[Link], Planeof focusis on dendriticbranchesthat stratify has been found for the (Y-Y and p-X cells of the cat’s retina
closestto the cell body. Right, Planeof focusshiftedabout 6-8 pm
farther from the cellbody; a secondgroupof dendriticbranchescomes (Wglssleet al., 1981a,b) and for the magnocellular-projecting
into [Link] cell thusappearedto be broadlystratifiedover about parasol cells of the macaqueand human retina (Watanabe and
a third of the depthof the IPL. Scalebar, 25 pm.B, Measurements of Rodieck, 1989; Silveira and Perry, 1991; Dacey and Petersen,
dendriticdepthsampledfrom vertical sectionsthroughthreeouterand 1992).One property of these,aswell asother identified ganglion
two [Link] outercells(shaded bars) 90%of the dendritic
measurements fell within 1o-47%ofthe totaldepthofthe IPL, indicated cell mosaics(Buhl and Peichl, 1986; Dann and Buhl, 1987;
by the upper bracket at the right. For innercells(openbars) 90% of the Dacey, 1989a;seealso WPssleand Boycott, 1991, for a review),
dendriticmeasurements fell between55%and85%depth,indicatedby is that the degreeof overlap betweenneighboringdendritic trees
the lower bracket. Inner midgetcellsdid not extendterminaldendrites is approximately the sameacrossthe retina, yielding a dendritic
into the innermost10%of the IPL. network of uniform and characteristic coverage (dendritic field
area x cell density). It was difficult, however, to envision how
a single inner and a single outer midget mosaic could account
drites suchthat the 30% density contour line fell near the outer for such a large variation in the dendritic field size and in the
border of the dendritic tree, and the contour plot did not reveal branching patterns of singlecellsobservedin the presentresults.
any “holes” within the dendritic field. The 70% contour line I addressedthis question directly by intracellularly filling all
bounded a single large, more proximal part of the field where neighboring inner or outer midget cells within a small patch of
density was higher and relatively uniform. retina. To do this I took advantageof the sizedifferencebetween
The number of clusters per cell tended to increasewith in- somata of the inner and outer cells, which was visible in the
creasingdistance from the fovea. All but one of six cells in the retina in vitro and could be used to target either population
databasebetween 2 and 3 mm eccentricity showed two small reliably for intracellular injection. It was necessaryto inject a
clusters within a larger dendritic field (Fig. 13; also seecell number of cells of the sametype that were nearestneighbors.
tracings in Fig. 6). As dendritic field size increased fourfold, This task was possiblein the mid-peripheral and far peripheral
from -50 to 200 Km diameter, the mean number of clusters retina where cell density was relatively low and the spacingof
per cell increasedfrom two to seven (Fig. 14A,C). However, at neighboring midget cells appearedvery regular. Using this ap-
any given eccentricity the variation in cluster number waslarge. proach I injected 2 1patchesof midget cells(11 inner cell patches
For example, amongthe 18 midget cellswith a diameter of 100 and 10 outer cell patches),containing from 4 to 30 cells/patch
+ 5 Km the number of clustersranged from one to eight (mean (mean + SD = 6 + 6 cells). Patcheswere located 7-12 mm
+ SD = 4.4 f 1.7). Three other cellswith comparabledendritic from the fovea.
 The Journal of Neuroscience, December 1993, 73(12) 5345

Figure 12. Methodfor quantifyingthe

tendencyof midget ganglioncellsto
form localizedregionsof highdendritic
densitywithin their dendriticfields.A,
Cameralucidatracingsof a midgetcell
from the nasalretinal periphery(left)
and a parasolcell (the other common
ganglioncelltypeof the primateretina)
alsofrom the retinal periphery(right).
To facilitate comparisonof branching
patterns,the two cell typesare shown
at differentmagnifications. B, Simpli-
fied tracingof the dendriticbranching
pattern of both cells(somaand axon
omitted).Themidgetcelldendritictree
showsmultipleregionsof highandlow
dendriticdensities; bycontrast,thepar-
asolcell showsa comparativelyeven,
space-fillingbranchingpattern.C, Con-
tour plots made from digitized and
smoothedimagesof dendritictreeout-
linesshownin B. Contourlinesfollow
the 30%and70%of maximumdensity
[Link] the midgetcellthis defines
distinct regionsof high and low den-
dritic [Link] regionsboundedby
the 70%contourlinesaredesignated as
[Link] the parasolcell,
dendritic density is lower in the pe-
riphery of the field and relatively uni-
form over the center of the field; no
I patchesof highandlow density,asde-
‘\ 1 J finedby the30%and70%contourlines,
-- - -- --- canbe seen.

Several featuresof the morphology of the cells indicated that (0.027 mm2) gave a coverage of 1.1. This calculated coverage
within each patch, all the membersof a singlemosaichad been suggestedthat there should be a some small overlap of the
injected. First, the arrangementof the cell bodies, evaluated by dendritic [Link], as shownin Figure 15C, the irregular,
measuringcell-to-cell spacing,was highly regular. Second, the interlocking shape of the dendritic trees served to reduce the
coverage factor was constant acrossa patch, and was the same real coverage to 1, or slightly lessthan 1. The regular spacing
for patchesfrom different retinal eccentricities. Third, the depth of the cell bodies in the patch is shown in Figure 15B. One
of stratification of all injected cells in a patch was identical. measure of this regularity is given by the ratio of the mean
Fourth, adjacentdendritic treeswere shapedto interdigitate and nearestneighbor distanceto its standard deviation (Wgissleand
fill the available spacelike interlocking piecesof a jigsaw puzzle Riemann, 1978). For the patch of peripheral inner midget cells
(Wglssleet al., 198lc). Finally, when a cell was not completely shown in Figure 15, the mean nearest neighbor distance was
filled, an obvious hole in the mosaicwas apparent. 145 pm and the standard deviation was 18 pm, giving a ratio
The largestpatch, containing 42 inner midget cells,illustrates of 8, which approachesthe high degreeof regularity (- 9.5) found
the regular intercell spacingand uniform coverage for this mo- for the peripheral red/green cone mosaic (Curcio et al., 1991).
saic (Fig. 15). For all neighboring cells the dendritic field cov- Comparisonof neighboringinner and outer patchesin a single
eragewasconsistently lessthan, but very closeto 1. That is, the retina confirmed that the outer midget cells were consistently
adjacent dendritic trees approached one another but showed about 30% smaller in dendritic field diameter than their inner
virtually no overlap. Moreover, single dendrites from neigh- counterparts (Fig. 16). Both the inner and outer mosaicshad a
boring cells rarely crossedover one another and in no casedid coveragevery closeto 1. As a consequence,the outer cellsformed
the dendritic trees intermingle. Multiplying the density of cells a mosaicof higher density than the inner [Link] density ratio
in this patch (40 cells/mm*) by the mean dendritic field area of outer to inner midget cells was 1.7:1. Thus, at least beyond
5346 Dacey l Human Midget Ganglion Cells

= 186 & 22 pm). Given that the ratio of inner to outer dendritic
3 2 mm eccentricity 3.0 field diameter was 1.3: 1, and assumingthat the outer mosaic
shows the same local scatter as the inner mosaic, a mean of
- 140 pm and a diameter range of - 100-200 km would be
predicted for the outer midget cell mosaicat the retinal location
shownin Figure 15. The predicted diameter rangefor all midget
cells at this location would therefore be - 100-270 pm. This
variation of nearly threefold is closeto that found for the midget
sampleas a whole and supportsthe hypothesisthat the sample
of midget cells shown in Figure 4 is derived from only two
ganglion cell mosaics.
The detailed dendritic structure within the patches revealed
the significant variation in the clustered dendritic morphology
of the midget cells (Fig. 14). As in the total sampleof midgets,
3.0 2.7 the cells in the patchesrangedfrom very denselybranching and
highly clustered, to sparselybranching and minimally clustered
(e.g., arrowheadsin Fig. 15A indicate the two extremes). Thus,
much if not all of the morphological variation that has been
describedfor the sampleof singlecells can also be observed in
any single small patch of cells.
The filled cell patchesalso revealed that the dendritic clusters
and holes within the dendritic field of single midget cells were
maintained in the cell mosaic as a whole (Figs. 17, 18). Based
on the dendritic networks formed by other ganglioncell mosaics
(e.g., Wassle et al., 198la,b; Dacey, 1989), it was anticipated
that the dendritic clusters and holes would interdigitate such
that the dendritic network formed by the completemosaicwould
display a uniform density. However, the midget cell dendritic
2.5 trees did not overlap, and the holes were not filled in by the
interdigitation of neighboringdendrites (Fig. 17). Contour plots
of dendritic density acrosscell patchesthus revealed the same
regions of high and low density observed in single cells (Fig.
18).
Discussion
Midget ganglion cells: two types, two mosaics
In this study human midget ganglion cells were identified by
their characteristic dendritic field sizeat a given retinal location.
Midget cell dendritic fields showedno overlap in size with those
of any other ganglion cell type at a given retinal location, and
by this measureformed a singledistinct grouping. The human
Figure 13. Dendritic clustering in midget cells is first observed between midgets were divided into two types, presumedto correspond
2.4 and 3.2 mm eccentricity. Smoothed gray scale images and contour to ON- and OFF-center cells, by the depth of stratification of
plots of midget cells from the near peripheral retina are shown. Contour their dendritic trees, in agreementwith previous studies(Perry
lines follow the 30% and 70% density values as in Figure 12. At this et al., 1984; Rodieck et al., 1985;Watanabeand Rodieck, 1989).
eccentricity the cells show a distinct tendency to divide into two separate
regions of high dendritic density. The dendritic clusters may be relatively Despite the general agreementamong previous studies, sup-
widely spaced as for the cells shown at 3.2,2.7, and 2.4 mm eccentricity, ported by the presentresults,regardingthe subdivision of midget
or they may be closely spaced as for the two cells shown at 3.0 mm. cells into only the inner and outer types, there is still some
Only one cell in the sample of filled cells over this eccentricity range dispute about the number of ganglion cell types in the midget
(lower left, 2.5 mm; camera lucida tracing for this cell is shown in the size range. A recent Golgi study of human retina hassuggested
upper left of Fig. 6) had an arbor that did not divide into clusters.
that the ganglion cells with the smallestdendritic fields at all
eccentricities can be subdivided by differencesin dendritic field
7 mm eccentricity, there must be five outer midget cells for size alone (Kolb et al., 1992) into the Pl cells (-5-15 pm di-
every three inner midget cells. The data on inner and outer ameter), and the P2 cells (- 1O-60 pm diameter). In the present
dendritic field size for eccentricities lessthan 7 mm (Fig. 10) study no suchsubdivision on the basisofdendritic field diameter
suggestedthat this density difference may also hold at more was possible(seeFig. 4). A distinct subgroupof midget cellsin
central locations. the far retinal periphery with dendritic field diameters in the
The range of dendritic field sizes observed within a single range of 15-20 pm, as reported by Kolb et al. (1992), was not
patch, whether of inner or outer cells, can also account for the observed. The smallest peripheral midget cells had diameters
largelocal scatter in dendritic field diameter illustrated in Figure of 35-45 Mmin the nasalretina; in the other retinal quadrants
4. For example, the inner cells in the patch shown in Figure 15 somecells as small as 50 Km in diameter were found between
ranged in diameter from 140 to 270 pm (n = 29; mean + SD 7 and 8 mm eccentricity (Fig. 4A). I can suggesttwo possible
 The Journal of Neuroscience, December 1993, 13(12) 5347

n= 118

.
..
.

.m. .
I
0
0 100 zoo 100 200 300
dendritic field diameter (pm) dendritrs field diameter Cm)

Figure 14. Numberandsizeof midget

cell dendritic clusters as a function of
eccentricity. A, Number of dendritic
clusters per cell plotted against den-
dritic field diameter. Average cluster
did = 180
number increases with increasing den-
dritic field size but cluster number is
highly variable. B, Mean cluster di-
ameter ranges between 10 and 20 pm
over most of the dendritic field size
range. C, Contour plots of smoothed,
gray scale imagesillustrate how the
number of clusters increases with den-
dritic field size but the range of cluster
sizes remains about the same. q,.Con-
tour plots illustrate the variabthty in
cluster number for cells of similar den-
did - 95
dritic field sizes. dfd, dendritic field di-
ameter (in pm).

explanations for this discrepancy. Given the great variability in the degree of scatter shown in Figure 4 without recourse to
the dendritic field size range for the inner and outer midgets, additional unidentified cell types in this size range.
such small cells in the retinal periphery could exist, especially
in the smaller, outer cell mosaic, but they would be very rare Midget cell coverageand density
and would not comprise a separate and complete mosaic (as- Assumingthat the coverageof 1 found for the midget cell patch-
suming a coverage of one or greater). Alternatively, the clustered es holds over the entire retina, it is possibleto calculate the
nature of the peripheral midget cell’s dendritic tree could be the density of midget cells usingthe dendritic field size data shown
basis for rare observations of peripheral midgets with small in Figure 4. Figure 19A plots midget ganglion cell density (cov-
dendritic fields. Some cells possess multiple, spatially distinct erage/dendritic field area in cells/mm*) as a function of retinal
arbors, each arbor much smaller than the total dendritic field eccentricity using the values from the curve fit to the data in
(e.g., Figs. 9, 12); a partial Golgi impregnation of such a ccl1 Figure 4. These values were compared with the mean density
could result in an artifactual image of an extremely small den- values from the temporal, upper and lower meridians for the
dritic field in the retinal periphery. total ganglioncell counts in human retinasgiven by Curcio and
Although the existence of Kolb’s Pl and P2 cell types could Allen (1990). This comparison suggeststhat the proportion of
not be confirmed in the present study, the great variability in midget cellssteadily increasesasthe fovea is [Link]-
midget cell dendritic field size did suggest that distinct subpo- sity wasrelatively constant at about 45-50% ofthe total ganglion
pulations beyond the inner and outer mosaics might exist in cell population from 15 to 9 mm eccentricity; from 8 to 3 mm
this sample. However, the range of dendritic field sizes within eccentricity, density increasedto about 95%of the total ganglion
each mosaic argues against this possibility. The inner midget cell population (Fig. 19B).
cells showed on average a 30% larger dendritic field diameter This picture of human midget ganglion cell topography con-
than their outer cell counterparts (Fig. 10); this difference is a trasts with resultsin macaque,where midget ganglioncellshave
property of the mosaic as a whole (Fig. 16). Taking both mosaics been reported to make up a constant proportion of the ganglion
together, there is an almost threefold variation in dendritic field cell population, about 80%, at all retinal locations (Perry et al.,
size at a given retinal location. This variation predicts nicely 1984), so it is necessaryto considerpossiblesourcesof error in
5348 Dacey l Human Midget Ganglion Cells

inner midget cells

P
,Y&+

dendritic network suoerior retina - 12 mm eccentncftv 200 urn

.
. l . .
.
. l l . . .
. l .
.
. l
. .
. . . l
. .
. .
.
. . .
. .
.
.
cell body mosaic l dendritic overlap

Figure 15. The midgetcell mosaic.A, Patchof 42 inner midgetcellsintracellularlyfilled with [Link] the dendritictreeswere
madefor all cells(29 cells)that showedcompletedendriticfilling. B, Injectedccl1bodies(solid circIes) showa highly regulardistribution.C,
Tracingsaroundthe dendriticfieldsof filled cellsrevealthat neighboringdendritictreesdo not overlapandshowa coverageof 1 or slightlyless.

the resultsplotted in Figure 19. The coverageof 1 for the midget highly unlikely, however, becauseit would meanthat the midget
cells is derived from cell patches at eccentricities of 7 mm or cellswould not have the minimum coveragerequired to sample
greater. If midget coverage progressively decreasedtoward the all points in the visual field. Alternatively, there is the possibility
fovea, the percentageof midget cellswould be lower. This seems that Curcio and Allen (1990) overestimated the total number

Figure 16. Differencein sizeandden-

sity of innerandoutermidgetcellmo-
[Link] bodiesare shownin darker
shadingand the dendritic fields are out-
Iined and lightly shaded. Adjacent
patches of inner and outer midget cells
show that both cell types show similar
coverage very close to 1. Outer cells have
on average smaller dendritic field di-
ameters, resulting in a mosaic of higher
density. Ratio of outer cell density to
inner cell density is 1.7: 1
 The Journal of Neuroscience, December 1993, 13(12) 5349

inner midget cells outer midget cells

Figure 17. Neighboring midget cell

dendritic trees avoid intersecting and
maintain a constant coverage. High-
magnification camera lucida tracings of
inner and outer midget cell patches show
that the individual dendrites of neigh-
boring dendritic trees avoid entering
into [Link] resultis that
the spatialinhomogeneities of the in-
dividual dendritictreesare preserved
within the cell patch. Outlinesof the
dendritictrees(insetat lower right) il-
superior retina - 12 mm eccentricity 50 urn lustratethelackof overlapof theneigh-
boringdendrites.

of ganglion cells in the retinal periphery, and that midget cells periphery to about 40: 1 in the central representation(Connolly
actually comprise a much greater proportion than the 45-50% and Van Essen,1984). The presentdata on midget cell density
shown in Figure 19. The number of amacrine cells present in thus provide renewed support for the hypothesis that the dif-
the ganglion cell layer increases greatly in the retinal periphery, ferent threshold gradientsobserved for the sustained-chromatic
and if some of the displaced amacrines were counted as ganglion and transient-luminance channelsof human vision are related
cells, ganglion cell density would be significantly overestimated. to the different density gradientsofthe magno-and parvocellular
In the macaque retina, immunostaining for GABA combined pathways (Drasdo et al., 1991; for review, seeDrasdo 1991).
with retrograde degeneration of ganglion cells has shown that
in the far periphery about 80% of the neurons in the ganglion Midget cell Nyquist frequency and visual acuity
cell layer were displaced amacrine cells (Whsle et al., 1990). In It is now well acceptedthat for central vision, spatial acuity is
the human retina, Curcio and Allen (1990) estimated the num- limited by the spacingof the cone photoreceptors (Bankset al.,
ber of displaced amacrine cells based on soma size and other 1987; Williams and Coletta, 1987; Hirsch and Curcio, 1989;
morphological criteria to reach about 75% in the far retinal Anderson et al., 1991). However, beyond about 6-7” eccentric-
periphery in fair agreement with the more conclusive findings ity, the maximum spatial resolution afforded by cone spacing-
in macaque. Thus, it seems unlikely that the human ganglion the Nyquist frequency-exceeds psychophysicalmeasuresof vi-
cell counts are in error by more than 5%. sual acuity. The cone Nyquist frequency falls off to about 10
The central-to-peripheral decrease in the proportion of midget cycles/degreeperipherally, whereasvisual acuity declinessharp-
cells supports the previous suggestion that the ratio of human ly to lessthan 1 cycle/degreein the peripheral temporal retina.
midget to parasol ganglion cells changes from about 30: 1 in the Two psychophysical studies have suggestedthat the density
central retina to about 3:l in the retinal periphery (Dacey and of parvocellular-projecting ganglion cells may limit spatial res-
Petersen, 1992). It has been reported that the ratio of the den- olution in the periphery (Thibos et al., 1987;Merigan and Katz,
sities of parvocellular to magnocellular neurons in the macaque 1990).In thesestudiesmeasuresofgrating acuity werecompared
dorsal lateral geniculate also changes from about 4: 1 in the far with Nyquist frequenciescalculated from published values of
5350 Dacey l Human Midget Ganglion Cells

ccc = 13 mm
A

Figure 18. Contour plots of dendritic

density show that inhomogeneities ex-
ist within the midget cell mosaic as a
whole. Holes in individual dendritic
trees are not filled in by neighboring
trees. A.&$, Tracings around the five
individual inner midget cell dendritic
trees that make UD a oatch show no
overlapping dendrites. kight, Contour
plot of 30% and 70% of maximum den-
sity of the smoothed gray scale image
of the patch. Retinal eccentricity = 13
mm. &Left, Tracings around four in-
dividual dendritic trees that make up a
more centrally located patch of inner
midget cells. Right, Contour plot as in
A. Retinal eccentricity = 9 mm.

midget cell density acrossthe macaqueretina. Merigan and Katz leaguessuggestthat such an orientation protocol is not “alias-
compared grating acuity in macaquemonkey with the Nyquist free” and can lead to acuity estimatesthat are up to 1.5 times
frequency derived from the estimateddensity of a singlemidget higher than the Nyquist frequency of the underlying mosaic
cell mosaic, that is, one-half the estimated total midget cell (Williams and Coletta, 1987). These authors therefore adopted
density, assumingthat a single mosaic representsan indepen- a discrimination protocol basedon the detection of the direction
dent samplingarray. Thibos and colleaguescompared human of motion of a grating, which they arguedwas free of sampling
grating acuity, usinginterference fringes imageddirectly on the artifacts. Anderson and colleaguesthen went on to compare
retina, to estimatesof total midget cell density in macaquemon- their results with the Nyquist frequency calculated from mea-
key. Despite the twofold variation in ganglion cell density that suresof total ganglioncell density in the human retina by Curcio
was employed, both studiessuggestedthat midget cell density and Allen (1990). Not surprisingly, they found that the Nyquist
setsthe acuity limit beyond about lo”, and this conclusion has frequency for the total ganglion cell population was higher, by
gained someacceptance(Lennie et al., 1990; Wlssle and Boy- a factor of -2, than achromatic acuity.
cott, 1991). The finding of a coverageof 1 for the human midget ganglion
There are two potential difficulties with the above resultsthat cells makesit now possibleto calculate the highest anatomical
have prompted another look at the relationship between the resolving power for the midget cell mosaic. The Nyquist fre-
midget cell mosaicand spatial acuity. First, the calculations in quency, assuminghexagonal packing of ganglion cells, equals
both studieswere based on ganglion cell density estimatesin l/v/3(0), where D is the intercell spacingin degrees(Snyder
which midget cells were assumedto make up a constant pro- and Miller, 1977). For the midget cells, D is equivalent to the
portion of the retinal ganglion cells acrossthe retina (Perry et dendritic field diameter. Figure 2OA plots spatial resolution in
al., 1984). Thus, their cell density estimates for the macaque cycles/degreefor the midget ganglion cells from 1” to 55” using
differed from those calculated here for the human. Second, a the dendritic field diametersgiven by the equation for the curve
more recentstudy (Anderson et al., 1991)hasfound significantly fit to the data in Figure 4. Thesedata are compared to the recent
lower values for human spatial resolution acrossthe visual field data of Anderson et al. (1991), Merigan and Katz (1990), and
than those measuredby Merigan and Katz or Thibos and col- Thibos et al. (1987)and the classicalfindingsof Wertheim (1891).
[Link] Merigan and Katz and Thibos and colleagues In the retinal periphery the grating acuities measuredby Thibos
measuredgrating acuity usinga discrimination protocol in which and colleaguesin human and by Merigan and Katz in macaque
the orientation of the grating was detected. Anderson and col- are about 1.5-2 times higher than the human midget cell Nyquist
 The Journal of Neuroscience, December 1993, f3(12) 5351

A
rl Midget ganglion cells:
coverage / dendritic field area

0 Total ganglion cells:

temporal, upper and lower meridians
(Curcio and Allen, ‘90)
L
a
: 0
0
0
0
0 0
n
0
O
0 0 0 0 0 0

. 0
. . 0
lo2 I I I I
0 4 a 12 16

Figure 19. Calculation of midget gan-

glion cell density as a function of retinal
eccentricity. A, Midget cell density is
calculated from the dendritic field area
given by the equation for the curve
shown in Figure 4 and a coverage of 1
across the retina. The midget cell den-
sities are compared with published val-
ues for total ganglion cell density (Cur-
cio and Allen, 1990). Total ganglion cell
density was taken as the mean of values
for the temporal, upper and lower me-
ridians. B, From the two sets of data
plotted in A the percentage of midget
cells of the total ganglion cell popula-
tion is shown as a function of eccen-
tricity. From 15 to 8 mm eccentricity,
the midget cells make up 45-50% ofthe
aandion cells. From 8 to 4 mm, relative
0 4 8 12 16 density increases dramatically’to reach
eccentricity (mm from the fovea) about 95% of the total ganglion cells.

frequency. By contrast, the data given by Anderson et al. (199 1) the retinal periphery relatesto the transition in dendritic mor-
for achromatic acuity closely match the midget Nyquist fre- phology seenfor midget cellsaround 6” eccentricity. This is the
quency from 55” to about 6-7”. In addition, the classicaldata eccentricity where visual acuity beginsto match the midget cell
of Wertheim, consideredan underestimateof peripheral acuity Nyquist limit. Figure 20B indicates that between 0” and 6”, all
by modem standards,also closely match both the anatomical midget ganglion cells show the minimum dendritic tree diam-
data and the resultsof Anderson and colleaguesfrom about 10” eters, 5-l 0 pm. At this eccentricity the ratio of ganglioncellsto
to the far periphery. The correspondencebetween thesediverse conesis greater than 2: 1(WPssleet al., 1989a;Curcio and Allen,
setsof data supportsthe conclusion that spatial resolution mea- 1990). There are therefore enough ganglion cells that signals
sured by Anderson and colleaguesis a function of the sampling from a single cone photoreceptor can be conveyed indepen-
density of the mosaic of midget cell dendritic fields. The pre- dently to both ON- and OFF-center midget ganglioncells. Con-
viously reported correspondencebetweenhuman and macaque sequently, the spatial resolution afforded by cone spacingcan
acuity and macaquemidget cell density may be a consequence be preserved in the output of both the ON and OFF midget
of overestimates of both acuity and midget cell density in the ganglion cell [Link] 6” eccentricity, midget cell den-
retinal periphery. dritic trees enlargeand form multiple dendritic [Link]
Further evidence that the midget cell mosaiclimits acuity in clusters suggestthe convergenceof input from more than one
5352 Dacey - Human Midget Ganglion Cells

l Midget cell Nyquist frequency based on dendritrc field drameter

- - - Human acurty - (Wertheim. 1891)

Figure 20. Comparison of the midget
cell Nyquist frequency with measure- + Human acurty grating orientation protocol (Thibos et al. ‘87)
ments of human spatial resolution. A,
Given a constant dendritic field overlap o Macaque acurty - gratrng onentatron protocol (Merigan and Katz ‘90)
of 1, the midget dendritic field diameter
is equivalent to intercell spacing and, 0 Human acuity drifting gratrng protocol (Anderson et al. ‘91)
assuming approximately hexagonal
packing, the Nyquist limit for human
midget cells = l/(43 x dendritic field
diameter). To generate the curve shown
in this figure for the midget cells, den-
dritic field diameter was taken from the
equation for the curve fit to the data
given in Figure 4. The curves for the
midget cells and the data of Anderson
et al. (199 1) are closely matched from
55” to about 6” eccentricity. The curve
0
shown for Wertheim’s classical data also .
matches the midget Nyquist frequency, 1
.
n A
as well as the data of Anderson and 1
0
colleagues, from about 10” out to the
far periphery. (Wertheim’s data were 0 10 20 30 40 50 60
given as normalized relative to fovea1 eccentncrty, temporal retrna (degrees)
resolution, his Table 2; absolute reso-
lution in cycles/degree was calculated
for the temporal retina by reference to I
the data provided in Table 1 of the orig- 2 mm from fovea
inal work). Other recent measures of
grating acuity in both human and ma-
caque (Thibos et al., 1987; Merigan and
Katz, 1990) give higher values for spa-
tial resolution in the periphery. B, Scat-
terplot of dendritic field diameter ver-
sus eccentricity for midget cells in the
central and near peripheral retina pro-
vides an anatomical explanation for the
point at which the Nyquist frequency
given by midget cell dendritic field di-
ameter matched spatial acuity. From 0
to 6“ midget cells have single arbors that
receive signals from only a single cone
photoreceptor; in this region cone spac-
ing limits spatial resolution. Beyond 6”
the midget cells show multiple den-
dritic clusters that are assumed to re-
ceive convergent input from a number
of cones and the acuity limit would
thereby be shifted to the midget cells at
this eccentricity. Open circles, tempo-
4 8 12 16
ral, upper and lower retina; solid circles,
nasal retina. eccentricity (degrees)

cone bipolar cell. Consistent with this conclusion, it has been (Dacey and Petersen, 1992). Such an asymmetry has not been
shown that at about this eccentricity the ratio of ganglion cells previously noted for either macaque or human ganglion cells,
to cones declines to lessthan 2: 1, so convergent input from though the presenceof size differencesbetween ON- and OFF-
multiple conesto singlemidget ganglioncellsbecomesprobable center ganglioncell pairs hasbeenobservedin other mammalian
(Whsle et al., 1990). Correspondingly, spatial acuity should no species(for review, seePeichl, 1991). This finding wasconfirmed
longer attain the cone Nyquist frequency but instead must be and extended in the presentstudy. Intracellular stainingof midg-
limited by the midget cell Nyquist frequency. et cell patchesshowedthat the dendritic coverage for both the
inner and outer cell mosaic was the [Link] a consequence,
An asymmetry in the spatial scaleof the ON and OFF midget the smaller-diameterouter cellsformed a higher-density mosaic
mosaic than the larger-diameter inner cells. Given a difference of about
In a previous report we showedthat the inner midget and parasol 30% in dendritic field diameter, the density ratio of outer to
ganglion cells(presumedON-center) were consistently larger in inner midget cellsshouldbe about 1.7: [Link] result alsopredicts
diameter than their outer (presumedOFF-center) counterparts that the spatial resolution afforded by the OFF-center mosaic
 The Journal of Neuroscience, December 1993. 13(12) 5353

Figure 21. Hypothesis that dendritic

clustering in peripheral midget cells re-
flects selective connections with a sub-
set of midget bipolar cells. The upper
midget ganglion cell dendritic tree panel shows a camera lucida tracing of
a small patch of presumed flat midget
cone bipolar cells from macaque mon-
key retina revealed by immunostaining
for the calcium-binding protein recov-
erin (Milam et al., 1993). The cell bod-
ies are stippled and the axon terminals
are solid. Also shown is a tracing of a
human midget ganglion cell at about
the same retinal eccentricity. In the low-
er left panel a tracing has been made
around the axon terminal arbors of all
bipolar cells to illustrate their mosaic
organization in the IPL. The middle
panel is a simplified tracing of the den-
dritic branches of the midget ganglion
cell to illustrate more clearly the ten-
dency of the dendrites to form clusters.
Scale bar, 25 pm. In the right panel the
ganglion cell tree has been superim-
posed upon the bipolar mosaic. The
dendritic tree is not arranged to make
uniform contact with all of the 17 or so
bipolar terminals within its field. The
size of the dendritic clusters approxi-
mates the size of one to three bipolar
axon terminal fields. The holes between
the clusters are regions where bipolar
terminals may make few if any contacts
with the midget cell dendritic tree.

shouldbe about 1.3 times that of the ON-center mosaic,at least In central retina the cone-specific center responsesof par-
in the retinal periphery where such a size difference has been vocellular-projecting ganglion cells are presumedto be a con-
directly observed. There is psychophysical evidence that the sequenceof the one-to-one connectivity betweena singlecone,
visual systemshowsa greater sensitivity to decrementsin light a midget bipolar cell, and a midget ganglion cell (Boycott and
level than to light increments(e.g., Bowen et al., 1989; Tyler et Dowling, 1969; Kolb, 1970; Kolb and Dekorver, 1991; Calkins
al., 1992; for review, seeFiorentini et al., 1990; Bowen et al., et al., 1992).However, color-opponent responseshave alsobeen
1992) but how this finding relates to receptive field sizes of observed beyond the central 6-7” (Wiesel and Hubel, 1966;
ON- and OFF-center neuronsis not clear. Human visual evoked Gouras, 1968; DeMonasterio and Gouras, 1975) where it is
potentials generatedby positive and negative contrast stimuli probable that midget ganglion cells receive convergent input
believed to isolatethe ON- or OFF-visual pathways suggestthat from more than one midget bipolar cell. At this eccentricity it
the responseto stimulation of the OFF-pathway shows finer is not at all clear how a cone-specificcenter responsewould be
spatial tuning than the ON-pathway response(Zemon et al., [Link] a recent review, Wassleand Boycott (199 1) spec-
1988),but this conclusion hasnot yet been supported by other ulated that color-opponent receptive fieldsmay ariseby selective
approaches. connections between midget bipolar cells and ganglion cells in
the IPL. They pointed out that, in peripheral retina, irregularity
Dendritic clustersand the color selectivity hypothesis in the dendritic morphology of the midget cells may reflect this
Beyond the central retina, midget cells possessdendritic trees selectivity to some degree. This idea is a logical extension of
that are not spatially homogeneous;terminal dendrites cluster recent findings indicating that horizontal cellsdo not make col-
into small regionsof high density. Moreover, becausethe den- or-selective connectionswith cones(Boycott et al., 1987;Whsle
dritic treesof neighboring midget cells do not overlap, the clus- et al., 1989a; Dacheux and Raviola, 1990), but that midget
tered dendritic morphology is preserved in the overall midget bipolar cells contact only singleconeswell out into the retinal
cell mosaic. Previous studies of midget cell morphology have periphery (Boycott and WHssle,1991). It has now been shown
not commented on the tendency for dendritic clustering, but a that the density of cone bipolar cells is great enough to deliver
careful review of the published camera lucida tracings reveals cone-specific signalsto the IPL, even in the periphery, where
many examplesof this type of morphology (Perry et al., 1984, the ratio of cone bipolar cells to conesallows for one ON and
their Fig. 2C-F, Watanabe and Rodieck, 1989, their Fig. 2d-f, one OFF midget bipolar cell per cone (Martin and Griinert,
Kolb et al., 1992, their Fig. 18) and Kolb and colleaguesdid 1992;Milam et al., 1993).Moreover, recent electrophysiological
note that in mid-peripheral retina midget cellssometimes“have results suggestthat both the centers and surrounds of color-
two dendritic heads.” The presentresultsshowthat the tendency opponent cellsreceive specificred coneor greencone input (Reid
for the midget cell to form spatially distinct dendritic clusters and Shapley, 1992) further reinforcing the requirement for col-
is a characteristic feature of the great majority, if not all, of the or-selective connectionsat the level of the midget cone bipolar-
midget cells beyond the central 2-3 mm. midget ganglion cell connection.
5354 Dacey - Human Midget Ganglion Cells

Is it possible that the dendritic clusters reflect selective con- Boycott BB, Hopkins JM, Sperling HG (1987) Cone connections of
nections made with a cone-specific subset of bipolar cells within the horizontal cells of the rhesus monkey’s retina. Proc R Sot Lond
[Biol] 229:345-379.
the midget cell’s dendritic field? In thinking about this question Buhl EH, Peichl L (1986) Morphology of rabbit retinal ganglion cells
it is helpful to compare the midget cell dendritic morphology projecting to the medial terminal nucleus of the accessory optic sys-
with that of the mammalian horizontal cell. The main dendrites tem. J Comp Neurol 253:163-174.
of horizontal cells give rise to multiple small clusters of finger- Calkins DJ, Schein SJ, Tsukamoto Y, Masarachia P, Sterling P (1992)
like terminals; each terminal marks a region of synaptic contact Parallel pathways to midget ganglion cells in macaque fovea. Invest
Ophthalmol Vis Sci 33: 1173.
with a single cone pedicle. Because the horizontal cell makes Connolly M, Van Essen D (1984) The representation of the visual
nonselective contact with every pedicle in its field, the dendritic field in parvicellular and magnocellular layers of the lateral geniculate
clusters provide an unambiguous map ofthe spacing and density nucleus in the macaque monkey. J Comp Neurol 226:544-564.
of the array of overlying cone pedicles. The dendritic clusters Creutzfeldt OD, Lee BB, Elepfandt A (1979) A quantitative study of
chromatic organization and receptive fields of cells in the lateral ge-
of the midget ganglion cells are not as distinctly defined as those niculate body of the rhesus monkey. Exp Brain Res 35~527-545.
of the horizontal cells but could be clearly delineated by making Curcio CA, Allen KA (1990) Topography of ganglion cells in human
topographic maps of dendritic density. Isodensity maps showed retina. J Comp Neurol 300:5-25.
the midget cells to have highly clustered dendritic trees when Curcio CA, Allen KA, Sloan KR, Lerea CL, Hurley JB, Klock IB, Milam
compared to the parasol ganglion cell type. For the horizontal AH (199 1) Distribution and morphology of human cone photore-
ceptors stained with anti-blue opsin. J Comp Neurol 3 12:6 1O-624.
cell dendritic clusters it is only the visualization of the overlying Dacey DM (1988) Dopamine-accumulating amacrine cells revealed
cone pedicles that clearly reveals their functional significance. by in vitro catecholamine-like fluorescence display a unique mor-
Similarly, superposition of the mosaic of midget cone bipolar phology. Science 240: 1196-l 198.
axon terminals and the midget ganglion cell dendritic tree might Dacey DM (1989a) Axon-bearing amacrine cells of the macaque mon-
key retina. J Comp Neurol 284:275-293.
demonstrate whether the dendritic clustering reflects selective Dacey DM (1989b) Monoamine-accumulating ganglion cell type of
connections between the two mosaics. the cat’s retina. J Comp Neurol 288:59-80.
Recently a selective immunostain for the presumed flat midg- Dacey DM (1990) The dopaminergic amacrine cell. J Comp Neurol
et bipolar cells in macaque retina (Milam et al., 1993) has made 301:461489.
such an exercise possible. Figure 2 1 compares the bipolar axon Dacey DM (1993) Morphology of a small field bistratified ganglion
cell type in the macaque and human retina. Vis Neurosci, in press.
terminal mosaic to the dendritic morphology of a single human Dacey DM, Petersen MR (1992) Dendritic field size and morphology
midget cell at 6 mm eccentricity. Overlaying the midget cell of midget and parasol ganglion cells of the human retina. Proc Nat1
dendritic field with the mosaic of biDolar axon terminals reveals Acad Sci USA 89:9666-9670.
two important features of their sDatia1 relationship. First. the Dacheux RF, Raviola E (1990) Physiology of HI horizontal cells in
the primate retina. Proc R Sot Lond [Biol] 239:2 13-230.
sizes ofthe midget cell dendritic clusters are appropriate to make Dann JF, Buhl EH (1987) Retinal ganglion cells projecting to the
selective contact with one or a few bipolar axon terminals in the accessory optic system in the rat. J Comp Neurol 262: 14 l-l 58.
field. Second, the “holes” in the dendritic field that separate the de Monasterio FM, Gouras P (1975) Functional properties ofganglion
clusters suggest that a number ofbipolar terminals in the midget cells of the rhesus monkey retina. J Phvsiol (Land) 25 1: 116-l 95.
cell’s field would not make direct contact with the midget cell Derrington AM, Krauskopfj, Lennie P (lG84) ‘Chromatic mechanisms
in lateral geniculate nucleus of macaque. J Physiol (Lond) 357:241-
dendritic tree. In the example shown there are about 18 bipolar 265.
axon terminals in the midget cell’s dendritic field, but only about Drasdo N (199 1) Neural substrates and threshold gradients of pe-
nine of them significantly overlap the midget ganglion cell den- ripheral vision. In: Limits ofvision (Kulikowski JJ, ed), Vo15, Vision
drites. It is possible, then, that a color-selective center response and visual dysfunction (Cronly Dillon JR, ed), pp 250-264. New
of the midget ganglion cell derives from a submosaic of midget York: Macmillan.
Drasdo N, Fowler CW (1974) Non-linear projection of the retinal
bipolar cells that convey signals from only one population of image in a wide-angle schematic eye. Br J ephthalmol 58:709-714.
cones, either red or green. If this were the case, then the bipolar Drasdo N. Thomoson CM. Deelev RJ (199 1) Psvchoohvsical evidence
cells that overlap with the “holes” in the midget cells dendritic of two gradients of neural samplings in p&ipceral -&ion. In: From
field could mediate the opponent surround by providing indirect pigments to perception (Valberg A, Lee BB, eds), pp 189-192. New
York: Plenum.
inhibitory input via a sign-inverting amacrine cell. This ana- Famiglietti EV Jr, Kolb H (1976) Structural basis for ON- and OFF-
tomical picture is also consistent with the evidence that for many center responses in retinal ganglion cells. Science 194: 193-l 95.
red-green opponent cells the “surround” does not in fact sur- Fiorentini A, Baumgartner G, Magnussen S, Schiller PH, Thomas JP
round the center but is coextensive with it (Reid and Shapley, (1990) The perception of brightness and darkness: relations to neu-
1992). ronal receptive fields. In: Visual perception: the neurophysiological
foundations (Spillmann L, Werner JS, eds), pp 129-l 6 1. San Diego:
Academic.
References Gouras P (1968) Identification of cone mechanisms in monkey gan-
glion cells. J Physiol (Lond) 199:533-547.
Anderson SJ, Mullen KT, Hess RF (199 1) Human peripheral spatial Heimer GV, Taylor CED (1974) Improved mountant for immuno-
resolution for achromatic and chromatic stimuli: limits imposed by fluorescent preparations. J Clin Path01 27:254-256.
optical and retinal factors. J Phvsiol (Land) 442:47-64. Hirsch J, Curiio CA (1989) The spatial resolution capacity of human
Banks MS, Geisler WS, Bennett-PJ (1987j The physical limits of fovea1 retina. Vision Res 29: 1095-l 101.
grating visibility. Vision Res 27: 19 15-l 924. Kaplan E, Lee BB, Shapely RM (1990) New views of primate retinal
Bowen RW, Poke-my J, Smith V (1989) Sawtooth contrast sensitivity: function. In: Progress in retinal research, Vol 9 (Osborne N, Chader
decrements have the edge. Vision Res 29: 1501-1509. G, eds), pp 273-336. New York: Pergamon.
Bowen RW, Pokomy J, Smith VC, Fowler MA (1992) Sawtooth con- Kolb H (1970) Organization ofthe outer plexiform layer ofthe primate
trast sensitivity: effects of mean illuminance and low temporal fre- retina. Philos Trans R Sot Lond [Biol] 258:261-283.
quencies. Vision Res 3211239-1247. Kolb H, Dekorver L (199 1) Midget ganglion cells of the parafovea of
Boycott BB, Dowling JE (1969) Organization of the primate retina: the human retina: a study by electron microscopy of serial sections.
light microscopy. Philos Trans R Sot Lond 1Bioll 255:109-184. J Comp Neurol303:617-636.
Boycott BB, W&sie H (199 1) Morphological’cla&fication of bipolar Kolb H, Linberg KA, Fisher SK (1992) Neurons of the human retina:
cells of the primate retina. Eur J Neurosci 3: 1069-1088. a Golgi study. J Comp Neurol 3 18: 147-l 87.
 The Journal of Neuroscience, December 1993, 13(12) 5355

Lennie P, D’Zmura M (1988) Mechanisms of color vision. CRC Crit ;;!;rt and broad-band channels of the visual system. Nature 343:
Rev Neurobiol 3:333400.
Lennie P, Trevarthen C, Van Essen D, Wlssle H (1990) Parallel pro- Shapley R, Perry VH (1986) Cat and monkey retinal ganglion cells
cessing of visual information. In: Visual perception: the neurophys- and their visual functional roles. Trends Neurosci 9:229-235.
iological foundations (Spillmann L, Werner JS, eds), pp 103-l 28. San Silveira LCL, Perry VH ( 199 1) The topography of magnocellular pro-
Diego: Academic. jecting ganglion cells (M-ganglion cells) in the primate retina. Neu-
Lennie P. Haake PW. Williams DR (199 1) The construction of chro- roscience 40:2 17-237.
matically opponent receptive fields. In:’ Computational models of Snyder AW, Miller WH (1977) Photoreceptor diameter and spacing
visual perception (Movshon JA, Landy SC, eds), pp 61-72. Cam- for highest resolving power. J Opt Sot Am 673696-698.
bridge, MA: MIT Press. Thibos LN, Cheney FE, Walsh DJ (1987) Retinal limits to the detec-
Leventhal AC, Rodieck RW, Dreher B (198 1) Retinal ganglion cell tion and resolution of gratings. J Opt Sot Am 4: 1524-l 529.
classes in the old world monkey: morphology and central projections. Tyler CW, Chan H, Lui L (1992) Different spatial tunings for ON and
Science 213:1139-l 142. OFF pathway stimulation. Ophthalmol Physiol Opt 12:233-240.
Lynch JJ, Silveira LCL, Perry VH, Merigan WH (1992) Visual effects Vaney DI (1985) The morphology and topographic distribution of AI1
of damage to P ganglion cells in macaques. Vis Neurosci 8:575-583. amacrine cells in the cat retina. Proc R Sot Lond [Biol] 224:475-488.
Martin PR, Griinert U (1992) Spatial density and immunoreactivity Vaney DI (1986) Morphological identification of serotonin-accumu-
of bipolar cells in the macaque monkey retina. J Comp Neurol 323: lating neurons in the living retina. Science 233:444-446.
269-287. Wassle H, Boycott BB (1991) Functional architecture of the mam-
Merigan WH (1989) Chromatic and achromatic vision of macaques: malian retina. Physiol Rev 7 1:447480.
role of the P pathway. J Neurosci 9:776-783. Wlssle H, Riemann HJ (1978) The mosaic of nerve cells in the mam-
Merigan WH, Eskin TA (1986) Spatio-temporal vision of macaques malian retina. Proc R Sot Lond [Biol] 200:44 l-46 1.
with severe loss of P@ retinal ganglion cells. Vision Res 26: 175 l- Wiissle H, Peichl L, Boycott BB (198 la) Morphology and topography
1761. of on- and off-alpha cells in the cat retina. Proc R Sot Lond [Biol]
Merigan WH, Katz LM (1990) Spatial resolution across the macaque 212:157-175.
retina. Vision Res 30:985-99 1. Wlssle H, Boycott BB, Illing R-B (198 1b) Morphology and mosaic of
Merigan WH, Katz LM, Maunsell JHR (199 1) The effects of parvo- on- and off-beta cells in the cat retina and some functional consid-
cellular lateral geniculate lesions on the acuity and contrast sensitivity erations. Proc R Sot Lond [Biol] 2 12: 177-l 95.
of macaque monkeys. J Neurosci 11:994-l 00 1. Wassle H, Peichl L, Boycott BB (198 lc) Dendritic territories of cat
Milam AH, Dacey DM, Dizhoor AM (1993) Recoverin immuno- retinal aanalion cells. Nature 292:344-345.
reactivity in mammalian cone bipolar cells. Vis Neurosci 10: 1-12. Wassle H: Boycott BB, Rohrenbeck J (1989a) Horizontal cells in the
Peichl L (199 1) Alpha ganglion cells in mammalian retinae: common monkey retina: cone connections and dendritic network. Em J Neu-
properties, species differences, and some comments on other ganglion rosci 1:421-435.
cells. Vis Neurosci 7: 155-169. Wlssle H, Grtlnert U, Riihrenbeck J, Boycott BB (1989b) Cortical
Perry VH, Cowey A (1988) The lengths of the fibres of Henle in the magnification factor and the ganglion cell density of the primate ret-
retina of macaque monkeys: implications for vision. Neuroscience ina. Nature 341:643-646.
25~225-236. Wlssle H, Grunert U, Rijhrenbeck J, Boycott BB (1990) Retinal gan-
Perry VH, Oehler R, Cowey A (1984) Retinal ganglion cells that project glion cell density and cortical magnification factor in the primate.
to the dorsal lateral geniculate nucleus in the macaque monkey. Neu- Vision Res 30: 1897-19 11.
roscience 12:1101-l 123. Watanabe M, Rodieck RW (1989) Parasol and midget ganglion cells
Polyak SL (194 1) The retina. Chicago: University of Chicago. of the primate retina. J Comp Neurol 289:434-454.
Reid RC, Shapley RM (1992) Spatial structure of cone inputs to re- Wertheim T (1891) Peripheral visual acuity. Reprint (Dunsky IL,
ceptive fields in primate lateral geniculate nucleus. Nature 356:7 16- transl), Am J Optom Physiol Opt 57:9 15-924, 1980.
718. Wiesel TN, Hubel DH (1966) Spatial and chromatic interactions in
Rodieck RW, Binmoeller KF, Dineen J (1985) Parasol and midget the lateral geniculate body of the rhesus monkey. J Neurophysio129:
ganglion cells of the human retina. J Comp Neurol 233: 115-l 32. 1115-1156.
Schein SJ (1988) Anatomy of macaque fovea and spatial densities of Williams DR, Coletta NJ (1987) Cone spacing and the visual reso-
neurons in fovea1 representation. J Comp Neurol 269:479-505. lution limit. J Opt Sot Am 4: 15 14-1523.
Schein SJ, de Monasterio FM (1987) Mapping of retinal and geniculate Zemon V, Gordon J, Welch J (1988) Asymmetries in ON and OFF
neurons onto striate cortex of macaque. J Neurosci 7:996-1009. visual pathways of humans revealed using contrast-evoked cortical
Schiller P, Logothetis NK, Charles ER (1990) Functions of the colour- potentials. Vis Neurosci 1: 145-150.