The Journal of Neuroscience, July 2, 2008 • 28(27):6807– 6817 • 6807

Behavioral/Systems/Cognitive

A Retinal Circuit That Computes Object Motion

Stephen A. Baccus,1,2 Bence P. Ölveczky,1 Mihai Manu,2 and Markus Meister1
1Department of Molecular and Cellular Biology and Center for Brain Science, Harvard University, Cambridge, Massachusetts 02138, and 2Department of
Neurobiology, Stanford University School of Medicine, Stanford, California 94305

Certain ganglion cells in the retina respond sensitively to differential motion between the receptive field center and surround, as produced
by an object moving over the background, but are strongly suppressed by global image motion, as produced by the observer’s head or eye
movements. We investigated the circuit basis for this object motion sensitive (OMS) response by recording intracellularly from all classes
of retinal interneurons while simultaneously recording the spiking output of many ganglion cells. Fast, transient bipolar cells respond
linearly to motion in the receptive field center. The synaptic output from their terminals is rectified and then pooled by the OMS ganglion
cell. A type of polyaxonal amacrine cell is driven by motion in the surround, again via pooling of rectified inputs, but from a different set
of bipolar cell terminals. By direct intracellular current injection, we found that these polyaxonal amacrine cells selectively suppress the
synaptic input of OMS ganglion cells. A quantitative model of these circuit elements and their interactions explains how an important
visual computation is accomplished by retinal neurons and synapses.
Key words: neural coding; eye movements; motion processing; computational model; neural circuit; inhibition

Introduction move with a trajectory different from that in a large surrounding

The retina has about a dozen types of output neurons. Each type region. A neuron designed to detect and flag a moving object in
of retinal ganglion cell forms a complete array that covers the the retinal image should therefore have the following desirable
visual field (Masland, 2001). Every such population represents a properties: it should monitor a patch of image and fire when that
specific computation on the raw visual image, and transmits the patch moves differently from the surrounding region. Further-
resulting output image to various stations in the brain. To under- more, to allow the tagging of all kinds of objects, the neuron
stand the function of the retina, one would like to identify for should produce this response regardless of the visual pattern dis-
each ganglion cell type what image operation it reports. Then, played in its image patch or in the surrounding region. Finally, to
one needs to examine how that computation is performed by the accommodate all kinds of observer motion, the neuron’s re-
circuits of interneurons between photoreceptors and ganglion sponse should be independent of the direction of image motion.
cells. Recently we described a class of retinal ganglion cells in- Remarkably, these capabilities are found in certain types of reti-
volved in an ecologically important visual computation: the de- nal ganglion cells (Ölveczky et al., 2003), termed “object motion
tection of moving objects (Ölveczky et al., 2003). Here we report sensitive (OMS)” (see Fig. 1).
how this specific computation is accomplished through the inter- The OMS neurons meet two seemingly conflicting require-
actions of certain bipolar, amacrine, and ganglion cells. ments: they are highly tuned to a condition of differential motion
Image motion on the retina has two components: one, of between the receptive field center and the surround, but at the
course, is the movement of objects in the scene. The other is same time remarkably insensitive to the actual visual pattern in
observer-induced motion, resulting from translation of the ani- the center or the surround. Here we explore how this is accom-
mal, movements of the head, large gaze-shifting eye movements, plished, by recording directly from many interneurons in the
and the incessant small eye movements that persist even during retina. This approach allowed us to trace the flow of signals dur-
gaze fixations (Skavenski et al., 1979; Engbert and Kliegl, 2004). ing the differential motion computation, identify which specific
As a result of observer or eye motion, even a static scene moves retinal interneurons are involved, and verify their connectivity.
across the retina continuously, with a motion trajectory that is
nearly uniform over large parts of the image. An object moving Materials and Methods
relative to the background causes a patch of image on the retina to Electrophysiology. Simultaneous intracellular and multielectrode record-
ing was performed as described previously (Baccus and Meister, 2002).
The isolated retina of a tiger salamander was placed on a flat array of 61
Received Sept. 13, 2007; revised April 1, 2008; accepted April 24, 2008. extracellular electrodes and held in place under a layer of 0.6% agarose
This work was supported by a National Research Service Award postdoctoral fellowship (S.A.B.), a Harvard Junior (type III-A: High EEO; Sigma) of ⬃100 ␮m thickness with a dialysis
Fellowship (B.P.Ö.), and grants from the National Institutes of Health (S.A.B., M. Meister).
membrane containing several 150 –300 ␮m holes. Sharp intracellular
Correspondence should be addressed to Stephen A. Baccus, Department of Neurobiology, Stanford University
School of Medicine, 299 West Campus Drive, Stanford, CA 94305. E-mail: [email protected].
microelectrodes were positioned over the retina under infrared illumi-
B. P. Ölveczky’s present address: Department of Organismic and Evolutionary Biology and Center for Brain Sci- nation, viewed through an infrared-sensitive CCD camera, and guided
ence, Harvard University, 16 Divinity Avenue, Cambridge, MA 02138. through the dialysis membrane and agarose into the retina. Electrodes
DOI:10.1523/JNEUROSCI.4206-07.2008 were filled with 2 M potassium acetate and 1% Alexa Fluor 488, with a
Copyright © 2008 Society for Neuroscience 0270-6474/08/286807-11$15.00/0 final impedance of 150 –250 M⍀.
6808 • J. Neurosci., July 2, 2008 • 28(27):6807– 6817 Baccus et al. • A Retinal Circuit That Computes Object Motion

Intracellular recordings were made from a total of 22 bipolar cells where F(x, y, ␶) is the linear response kernel of the bipolar cell at position
(resting membrane potential ranging between ⫺45 and ⫺55 mV), 4 (x, y) and delay ␶; s(x, y, t) is the stimulus intensity at position (x, y) and
horizontal cells (⫺40 to ⫺50 mV), 12 OMS ganglion cells (⫺60 to ⫺70 time t, normalized to zero mean and unit variance; b(t) is the membrane
mV), 21 polyaxonal amacrine cells (⫺55 to ⫺65 mV), and 14 amacrine potential of the bipolar cell; and T is the duration of the recording. To
cells of other types, including slow ON–OFF cells, ON cells, and OFF cells predict the bipolar cell response to a jittering grating, the spatiotemporal
(⫺50 to ⫺65 mV). Not all stimuli were presented to all cells. The soma of filter, F(x, y, ␶), was convolved with the stimulus, yielding the predicted
each neuron was situated in the center of the “object region” of the response
stimulus, which was more than twice as large as the receptive field centers
of the bipolar, amacrine, and ganglion cells from which we recorded. A T

cell was identified during the experiment by its response to flashes and
global motion and differential motion stimuli. After the recording ses-
b⬘ 共 t 兲 ⫽ 冕 s共 x,y,␶兲␣F共 x,y,t ⫺ ␶兲dxdyd␶, (2)
o
sion, the neuron was filled with dye iontophoretically, and the cell type
was confirmed by imaging in the live preparation using a 40⫻ water- where ␣ is a scaling factor set to yield the best prediction. To predict the
immersion objective. Selected neurons were micrographed using a CCD OMS ganglion cell response from measured bipolar cell responses, we
camera and their processes traced from three-dimensional image stacks. postulated a simple nonlinear transformation followed by integration
Cell nomenclature. The five major classes of retinal neurons can be over many bipolar cells (see Fig. 4 D):
recognized unambiguously. Within each class there exists a wide range of
cell shapes and physiological properties, and it is thought that these re-
flect discrete genetic types of neuron. In the salamander retina, the ana-
g⬘ 共 t 兲 ⫽ 冘␾
N 共 b ␾ 共 t 兲兲 , (3)

tomical and physiological properties of these cells tend to form a contin-

uum rather than breaking naturally into discrete clusters. Therefore the where g⬘(t) is the predicted membrane potential of the ganglion cell;
type designations can vary according to the methods applied. For the b␾(t) is the membrane potential of a bipolar cell at phase ␾ relative to the
major neurons involved in the circuit described here, we provide ana- grating; and N(b) is a nonlinear transformation of the bipolar cell output.
tomical tracings, receptive fields, and light responses to standard stimuli Bipolar cell responses b␾(t) were recorded using at least four equally
(see Figs. 2, 3, 6) by which they may be recognized. In addition we give spaced phases of the grating. The nonlinearity N(b) consisted of a piece-
here some correspondence to type designations used previously. The wise linear fit through five points, adjusted so that the prediction g⬘(t)
OMS ganglion cells are the “fast OFF” cell of Warland et al. (1997), the most closely approximated the measured ganglion cell response.
“biphasic OFF” cell of Segev et al. (2006), and resemble the “small highly To predict the OMS ganglion cell response to local motion (see Fig. 5)
complex” morphological type of Costa and Velte (1999). The OFF bipo- or the polyaxonal amacrine response to background motion (see Fig. 9)
directly from the stimulus, the bipolar cell responses were predicted for
lar cells with rapid kinetics (see Figs. 2– 4) resemble “type 5” of Wu et al.
each spatial phase, passed through the nonlinearity, and then summed:
(2000). The polyaxonal amacrine cells identified in this study are distinct
from other amacrines by having a field of axons that spreads laterally over
several millimeters, whereas the dendrites and consequently the receptive
g⬘ 共 t 兲 ⫽ 冘
␾
N 共 b⬘␾ 共 t 兲兲
field are constrained to a radius of ⬃200 ␮m (see Fig. 2) (see also Ölvec-

冢冕 冣
zky et al., 2003). Their light response is OFF dominated. Their correspon-
冘
T

dence to the literature is uncertain; many reports mention wide-field

⫽ N s 共 x,y, ␶ 兲 ␣ F ␾ 共 x,y,t ⫺ ␶ 兲 dxd yd ␶ . (4)
amacrine cells in the salamander retina, but their images are mostly from
␾
vertical sections, which obscures the distinctive morphology. 0

Visual stimulation. Visual stimuli were projected onto the retina from
Judging from the receptive field sizes (see Fig. 2), an OMS ganglion cell
a video monitor, at a photopic mean intensity of ⬃8 mW/m 2. The cir-
receives input from at least 20 bipolar cells. For the simulation of Equa-
cular object region was 800 ␮m in diameter, and the background region
tion 4, we therefore replicated the measured bipolar cell receptive field
measured 5900 ⫻ 4400 ␮m. The two regions were separated by a 92 ␮m
F␾(x, y, t ⫺ ␶) at 20 different spatial phases relative to the grating. Al-
gray annulus, although this annulus is not required for differential mo-
though the actual number of contributing bipolar cells may well be
tion selectivity (Ölveczky et al., 2003). The eye movement during fixa- larger, a finer phase spacing makes little difference for the predicted OMS
tional drift resembles a random walk (Skavenski et al., 1979; Engbert and response.
Kliegl, 2004), with horizontal and vertical motion occurring indepen- To characterize the speed tuning of a bipolar cell, we estimated its
dently. Statistics of these eye movements are qualitatively similar across linear response to a grating stimulus (184 ␮m period) moving at constant
species, including humans and salamander (Manteuffel et al., 1977; En- speed, by filtering the stimulus with the cell’s spatiotemporal receptive
gbert and Kliegl, 2004). To approximate the trajectory of fixational drift, field (Eq. 2). This generated a periodic output, whose amplitude was
grating stimuli consisting of black and white bars with a periodicity of taken as the cell’s sensitivity at that speed. The high-speed cutoff was
184 ␮m were jittered in one dimension. The trajectory was generated by defined as the speed at which the sensitivity fell to 90% of its peak value.
stepping the grating randomly every 15 ms with a step size of 9.2 ␮m. In Fast, transient OFF-type bipolar cells were identified from their receptive
the global motion condition, the seeds for generating the random walk fields as those cells with a speed tuning cutoff of ⬎300 ␮m/s for a 184 ␮m
were the same for the object and background regions, producing the period grating.
same motion trajectory in both regions. In the differential motion con- To predict the OMS ganglion cell response to differential motion, the
dition, the seeds were different for the object and background regions. In amacrine response was used to reduce the gain of the input to the recti-
the local motion condition, the background region was gray. To map fication stage in the ganglion cell receptive field center,
receptive fields, the stimulus was a randomly flickering checkerboard,

冢 冣
with square fields 18 –92 ␮m in width, each modulated independently by T

white noise (Meister et al., 1994).

Analysis and simulation. The spatiotemporal receptive field of a bipolar
冕 s共 x,y,␶兲␣F 共 x,y,t ⫺ ␶兲dxd yd␶
␾

cell was computed from the response to random checkerboard stimula-

tion by the standard method of reverse correlation:
g⬘ 共 t 兲 ⫽ 冘␾
N
0

1 ⫹ ␤ a⬘ 共 t 兲
, (5)

T
where a⬘(t) is the amacrine cell potential scaled to range between a value
F 共 x,y, ␶ 兲 ⫽ 冕 s共 x,y,t ⫺ ␶兲b共t兲dt, (1)
of zero and one, and ␤ is an optimized gain factor. This best value of ␤
was found to be 1.04, meaning that at times of greatest inhibition, the
0 gain of the bipolar subunit input was reduced by a factor of 0.49. This
Baccus et al. • A Retinal Circuit That Computes Object Motion J. Neurosci., July 2, 2008 • 28(27):6807– 6817 • 6809

c共␶兲 ⫽ 冕 a共t兲 g共t ⫹ ␶兲dt. (7)

0

The time of the peak of c(␶) was taken as the

average delay between the two cells.
The receptive fields of amacrine cells were
computed from random checkerboard stimula-
tion as for bipolar cells (Eq. 1). For OMS gan-
glion cells, we used the spike train as the re-
sponse variable instead of the membrane
potential. The receptive field profiles in Figure 2
show the response kernel F(x, y, ␶) at a time ␶
near the peak amplitude.
Averaged traces are computed across trials of
identical stimuli.

Results
We presented to the isolated salamander
retina visual stimuli with different combi-
nations of object and background motion
(Fig. 1). The display was divided into a cen-
tral circular object region and a surround-
ing background region. Stripe gratings in
each region were jittered in a random walk
with the characteristics of drifting fixa-
tional eye movements (Manteuffel et al.,
1977; Engbert and Kliegl, 2004), but lim-
ited to one dimension for simplicity. The
jitter trajectories were chosen to produce
three motion conditions (Fig. 1 B–D):
global motion simulated the image of a
static scene scanned over the retina by fixa-
Figure 1. Stimuli and working model for studying the OMS circuit. A, Diagram of object and background regions in the tional eye movements; differential motion
stimulus display. B–D, First row, Space–time plot of a vertical cross section through the center of the stimulus (line in A), showing simulated an object moving over a static
trajectories for global motion, differential motion, and local motion. Second row, Firing rate of a sample OMS ganglion cell in background and scanned by eye move-
response to 10 repeats of each stimulus sequence. Third row (D), Response to a local motion stimulus having the same trajectory ments; local motion was an artificial stim-
but reversed grating phase (180°), with black and white bars exchanged relative to 0° phase. E, Working model of the OMS circuit. ulus used to isolate the effects of the central
An OMS ganglion cell (G) receives excitatory input in the object region from multiple small subunits. Each subunit applies a linear object region. The motion trajectory of the
spatiotemporal filter to the stimulus in its receptive field. The result is then rectified and summed with the output of other
object was identical in all three cases, and
subunits. An inhibitory amacrine cell (A) in the background region receives input from a similar set of rectified subunits. The
output of the amacrine cell then inhibits the ganglion cell. Both inhibition and excitation are temporally sparse. Traces show the
we recorded from OMS ganglion cells
excitatory and inhibitory components of the model’s response to global motion. Numbers identify the key circuit elements to be whose receptive field centers were con-
identified, as listed in the text. tained in this object region. These neurons
respond robustly to differential or local
motion, but are nearly silent during global
divisive interaction of amacrine and bipolar input yielded a better fit than motion (Fig. 1 B–D). Motion in the background has little effect
a subtractive interaction. on the response to the object (Fig. 1, differential vs local), unless
To compute correlation coefficients between actual and model mem- its trajectory matches that of the object (Fig. 1, differential vs
brane potential responses, spikes were removed by setting a threshold for
global), in which case it suppresses firing almost completely.
the derivative of the membrane potential, and then each response was
smoothed with a box filter of width 20 ms. The correlation coefficient r Thus, the circuits of the retina must somehow compare the mo-
between two responses x(t) and y(t) was then calculated as tion trajectories in the object and background regions, and the
main challenge lies in finding how this is implemented.
T

冕 共 x共t兲 ⫺ 具x典兲共 y共t兲 ⫺ 具y典兲dt Working model of object motion circuitry

In previous work (Ölveczky et al., 2003), the spiking response of

冑 冑
0

r⫽ , (6) OMS ganglion cells has been described by a computational model

T T

冕 共 x共t兲 ⫺ 具x典兲 dt 冕 共 y共t兲 ⫺ 具y典兲 dt

2 2
with several layers (Fig. 1 E). The stimulus is first processed by
linear subunits with a small receptive field and transient dynam-
0 0 ics. In the next layer, the output from these subunits is strongly
where T is the duration of the experiment, and rectified and then summed within the object and within the back-
冓. . .冔 denotes the time average. ground region. Finally, the output from the background inhibits
To measure the delay between amacrine and ganglion cell responses, the response to the object region.
we computed the cross-correlation c(␶) between two responses a(t) and OMS ganglion cells respond with nearly the same firing se-
g(t) as follows: quence regardless of the specific spatial pattern of the moving
6810 • J. Neurosci., July 2, 2008 • 28(27):6807– 6817 Baccus et al. • A Retinal Circuit That Computes Object Motion

Figure 2. Receptive fields and morphology of relevant neurons. A, Top, Spatial profile of the receptive field for two OFF bipolar cells; the same scale bar applies to receptive fields in B and C.
Bottom, Tracing of an OFF bipolar cell in tangential and radial views. Dotted lines indicate borders of the inner plexiform layer. The axon terminals ramify in the outer (OFF) sublamina. B, Top,
Receptive field of an OMS ganglion cell. Bottom, Tracing in tangential view. C, Top, Receptive field of a polyaxonal amacrine cell. Bottom, Tracing in tangential view showing ON dendrites (red), OFF
dendrites (blue), and axons (black). Dashed lines indicate missing image information. The axons were followed beyond the pictured region and extended ⬎3 mm from the soma. Receptive fields
and tracings are from different cells.

object. An example of this insensitivity to visual pattern can be mammalian ganglion cells, it has been proposed that these recti-
observed by reversing the phase of the grating; swapping the black fied subunits correspond to individual bipolar cells (Victor and
and the white bars has almost no effect on firing (Fig. 1 D). Sim- Shapley, 1979; Demb et al., 2001).
ilarly, changing the period of the grating leaves the firing pattern
virtually unchanged (Ölveczky et al., 2003). Such invariance to The role of bipolar cells
the spatial pattern of the stimulus can be understood as a direct To identify directly the subunits of the OMS ganglion cell, we
consequence of the postulated summation over rectified subunits recorded intracellular light responses from bipolar cells. First, we
(Shapley and Victor, 1979; Ölveczky et al., 2003). mapped each cell’s spatiotemporal receptive field (see Materials
The second important computational property of the OMS and Methods). Bipolar cell spatial receptive fields measured 50 –
circuit, selectivity for differential motion, arises in this model 150 ␮m in diameter (Fig. 2 A). They were considerably smaller
from the sparse nature of the neural signals (Fig. 1 E). Both the than those of OMS ganglion cells (⬃400 ␮m) (Fig. 2 B) or of the
excitation from the object region and the inhibition from the polyaxonal amacrine cells (⬃400 ␮m) (Fig. 2C), interneurons
background region are delivered to the ganglion cell in a sparse that might mediate long-range inhibition (Ölveczky et al., 2003).
sequence of transient pulses. If the background trajectory In both the receptive field center and the antagonistic surround
matches the object trajectory, then inhibition arrives coincident region, the bipolar cell’s temporal filter followed a biphasic time
with excitation, silencing the ganglion cell. However, when the course (Fig. 3A). The surround response was delayed and in-
two trajectories are different, inhibition and excitation arrive at verted in sign relative to the center.
different times, and the ganglion cell fires. When stimulated with a jittering grating, bipolar cells pro-
Although this block diagram describes the response of the duced strong fluctuations in membrane potential, whose time
OMS ganglion cell accurately, it does not specify what neuronal course could be predicted accurately by passing the visual stimu-
circuit implements the computation. To test whether the retina lus through each cell’s measured spatiotemporal receptive field
indeed uses this algorithm and to flesh out the schematic with (Fig. 3B). The correlation coefficient between actual and pre-
actual neural elements, one needs to answer the following ques- dicted responses (see Materials and Methods) was r ⫽ 0.70 ⫾ 0.02
tions (Fig. 1 E): (1) What is the cellular identity of the subunits? (4 cells), nearly as large as the correlation between repeats of the
(2) How is the output of these subunits integrated? (3) At what identical stimulus (r ⫽ 0.78 ⫾ 0.05). Thus, in response to object
level is the signal from background motion combined with that motion that can be discriminated by OMS ganglion cells, these
from the object region? (4) What is the identity of the inhibitory bipolar cells essentially applied a linear spatiotemporal filter to
cell that reports on the background motion? the stimulus. In particular, the voltage response changed sign
when the phase of the jittering grating was reversed (Fig. 3B),
Basis of insensitivity to spatial pattern unlike what happens in OMS ganglion cells (Fig. 1 D). Clearly the
A key feature of the working model in Figure 1 E is the summation characteristic of pattern invariance is not yet elaborated at the
over small receptive field subunits. Based on recordings from level of bipolar cells.
Baccus et al. • A Retinal Circuit That Computes Object Motion J. Neurosci., July 2, 2008 • 28(27):6807– 6817 • 6811

We further investigated whether the

transformation from bipolar response to
ganglion cell response conformed to the
rules of the OMS working model (Fig. 1 E).
We obtained bipolar cell responses at mul-
tiple positions relative to the jittering grat-
ing by recording from a single cell while
repeating the grating’s motion trajectory at
different spatial positions, or phases (Fig.
4 A). To predict the response of the OMS
ganglion cell, these different bipolar cell re-
sponses were then individually rectified
and summed (Fig. 4 B). The shape of the
rectifier function was calculated to opti-
mally match the observed ganglion cell re-
sponse (Fig. 4C). This resulted in a steep
nonlinearity (Fig. 4 D). The resulting sum
of rectified bipolar signals closely approxi-
mated the intracellular responses recorded
from OMS ganglion cells (Fig. 4C, see also
Fig. 5).
Figure 3. Bipolar cells encode the spatial pattern. A, Bipolar cell spatiotemporal receptive field. Left, Spatial profile of a bipolar
The OMS ganglion cell response was
cell receptive field with OFF center and ON surround. Right, Temporal profile of the receptive field, normalized to the peak found to be invariant to the spatial phase of
sensitivity, and summed over all pixels (total) or only pixels in the center or the surround. B, Bipolar cell response to a jittering local the grating stimulus, not only at the level of
motion stimulus. Top, Traces from two identical presentations of the same trajectory. Bottom, The same trajectory with the spikes, but in its detailed subthreshold time
grating reversed (180° spatial phase shift). Red line shows a prediction of the bipolar cell response generated from Equation 2, course (Fig. 4C). The sum of rectified bipo-
using only the measured spatiotemporal receptive field and the stimulus trajectory. lar cell responses predicted many of those
subthreshold fluctuations correctly. Given
that the bipolar cell response still varies linearly with the stimulus
(Fig. 4 A), this rectification must happen after the bipolar cell
soma, presumably in the process of transmission to the ganglion
cell. Although the mechanism for rectification is unknown, it
may involve the voltage dependence of calcium influx at the bi-
polar cell presynaptic terminal (Heidelberger and Matthews,
1992). If the mechanism is postsynaptic, it must be tightly local-
ized in the ganglion cell dendrite, before many bipolar cell signals
are summed across the dendritic tree.

A refined subunit model

By substituting concrete bipolar cells for the theoretical subunits,
one obtains a refined model of the OMS ganglion cell response
(Fig. 5) in which all parameters are derived from measurements.
The stimulus is passed through the bipolar cell spatiotemporal
receptive field, yielding predictions of responses in the popula-
tion of bipolar cells. These responses are then rectified according
to the measured bipolar-to-ganglion cell transformation, and the
result is summed within the ganglion cell (Fig. 5A). This model
matches the time course of the observed OMS ganglion cell re-
sponse with remarkable accuracy (Fig. 5B): the predicted and
observed response had a correlation coefficient of r ⫽ 0.65 (see
Materials and Methods), similar to the correlations between
measured responses on repeats of an identical stimulus (r ⫽
0.71 ⫾ 0.02).
Not all bipolar cells produced a subunit model with this de-
gree of accuracy. Indeed, the bipolar cells in our sample varied
Figure 4. The transformation from bipolar cells to OMS ganglion cell involves rectification substantially in their spatiotemporal receptive fields. To charac-
and summation. A, Intracellular recordings from a single bipolar cell responding to a jittering terize the receptive field in the context of moving grating stimuli,
local motion stimulus, with the grating positioned at four different phases. The top plot in-
we computed for each bipolar cell the tuning curve for the speed
cludes two traces for identical stimuli to illustrate reproducibility of the response. B, The
measured bipolar cell responses were each rectified (see D) and then summed to yield a of a steadily moving grating (see Materials and Methods). These
prediction for the ganglion cell response. C, Measured response of an OMS ganglion cell to four tuning curves varied greatly in shape (Fig. 5C); in particular, the
phases of a jittering local motion stimulus. D, The nonlinear function used to rectify the bipolar highest speed a neuron could track ranged from ⬍100 ␮m/s to
output. This shape was chosen to optimize the fit between the trace in B and the traces in C, ⬎400 ␮m/s. We inserted each of these response types into the
ignoring the action potentials. Vertical lines indicate action potentials of the ganglion cell. model of Figure 5A, and the resulting predictions for the OMS
6812 • J. Neurosci., July 2, 2008 • 28(27):6807– 6817 Baccus et al. • A Retinal Circuit That Computes Object Motion

ate to predict the OMS ganglion cell re-

sponse (Fig. 5C,D). Although OMS
ganglion cells in the salamander are domi-
nated by OFF-type input (Ölveczky et al.,
2003), they also receive some input from
the ON pathway (Geffen et al., 2007). Thus,
ON bipolar cells may well contribute to the
response of OMS ganglion cells.

Suppression of responses by
background motion
To measure how object and background
motion are integrated at the level of gan-
glion cell synaptic input, we recorded
membrane potential responses from an
OMS ganglion cell while presenting either
differential motion or global motion (Fig.
6). As observed previously, spiking was al-
most completely suppressed by global mo-
tion. In addition, the subthreshold mem-
brane potential fluctuations were smaller
under global motion, decreasing to a frac-
tion of 0.45 ⫾ 0.08 (5 cells) of their value
during differential motion (Fig. 6 B). How-
ever, there were no large IPSPs as might
result from strong hyperpolarizing inhibi-
tion, and the decrease in response was con-
sistent with either shunting or presynaptic
inhibition. OMS ganglion cell responses
were very homogeneous, in that each cell’s
intracellular recording was highly corre-
lated to the average OMS cell response (r ⫽
0.78 ⫾ 0.03, 7 cells).

The effects of background motion

on interneurons
According to the working model of Figure
1 E, when object and background regions
experience the same motion trajectory, the
ganglion cell receives inhibitory input from
amacrine cells in the background region
that is synchronous with excitatory input
Figure 5. Model of the OMS ganglion cell excitatory input. A, Detailed structure of a subunit based on bipolar cell measure- from the central object region. Thus, we
ments. The stimulus trajectory of the object region during local motion is filtered by the bipolar cell spatiotemporal receptive field, inspected the responses of various inhibi-
yielding the bipolar cell response. Each bipolar response is then passed through the nonlinear transfer function from Figure 4 D tory interneurons to a jittering back-
and summed, yielding the ganglion cell membrane potential. B, Comparison of ganglion cell membrane potential and model
ground, and compared their timing to the
output. C, Speed tuning curves of bipolar cells to a moving 184 ␮m period grating (see Materials and Methods). Each trace shows
the normalized sensitivity from a different bipolar cell, grouped into four types of tuning profile. Cells with faster responses are at
responses of OMS ganglion cells under lo-
right. D, Each bipolar cell receptive field was used in a separate model of the OMS response (A). The quality of the model fit was cal motion with the same trajectory (Fig.
measured by the correlation coefficient between predicted and observed OMS ganglion cell response (B). The high speed cutoff 6C–H ). Among these recordings, the re-
of the bipolar cell was measured by the speed at which the sensitivity falls to 90% of the peak value (C). Here, the model quality sponse of a polyaxonal amacrine cell type
is plotted against the high speed cutoff. The dotted line indicates the correlation coefficient calculated between ganglion cell was appropriately timed to mediate sup-
responses to repeated jittering stimuli. pression from the background region: its
depolarizations were closely synchronized
response varied greatly in accuracy: from r ⫽ ⫺0.25 to r ⫽ 0.65. to those of the OMS ganglion cell (Fig. 6 D). The correlation
A bipolar cell’s upper speed cutoff, in turn, was a very sensitive coefficient between a polyaxonal amacrine cell and an OMS gan-
indicator of whether it would generate a useful model (Fig. 5D). glion cell responding to the same stimulus trajectory was 0.63 ⫾
Those cells sensitive to the highest speeds (cutoff above ⬃300 0.03 (7 cells), compared with 0.71 for repeats of an identical
␮m/s) produced an accurate prediction (r ⫽ 0.61 ⫾ 0.02, 4 cells), stimulus to a ganglion cell. Like OMS ganglion cells, polyaxonal
whereas the others did not (r ⫽ 0.1 ⫾ 0.09, 7 cells). This suggests amacrine cells were also very homogeneous; each cell’s response
that the OMS circuitry makes selective use of the fastest bipolar was very similar to the average response (r ⫽ 0.65 ⫾ 0.01, 7 cells).
cell types. In contrast, a different amacrine cell type, characterized by a slow
Interestingly, in addition to OFF-type bipolar cells, an ON- ON–OFF response, produced depolarizations at the wrong times
type bipolar cell was encountered with a receptive field appropri- (Fig. 6 E). Its signal had a correlation coefficient of only 0.13 ⫾
Baccus et al. • A Retinal Circuit That Computes Object Motion J. Neurosci., July 2, 2008 • 28(27):6807– 6817 • 6813

will refer to these amacrines simply as “polyaxonal,” bearing in

mind that they are also identified by the time course of their
motion response.
To further understand the interaction of signals from bipolar,
amacrine, and ganglion cells, we used a different stimulus that
clearly separates the effects of the center and surround regions
(Fig. 7). Here the object grating and the background grating each
moved back and forth in short 40 ␮m steps every 1 s. In the global
motion condition, the two gratings stepped synchronously,
whereas under differential motion they stepped in alternation
(Fig. 7A). This stimulus distills the essential ingredients of the
random jitter stimulus: brief periods of acceleration whose tim-
ing was either identical or different between object and back-
ground region. We recorded the responses of bipolar, OMS gan-
glion, and polyaxonal amacrine cells in the region of the object
grating.
OMS ganglion cells produced a transient depolarization on
every step of the object grating (Fig. 7B). Under global motion,
this depolarization was considerably smaller than under differen-
tial motion, where the depolarizations led to spikes (amplitude
ratio global/differential ⫽ 0.52 ⫾ 0.03, 4 cells). Steps of the back-
ground region alone caused virtually no response, either depolar-
izing or hyperpolarizing (amplitude ratio background/object ⫽
0.04 ⫾ 0.05, 4 cells).
We recorded from fast, transient OFF bipolar cells that yielded
an accurate prediction of the OMS ganglion cell response when
inserted into the model of Figure 5A (see Materials and Meth-
ods). The dynamics of these bipolar cells resemble those de-
Figure 6. Signals of inhibitory interneurons during the OMS response. A, B, Intracellular scribed previously (Wu et al., 2000) for cells that arborize in the
recording from an OMS ganglion cell responding to differential motion or global motion. C–H, OFF sublamina near the middle of the inner plexiform layer (Fig.
Response of an OMS ganglion cell and a panel of different inhibitory interneurons to the same 2 A). Such a bipolar cell depolarized when a dark bar of the object
motion trajectory: local motion for the OMS ganglion cell, and global motion for the inhibitory grating stepped into its receptive field center (Fig. 7C). These
interneurons (different trajectory from panels A and B). Vertical lines indicate action potentials
depolarizations had very similar magnitude under global and
of the OMS ganglion cell. Right, Response of the same neuron to a periodic stimulus, used to
characterize the cell type: a contrast-reversing grating in D and E, and a uniform field flash in
differential motion (ratio global/differential ⫽ 0.93 ⫾ 0.02, 7
F–H. I, Expanded time scale comparing the polyaxonal amacrine cell response to global motion cells). This shows that bipolar cell excitation, as measured at the
with the OMS ganglion cell response to local motion. AC, Amacrine cell; GC, ganglion cell; HC, soma, is not suppressed by global motion. On the reverse step of
horizontal cell. the object grating, bipolar cells hyperpolarized, with a somewhat
slower time course than in the preceding depolarization. This is
likely a result of light adaptation, which makes for more sluggish
0.06 (6 cells) with the OMS cell depolarizations. Other amacrine responses at lower luminance (Naka et al., 1979). Finally, during
cell types (Fig. 6 F, G) also produced unrelated signals with corre- differential motion, bipolar cells incurred brief hyperpolariza-
lations ranging from ⫺0.10 to 0.29 (8 cells). Horizontal cells, of tions on every step of the background grating (ratio background/
which the salamander also has multiple types, showed virtually object ⫽ 0.34 ⫾ 0.05, 7 cells). This input was identical for steps of
no response to the jittering grating (Fig. 6 H) (4 cells), and thus either sign, and thus must derive from an inhibitory interneuron
are unlikely to contribute to suppression from the background with rectified responses. Horizontal cells are an unlikely source,
region. because they have largely linear properties and hardly respond at
Based on this survey, the polyaxonal amacrine cell (Fig. 6 D) all to these fine gratings (Fig. 6H). Alternatively, this signal may
appears to be a plausible candidate to transmit inhibition from derive from amacrine cells that inhibit the bipolar cell synaptic
the background region. Indeed the detailed timing of its depolar- terminal, close to the site of transmission but at some electrotonic
izations would enable an effective suppression of excitatory input distance from the soma. This could help explain why the hyper-
to the OMS ganglion cell (Fig. 6 I). By computing the cross- polarizing signals in somatic recordings are small (Fig. 7C), espe-
correlation between the polyaxonal amacrine cell response to cially in the global motion condition, whereas the synaptic input
global motion and the OMS ganglion cell response to local mo- to ganglion cells is strongly suppressed (Fig. 7B).
tion of the same trajectory, we found that the amacrine cell pre- A typical polyaxonal amacrine cell with receptive field in the
ceded the ganglion cell by 26 ⫾ 7 ms (6 cell pairs; see Materials object region was strongly depolarized by movement of the object
and Methods). This fits well with previous results indicating that grating (Fig. 7D). Surprisingly, these amacrine cells showed a
the polyaxonal amacrine cell membrane potential precedes the substantial depolarizing response also to movement in the back-
suppression of OMS ganglion cell spiking by ⬃25 ms (Ölveczky ground region (amplitude ratio background/object ⫽ 0.45 ⫾
et al., 2003). Its broad output arborization would allow for rapid 0.08, 12 cells). The linear receptive field of these cells is relatively
inhibition over long distances. Judging from experience in other small, consistent with a dendritic field of ⬃200 ␮m radius (Fig.
species, there may well exist multiple kinds of amacrine cells with 2C), and is contained entirely within the object region. In addi-
polyaxonal morphology, but in our survey so far all such ama- tion, however, polyaxonal amacrine cells show extensive tracer
crines were correlated to the OMS ganglion cells. For brevity, we coupling, at least in mammalian retinas (Völgyi et al., 2001;
6814 • J. Neurosci., July 2, 2008 • 28(27):6807– 6817 Baccus et al. • A Retinal Circuit That Computes Object Motion

Given the fine laminar organization of

the inner plexiform layer, one expects a pri-
ori that any given ganglion cell type should
interact with only a small fraction of ama-
crine cells. Thus the positive finding of a
connection between OMS ganglion cells
and polyaxonal amacrines is significant.
Nevertheless, we tested whether OMS gan-
glion cells are somehow particularly pro-
miscuous, by injecting current into a dif-
ferent type of amacrine with slow ON–OFF
responses (Fig. 6 E). This amacrine cell type
inhibited certain non-OMS cells but had
no effect on OMS cells (Fig. 8 D, E). This
demonstrates that the OMS ganglion cell
derives its inhibitory input with some se-
lectivity from polyaxonal amacrine cells.
Vice versa, the polyaxonal amacrine cell
targets its inhibitory output selectively to
OMS ganglion cells, perhaps directly or via
intervening neurons. This reinforces the
precision of wiring in the inner plexiform
layer, where many types of amacrine and
bipolar cells are connected in specific sub-
circuits to a dozen types of ganglion cells
(Roska and Werblin, 2001).

Model of object motion sensitivity

These results were assembled into a full
model of differential motion sensitivity
(Fig. 9). To predict the response of polyax-
Figure 7. Convergence of signals from object and background. Periodic jitter stimuli were composed of an object and back-
ground grating shifting back and forth 40 ␮m every 1 s. A, Space–time plot as in Figure 1 B. The two gratings were shifted in
onal amacrine cells in the background re-
synchrony for global motion and in alternation for differential motion. B–D, Membrane potential responses to periodic jitter. First gion, an identical subunit model was used
row, Global motion. Second row, Differential motion. Third row, Average of 15 responses under each condition. B, OMS ganglion for the excitatory input of OMS ganglion
cell. The third row shows the average of the subthreshold potential after spikes were removed. C, Fast, transient OFF-type bipolar cells (Fig. 5). As expected from the obser-
cell. D, Polyaxonal amacrine cell. All these neurons had a receptive field center in the object region. vation that polyaxonal amacrine cells and
OMS ganglion cells have synchronous re-
sponses to identical motion trajectories
Wright and Vaney, 2004). Thus it is possible that a weak excita- (Fig. 6C,D), this model predicted the amacrine response accu-
tory input arrives through electrical synapses from distant poly- rately (Fig. 9B), yielding a correlation between the actual and
axonal amacrine cells, whose contribution is strong under global predicted response of r ⫽ 0.73.
motion, when the population in the entire background region is To predict the response of an OMS ganglion cell to differential
driven in synchrony. Under differential motion, only the ama- motion, amacrine inhibition from the background was com-
crines in the object region are synchronized with the OMS cell, bined with bipolar input from the object region before the stage
and their action alone is clearly not sufficient to silence OMS cell of rectification, representing inhibition as observed at the presyn-
firing (Fig. 7B, bottom). aptic terminal (Fig. 7C). We used a divisive form of inhibition, in
which the depolarization of the amacrine cell scales down the
The role of polyaxonal amacrine cells bipolar cell synaptic output (see Materials and Methods). This
The observed correlation of signals from polyaxonal amacrines model produced accurate predictions of the OMS cell’s response
and OMS cells is suggestive, but does not demonstrate a func- to differential motion (Fig. 9C). The best fits (r ⫽ 0.65) were
tional connection. To directly test for an inhibitory interaction, obtained when the strength of inhibition was chosen to scale the
we recorded from a polyaxonal amacrine cell intracellularly while bipolar output by at most a factor of 0.49. This corresponds well
simultaneously recording from the ganglion cell population with to the ⬃50% reduction in depolarization seen in the OMS gan-
a multielectrode array. A contrast-reversing grating was used to glion cell during global motion versus differential motion (Figs.
stimulate the ganglion cells. Simultaneously, we injected depolar- 6 A, B, 7B).
izing current into the amacrine cell and measured whether that
suppressed the ganglion cell’s light response. By this measure, Discussion
polyaxonal amacrine cells clearly inhibited the light response of This study provides a circuit explanation for a prominent com-
OMS ganglion cells (by 17% on average), but not of other types of putation in the retina: the detection of differential motion. Tran-
ganglion cells (Fig. 8 B,C). This effect was smaller than the com- sient responses, small spatial subunits, strong rectification, long
plete suppression that occurs during global motion, presumably axonal projections, and specific functional connections combine
because it involved injection of just a single polyaxonal amacrine to emphasize object motion while rejecting background motion
cell. resulting from eye movements. We subjected a basic working
Baccus et al. • A Retinal Circuit That Computes Object Motion J. Neurosci., July 2, 2008 • 28(27):6807– 6817 • 6815

synapse. (3) The inhibition is likely presynaptic on the bipolar

terminal. This implies that differential motion is computed one
neuron earlier and across a smaller spatial scale than previously
proposed (Ölveczky et al., 2003). (4) Among all the interneurons
tested, only the polyaxonal amacrine cells have the response
properties required to implement inhibition from global motion.
They integrate motion signals from beyond the dendritic tree,
perhaps via electrical coupling. Horizontal cells are not involved
in the computation. Furthermore, the polyaxonal amacrine cells
do indeed suppress the responses of OMS ganglion cells, but not
other ganglion cell types. In addition, we found the following: (5)
Because polyaxonal amacrine cells are not themselves sup-
pressed, they receive input from a different set of bipolar cell
terminals than the OMS cells. (6) Finally, a mathematical model
of the proposed circuit, when supplied with the measured signals
from interneurons, can predict the responses at each stage in the
circuit in a quantitative manner.

Summary of signal flow

An essential part of the computation is the nonlinear spatial sum-
mation over small subunits in the ganglion cell receptive field
center. Among all the interneurons sampled in our experiments,
bipolar cells are the only ones with receptive field properties that
match the requirements for this role. The retina contains many
kinds of bipolar cells, which differ both in morphology and in the
time course of the light response (Burkhardt and Fahey, 1998;
DeVries, 2000; Wu et al., 2000; Ghosh et al., 2004). The OMS
circuitry involves only certain of these neurons; in particular, the
dynamic properties of the OMS response are consistent only with
the fastest among the bipolar cells we sampled (Fig. 5).
Motion in the object region drives these bipolar cells, and their
outputs are rectified before summation by the ganglion cell. We
showed that this rectification occurs after the bipolar cell soma,
not at the photoreceptor– bipolar cell synapse, counter to a pro-
posal made for mammalian Y-cells (Demb et al., 2001).
Motion in the background region drives, among others, a
population of polyaxonal amacrine cells. The response of these
neurons matches almost perfectly that of the OMS ganglion cell
Figure 8. Polyaxonal amacrine cells selectively suppress OMS ganglion cells. A, Schematic to the same trajectory (Fig. 6). It appears likely, therefore, that the
diagram of experiment. Contrast reversal of a grating was used to visually stimulate the retina polyaxonal amacrines are driven by a similar network of rectified
either alone or synchronous with a depolarizing current pulse (500 pA, 0.5 s) delivered intracel- bipolar cell inputs (Fig. 10). In addition, these amacrine cells may
lularly to an amacrine cell. Spiking activity from OMS and non-OMS ganglion cells (G) was be coupled electrically, because they receive excitatory input from
recorded with a multielectrode array. Responses were analyzed from ganglion cells within 200 far beyond the dendritic field (Fig. 7). Furthermore, we showed
␮m of the amacrine cell (A). B, C, Results from current injection into a polyaxonal amacrine cell by direct single-neuron stimulation that polyaxonal amacrine
(AC). B, Left, Average firing rate over 30 trials of an OMS ganglion cell (GC) responding to the
cells are wired to suppress the visual response of OMS ganglion
visual stimulus with and without the current pulse. Trials for the two conditions were inter-
leaved. Right, Fractional inhibition of ganglion cell firing by the current pulse (mean ⫾ SEM, 10
cells (Fig. 8). In principle, this may occur through a direct inhib-
GCs, 3 ACs). C, Left, Same as B for a non-OMS OFF-type ganglion cell. Right, Fractional inhibition itory synapse or through presynaptic inhibition of a bipolar ter-
of non-OMS ganglion cells (13 GCs, 3 ACs). D, E, Results from current injection into a slow minal (Cook and McReynolds, 1998). Our observations favor the
ON–OFF amacrine cell, presented as in B and C. D, Left, OMS ganglion cell light response with latter (Fig. 10), because the inhibition can be detected in the
and without the current pulse. Right, Fractional inhibition of OMS ganglion cells (10 GCs, 3 ACs). bipolar cell signal, whereas ganglion cells show no overt inhibi-
E, Left, Same as D for a non-OMS OFF-type ganglion cell. Right, Fractional inhibition of non-OMS tory potentials (Fig. 7). Further support for presynaptic inhibi-
ganglion cells (10 GCs, 3 ACs). tion comes from a recent study on adaptation in the OMS re-
sponse (Ölveczky et al., 2007). We found this effect from
hypothesis of the circuit (Fig. 1 E) to many experimental tests, polyaxonal amacrines to be specific to the OMS ganglion cell type
and elaborated on this model as summarized in Figure 10. Prob- (Fig. 8). It is encouraging to find that a single amacrine cell has a
ing the various interneurons intracellularly fleshed out the com- measurable effect on a nearby ganglion cell, and paired record-
ponents of the circuit and uncovered new aspects of processing. ings of this type will be invaluable in unraveling the various sub-
The answers to the four questions posed at the outset are as circuits of the retina (Geffen et al., 2007). A further important
follows: (1) The small subunits are bipolar cells. Among bipolars, aspect of the polyaxonal amacrine network is the speed of prop-
there is a range of response kinetics, and only the fastest ones agation of axon impulses, to ensure that the inhibition can arrive
account for the function of this circuit. (2) The excitatory gan- fast enough to suppress excitatory input. It appears that propa-
glion cell input is rectified, whereas the bipolar cell response is gation and inhibitory synaptic transmission occur within ⬃25 ms
linear. This places the rectification at the bipolar-to-ganglion cell (Fig. 6 I) (Ölveczky et al., 2003).
6816 • J. Neurosci., July 2, 2008 • 28(27):6807– 6817 Baccus et al. • A Retinal Circuit That Computes Object Motion

This algorithm for computing differen-

tial motion owes its success to several spe-
cial constraints in the ecology of vision.
First, the global motion signal is driven by
fixational eye movements. The random-
walk nature of this image trajectory (Eng-
bert and Kliegl, 2004) includes frequent ac-
celerations, which, when filtered through
the biphasic temporal receptive field of bi-
polar cells, lead to a pulsatile output.
Transmission to the ganglion cell seems to
apply a high threshold (Fig. 5), resulting in
a sparse sequence of excitatory pulses,
which allows for a selective temporal com-
parison between the sequences derived
from the object and the background. Only
an object trajectory identical to that of the
background is silenced; others are encoded
faithfully, except for occasional coinci-
dences between eye motion and object mo-
tion. This would not be possible if the back-
ground image motion were smooth with
Figure 9. Model of object motion sensitivity. A, Amacrine and ganglion cells each sum the rectified output of many linear constant velocity.
bipolar cells. Amacrine inhibition is delivered to the output of the bipolar cell before rectification, for example, at the bipolar cell Second, the algorithm compares only
presynaptic terminal. The output of the amacrine cell scales the bipolar output by a factor ranging between ⬃0.5 and 1 (see the speed of motion of the object and the
Materials and Methods). B, Comparison of actual and model amacrine cell membrane potential response to background motion. background, not the direction. Indeed, we
C, Comparison of actual and model ganglion cell membrane potential response to differential motion. The motion trajectory is showed that the OMS ganglion cells are
different from that in B. duly suppressed even when object and
background move in opposite direction
but with the same instantaneous speed (Ölveczky et al., 2003).
During natural vision, a measurement of speed is sufficient to
distinguish eye from object velocities, because it is highly unlikely
that the eye and the object conspire to produce an image speed
matching that of the background.
Thus ecological circumstances permit a differential motion
algorithm that never computes the direction of motion. A parallel
can be found in the jamming avoidance response of wave-type
electric fish, whose algorithm computes the difference in the fre-
quency of two signals without ever encoding either frequency
(Heiligenberg, 1989). Again, a combination of ecological con-
straints makes this a viable approximation. However, it should be
noted that differential motion is computed again at later stages of
the visual system, for example in the bird’s optic tectum (Frost
and Nakayama, 1983) and in area MT of visual cortex (Born and
Tootell, 1992). Those computations are indeed selective for di-
rection, and furthermore, they work well with smooth trajecto-
ries of global movement. Presumably they serve a different func-
Figure 10. A circuit for object motion sensitivity. Proposed neural circuitry underlying the tion, for example, the analysis of optic flow generated by observer
OMS response, linking bipolar cells (B), polyaxonal amacrine cells (A), and OMS ganglion cells motion.
(G). The bipolar cells have an OFF-type transient response, and their synapses are rectifying. The
polyaxonal amacrine cells have a restricted dendritic field that pools excitation from many Selectivity and invariance
bipolars and are probably coupled electrically to more distant amacrines in the population. The The object motion-sensitive neurons of the retina show a high
OMS ganglion cell also pools over many bipolars, but their terminals are inhibited by the poly- selectivity for differential motion, and at the same time almost
axonal amacrine cells. Direct, shunting inhibition of the OMS ganglion cell may also exist. Note complete invariance with regard to the spatial pattern that is
that the bipolar cell synapses onto amacrines do not receive presynaptic inhibition.
moving. These are, of course, highly desirable properties for a
general detector of movement within the visual scene. From a
An algorithm for differential motion detection broader viewpoint, the combination of selectivity and invariance
At a basic level, the comparison of motion of the object and the is commonly regarded as a hallmark of sophisticated computa-
background is performed through the time domain (Fig. 1 E). tion in higher stages of the brain, for example, in position-
The motion trajectories of both the object and background are invariant face-selective neurons (Hung et al., 2005). It is instruc-
converted into a sequence of depolarizing pulses. Then the two tive to see how it comes about in the experimentally tractable
pulse sequences are compared. If the timing matches, the gan- circuits of the retina.
glion cell remains silent; otherwise, it fires. Pattern invariance is computed early in the circuit. It arises
Baccus et al. • A Retinal Circuit That Computes Object Motion J. Neurosci., July 2, 2008 • 28(27):6807– 6817 • 6817

already within the background region, before any comparison to Frost BJ, Nakayama K (1983) Single visual neurons code opposing motion
motion in the object region. Essentially, it is guaranteed by the independent of direction. Science 220:744 –745.
pooling over rectified bipolar cells within the dendritic field of a Geffen MN, de Vries SE, Meister M (2007) Retinal ganglion cells can rapidly
change polarity from Off to On. PLoS Biol 5:e65.
polyaxonal amacrine cell (Fig. 10). This nonlinear summation Ghosh KK, Bujan S, Haverkamp S, Feigenspan A, Wässle H (2004) Types of
makes the depolarizations of the amacrine cell largely indepen- bipolar cells in the mouse retina. J Comp Neurol 469:70 – 82.
dent of pattern phase or spatial frequency (Hochstein and Shap- Heidelberger R, Matthews G (1992) Calcium influx and calcium current in
ley, 1976; Ölveczky et al., 2003). The selectivity for differential single synaptic terminals of goldfish retinal bipolar neurons. J Physiol
motion arises one synapse later, when signals from the object and 447:235–256.
background regions converge, and the timing of excitation and Heiligenberg W (1989) Coding and processing of electrosensory informa-
inhibition are compared. This may occur at the OMS ganglion tion in gymnotiform fish. J Exp Biol 146:255–275.
cell, or more likely at the bipolar cell terminal (Fig. 10). Pattern Hochstein S, Shapley RM (1976) Linear and nonlinear spatial subunits in Y
cat retinal ganglion cells. J Physiol 262:265–284.
invariance within the object region is again accomplished by
Hubel DH, Wiesel TN (1962) Receptive fields, binocular interaction and
pooling within the dendritic field of the OMS ganglion cell. functional architecture in the cat’s visual cortex. J Physiol 160:106 –154.
This scheme differs significantly from the proposed algo- Hung CP, Kreiman G, Poggio T, DiCarlo JJ (2005) Fast readout of object
rithms in other neural systems, where selectivity is computed identity from macaque inferior temporal cortex. Science 310:863– 866.
first. For example, complex cells in the primary visual cortex are Konishi M (2003) Coding of auditory space. Annu Rev Neurosci 26:31–55.
quite selective for the orientation of a line, but not for the line’s Manteuffel G, Plasa L, Sommer TJ, Wess O (1977) Involuntary eye move-
position. In the classic Hubel–Wiesel model, it is assumed that ments in salamanders. Naturwissenschaften 64:533–534.
they function by pooling over many simple cells, each of which is Masland RH (2001) The fundamental plan of the retina. Nat Neurosci
4:877– 886.
selective for both orientation and position (Hubel and Wiesel,
Meister M, Pine J, Baylor DA (1994) Multi-neuronal signals from the retina:
1962). Similarly, the owl’s auditory midbrain contains neurons acquisition and analysis. J Neurosci Methods 51:95–106.
that are selective for the difference in latency of a sound at the two Naka KI, Chan RY, Yasui S (1979) Adaptation in catfish retina. J Neuro-
ears, but not for the frequency of the sound. It is thought that physiol 42:441– 454.
these neurons pool over many presynaptic cells, each of which is Ölveczky BP, Baccus SA, Meister M (2003) Segregation of object and back-
tuned for interaural time difference, but also selective for a cer- ground motion in the retina. Nature 423:401– 408.
tain frequency (Konishi, 2003). Our study shows that this is not Ölveczky BP, Baccus SA, Meister M (2007) Retinal adaptation to object mo-
the only route: in the case of differential motion detection, stim- tion. Neuron 56:689 –700.
Roska B, Werblin F (2001) Vertical interactions across ten parallel, stacked
ulus generalization begins early in the circuit, and the selection of
representations in the mammalian retina. Nature 410:583–587.
a specific stimulus quality occurs one synapse later. Segev R, Puchalla J, Berry MJ 2nd (2006) Functional organization of gan-
glion cells in the salamander retina. J Neurophysiol 95:2277–2292.
References Shapley RM, Victor JD (1979) Nonlinear spatial summation and the con-
Baccus SA, Meister M (2002) Fast and slow contrast adaptation in retinal
trast gain control of cat retinal ganglion cells. J Physiol 290:141–161.
circuitry. Neuron 36:909 –919.
Skavenski AA, Hansen RM, Steinman RM, Winterson BJ (1979) Quality of
Born RT, Tootell RB (1992) Segregation of global and local motion process-
ing in primate middle temporal visual area. Nature 357:497– 499. retinal image stabilization during small natural and artificial body rota-
Burkhardt DA, Fahey PK (1998) Contrast enhancement and distributed en- tions in man. Vision Res 19:675– 683.
coding by bipolar cells in the retina. J Neurophysiol 80:1070 –1081. Victor JD, Shapley RM (1979) The nonlinear pathway of Y ganglion cells in
Cook PB, McReynolds JS (1998) Lateral inhibition in the inner retina is the cat retina. J Gen Physiol 74:671– 689.
important for spatial tuning of ganglion cells. Nat Neurosci 1:714 –719. Völgyi B, Xin D, Amarillo Y, Bloomfield SA (2001) Morphology and phys-
Costa LF, Velte TJ (1999) Automatic characterization and classification of iology of the polyaxonal amacrine cells in the rabbit retina. J Comp Neu-
ganglion cells from the salamander retina. J Comp Neurol 404:33–51. rol 440:109 –125.
Demb JB, Zaghloul K, Haarsma L, Sterling P (2001) Bipolar cells contribute Warland DK, Reinagel P, Meister M (1997) Decoding visual information
to nonlinear spatial summation in the brisk-transient (Y) ganglion cell in from a population of retinal ganglion cells. J Neurophysiol 78:2336 –2350.
mammalian retina. J Neurosci 21:7447–7454. Wright LL, Vaney DI (2004) The type 1 polyaxonal amacrine cells of the
DeVries SH (2000) Bipolar cells use kainate and AMPA receptors to filter rabbit retina: a tracer-coupling study. Vis Neurosci 21:145–155.
visual information into separate channels. Neuron 28:847– 856. Wu SM, Gao F, Maple BR (2000) Functional architecture of synapses in the
Engbert R, Kliegl R (2004) Microsaccades keep the eyes’ balance during inner retina: segregation of visual signals by stratification of bipolar cell
fixation. Psychol Sci 15:431– 436. axon terminals. J Neurosci 20:4462– 4470.