The impact of photoreceptor noise on retinal gain controls
Felice A Dunn1 and Fred Rieke2,3
Multiple retinal mechanisms preserve visual sensitivity as
the properties of the light inputs change. Rapid gain controls
match the effective signaling range of retinal neurons to
the local image statistics. Such gain controls trade an
increased sensitivity for some aspects of the inputs for a
decreased sensitivity to others. Rapid, local gain control
comes at another cost: noise in the signal controlling gain
(e.g. from the photoreceptors) will cause gain itself to vary
even when the statistics of the light input are constant.
Recent advances in identifying retinal pathways and the
sites and mechanisms of mean and contrast adaptation have
begun to clarify the tradeoffs associated with different gain
control locations and how these tradeoffs differ for rod and
cone vision.
Addresses
1
Program in Neurobiology and Behavior, University of Washington,
Seattle, WA 98195, USA
2
Howard Hughes Medical Institute, University of Washington, Seattle,
WA 98195, USA
3
Department of Physiology and Biophysics, University of Washington,
Seattle, WA 98195, USA
Corresponding author: Rieke, Fred ([email protected])
Current Opinion in Neurobiology 2006, 16:363–370
This review comes from a themed issue on
Sensory systems
Edited by Yang Dan and Richard D Mooney
Available online 11th July 2006
0959-4388/$ – see front matter
# 2006 Elsevier Ltd. All rights reserved.
DOI 10.1016/j.conb.2006.06.013
Introduction
Vision operates under an enormous range of lighting
conditions — from moonless nights to sunny days. Visual
fidelity under these diverse conditions relies on gain
controls that adjust sensitivity to multiple aspects of
the light inputs (reviewed by [1–3]), such as changes in
mean intensity, variability about the mean (i.e. contrast),
and spectral composition. Some of these gain controls
operate slowly, for example to accommodate the rising or
setting sun. Others operate more quickly, serving to
match gain to the statistics of the inputs from local regions
of the visual field. Considered together, these gain controls are part of what makes vision so difficult to emulate
with man-made imaging devices.
Here, we focus on three constraints on rapid gain control
mechanisms in the retina and their relationships to recent
www.sciencedirect.com
insights into retinal circuitry and processing. First, the
statistics of natural visual inputs, for example the temporal and spatial scale over which the mean light intensity
changes, set the landscape in which gain controls operate
(reviewed by [4,5]). Second, the signals controlling gain
must be derived from the photoreceptor signals, which
poses an inherent limit to the accuracy with which gain
can be controlled. Third, gain controls do not operate in a
vacuum, but instead interact with other computations
carried out in the retinal circuitry; these interactions
can make some gain control locations more sensible than
others.
The challenges the dynamic range of vision poses for our
understanding of neural function recur throughout the
nervous system. In particular, nearly all sensory systems
must accurately represent a wide range of input signals in
their output spike trains. The auditory system, for example, can detect sounds producing movements of the hair
cell stereocilia similar in magnitude to the diameter of
a hydrogen atom (reviewed by [6]), and yet can also
avoid saturation for sounds a trillion times louder [7].
Studies of retinal gain controls provide an excellent
opportunity to understand these general issues because
the circuitry is well defined and retinal signals can be
compared directly to over a century of behavioral measures of visual fidelity.
Gain controls and efficient coding
Figure 1 illustrates some of the challenges of encoding
realistic visual inputs. The statistics of the light inputs in
different regions of a scene differ substantially [8]
(reviewed by [4]). In Figure 1 and below, we emphasize
changes in mean intensity and in contrast, that is, local
variations in intensity about the mean. Efficient encoding of visual images requires matching the gain of
visual neurons to the image statistics so as to make good
use of the limited dynamic range of the responses of a
cell.
Eye movements change the region of the scene sampled
by the receptive field of a cell. This causes spatial
differences in mean and contrast to produce corresponding temporal changes in the signals encountered by the
cell [9,10]. Local eye movements (microsaccades) will
cause the temporal contrast — the temporal variations
about the mean — of the signal encountered to reflect the
local spatial structure in the scene (e.g. upper left trace in
Figure 1). Under free-viewing conditions, larger eye
movements (saccades) change fixation on a time scale
of 200 msec [11]. Saccades will often change the
mean (Figure 1 top) and contrast (Figure 1 bottom)
Current Opinion in Neurobiology 2006, 16:363–370
364 Sensory systems
Figure 1
Local image statistics and adaptation. Changes in fixation within a scene will cause changes in the mean light intensity (top) and contrast
(bottom) encountered within the receptive field of a visual neuron. Left and middle graphs above and below the picture depict intensities
encountered by local movements of the receptive field of a cell within the boxed region of corresponding color. Rapid gain controls match the
stimulus–response relationships of retinal neurons to the range of signals encountered (upper and bottom right).
encountered within the receptive field. Efficient encoding under these conditions requires that gain is controlled
more rapidly than fixation changes. Indeed, some retinal
gain controls operate in less than 100 msec following a
change in mean intensity [12,13] or contrast [14]; such
gain controls are well suited for adjusting sensitivity to
match the statistics of a local region of the visual scene
during a single fixation.
Gain controls that are engaged following a change in
mean or contrast have different effects on the relationship between light input and neural response. Increases
in mean intensity shift and broaden the stimulus–
response relationship (reviewed by [1,2]), centering it
on the mean intensity (Figure 1, top right). Increases in
temporal contrast broaden the stimulus–response relationship [15,16], matching the dynamic range of a cell to
Current Opinion in Neurobiology 2006, 16:363–370
the range of input signals encountered (Figure 1, bottom
right).
Constraints imposed by photoreceptor
signals and image statistics
Like most things in life, adaptation does not come for
free. Adaptation poses two potential problems for the
visual system. First, it involves decreasing sensitivity to
some aspects of the light inputs while increasing sensitivity to others. Simple examples are our poor ability to
estimate mean light intensity (try setting your camera
exposure manually) or visual illusions that exploit ambiguities created by adaptation of selected stimulus features. Second, gain controls necessarily operate on a finite
spatial and temporal scale. As a consequence, gain is
likely to vary even when the statistics of the input signals
are constant.
www.sciencedirect.com
Retinal gain controls Dunn and Rieke 365
An unavoidable source of noise in the signals controlling
gain comes from variability in the photoreceptor signals
themselves. This noise can be reduced by averaging over
multiple photoreceptors and/or over time. The effectiveness of such averaging, however, is limited by the extent
of spatial and temporal correlations in the visual inputs.
Furthermore, in some cases it is advantageous to adapt
before combining photoreceptor signals. As summarized
below, these tradeoffs depend on light level and differ
substantially for rod- and cone-mediated signals.
The good, bad and ugly of adapting in
single photoreceptors
Figure 2c shows responses of a primate cone to an
increase in light intensity from darkness to a level roughly
equivalent to a typical interior setting. The responses to
superimposed flashes monitor gain. This increase in background light intensity causes a noticeable change in the
gain of cone signals [17] (25% in Figure 2c) and of
downstream neurons in primate retina [18–20].
At typical interior light levels, both physiological and
behavioral experiments indicate that gain controls operating in single cones contribute substantially to the overall control of the gain of cone-mediated signals. The
intensity dependence of the response gain of single
primate cones resembles that of horizontal cells
[13,17,20]. Furthermore, behavioral experiments show a
nonlinear mechanism indicative of adaptation that operates on a spatial scale consistent with that of a single cone
[21]. Taken together, these observations suggest an
important site of adaptation operating on the signals
generated by a single cone, that is, in the cone phototransduction cascade or output synapse.
The gain of cone-mediated signals changes within
100 msec following a change in mean light level
[12,13], well-matched to the time between saccades. With
such rapid mechanisms in mind, Figure 2d shows a
400 msec section of the cone response centered on the
time of the increase in mean intensity. Detecting the
change in mean is possible but difficult on this time scale
because of noise intrinsic to the cone signals. The intrinsic dark noise of primate cones is equivalent to 4000
absorbed photons per second [17]; at mean light levels
below this the cone signals are dominated by intrinsic
noise.
The high level of noise in the cone responses suggests
that gain fluctuates over time even when the mean light
Figure 2
Effect of mean and contrast on voltage responses of a monkey long-wavelength sensitive cone. (a) Cone response to a step in light intensity
producing 200 absorbed photons per cone per second. Superimposed flashes (timing marked by the vertical lines below the trace) monitor
the gain of the cone response. The boxed region is expanded on the right. This mean light level decreases the gain of parasol ganglion cell
responses but not cones (see text). (b) Short section of the cone response from (a) centered on the step onset. (c) Response to a step producing
2000 absorbed photons per second. This light level decreases the gain of the cone responses (smaller responses to superimposed test flashes)
and downstream cells in the retina. (d) Short section of the response centered on the step onset. (e) Cone responses to a step by contrast from
0 to 25%; stimulus bandwidth 0–60 Hz. Ganglion cell responses adapt to similar contrast changes. (f) Short section of the response centered
on the time of the contrast change. The cone responses illustrated are dominated by cellular rather than instrumental noise.
www.sciencedirect.com
Current Opinion in Neurobiology 2006, 16:363–370
366 Sensory systems
level is constant. Figure 3 illustrates this issue. The
smooth curve summarizes measurements of the dependence of the gain of primate cone responses on mean light
intensity [17]. Gain, however, is not directly controlled by
mean light intensity but instead by a neural signal that
depends on the mean intensity, for example, the mean
cone voltage. Poisson fluctuations in photon absorption
and noise intrinsic to the cone produce noise in the signal
controlling gain. This noise will masquerade as a real
change in mean intensity and hence produce changes in
gain. Expressing the variations in the signal controlling
gain as equivalent variations in input light level, we can
use the relationship between gain and mean intensity to
estimate the gain fluctuations (dashed lines in Figure 3).
Because gain is controlled on a time scale similar to the
integration time of the cone responses, fluctuations in
gain will introduce noise comparable to that produced by
quantal fluctuations in the incident light and the cone
dark noise. This relatively large contribution of gain
fluctuations to cone noise will hold true under all conditions in which gain controls are active. Noise introduced
by gain fluctuations are a necessary evil of rapid gain
control.
Adaptation at the level of single cones is not without
benefits. For example, adaptation is naturally cone-type
specific, that is, changes in the photon absorption rate and
response gain of one cone type do not influence gain in
other cone types. Both gain changes measured in the
horizontal cells [22] and behavioral color matching experiments [23] show cone-type specific gain controls.
The possible locations for controlling the gain of signals
from a single cone type are limited because signals from
different cone types are mixed early in retinal processing.
Gap junctions between cones [24] and non-selective
feedback from horizontal cells to the cone synaptic terminal [25] mean that cone signals are already mixed in the
cone synaptic output. Additional mixing of cone signals
occurs in convergence of signals from different cone types
within the retinal circuitry. In both primate and rodent,
signals from short-wavelength sensitive (S) cones remain
segregated from those of middle- (M) and long-wavelength (L) sensitive cones in the receptive field centers of
specific bipolar and ganglion cells [26–28,29,30],
whereas in primate retina the surrounds of these cells
receive mixed M and L cone input. In primate, most
evidence points to a mixing of M and L cone signals
[31,32] (but see [33]), except in the receptive field
centers of cells receiving single cone inputs, that is,
midget bipolar cells and midget ganglion cells in the
central retina. In addition to this mixing, gain controls
Figure 3
Rapid gain controls introduce noise. Controlling gain quickly necessarily introduces fluctuations in gain. The smooth curve approximates the
dependence of cone gain (response amplitude normalized by response in darkness) on mean intensity (in photon absorptions per cone per second,
P*/cone/sec, see [17]), plotted on linear axes rather than the more typical logarithmic axes. Dotted lines illustrate how variability in the number
of photons translates into variability in gain values. Noise in the cone response and the finite temporal integration time of the mechanism controlling
the gain of cone-mediated signals will cause gain itself to vary in time even when the mean intensity is constant.
Current Opinion in Neurobiology 2006, 16:363–370
www.sciencedirect.com
Retinal gain controls Dunn and Rieke 367
located in cells receiving single cone input cannot benefit
from convergence to decrease cone noise and the consequent gain fluctuations (see below). Thus, cone-type
specific adaptation is best achieved in cone outer
segments.
Another benefit of adaptation in cones is that computations within the circuitry can benefit from the normalization of signals provided by adaptation. For example,
the inverse scaling of gain with mean intensity provided
by Weber adaptation causes a cell to encode contrast
rather than light intensity; such encoding simplifies the
task of subsequent mechanisms responsible for contrast
adaptation.
Convergence and network adaptation
Gain controls operating downstream of the photoreceptors can exploit convergence to obtain more reliable
estimates of the relevant scene statistics than is possible
based on responses of a single photoreceptor. This spatial
averaging expands the conditions under which gain can
be controlled effectively.
Cone-mediated signals
The retinal circuitry reading out the cone signals provides
multiple opportunities for gain control. Indeed, the gain
of cone-mediated responses measured in parasol (magnocellular projecting) ganglion cells, which can receive
input from hundreds of cones, is altered by mean intensities that do not affect the gain of the cone responses
themselves (Figure 4, middle) [18,19]. The situation is
less clear for midget (parvocellular projecting) ganglion
cells (Figure 4, left).
Figure 2a shows responses of a primate cone to a step
increase in light intensity that produces a change in
parasol but not cone gain; Figure 2b shows a 400 msec
period centered on the step onset. The step produces a
small (0.2 mV) hyperpolarization that is nearly impossible to detect on a short time scale based on the response
Figure 4
Summary of principal locations of gain controls for cone and rod circuitry. In cone circuitry, a midget ganglion cell in central retina (<2 mm
eccentricity; reviewed by [61]) receives input from a single midget cone bipolar, which is contacted by a single cone in the center of its receptive
field (left). A parasol ganglion cell receives input from a few hundred cones through diffuse cone bipolars (middle). The diagram does not account
for receptive field surrounds involving horizontal cells (H1). At low light levels rod signals transit the retina through the rod bipolar pathway (right):
rods ! rod bipolars (RB) ! AII amacrines ! cone bipolars ! ganglion cells.
www.sciencedirect.com
Current Opinion in Neurobiology 2006, 16:363–370
368 Sensory systems
of a single cone. Thus, controlling the gain of conemediated signals at these light levels requires averaging
over multiple cones. Although such averaging makes gain
control possible, noise in the signal controlling gain is still
likely to cause gain fluctuations such as those illustrated
in Figure 3.
Contrast gain controls face similar issues. Figure 2e shows
the response of a cone to a step from 0 to 25% contrast.
Similar changes by contrast strongly influence the gain of
ganglion cell responses [16,34] (reviewed by [3]), with
some mechanisms operating in 100 msec [14]. Figure 2f
shows a 400 msec segment of the cone response centered
on the time of the contrast change. Cone noise obscures
the contrast change on this time scale, as it did the
response to the light step in Figure 2a. Indeed, contrast
affects the gain of the retinal readout of the rod and cone
responses, but not the gain of the photoreceptor responses
themselves [34,35]. Much of the rapid (100 msec) contrast gain control appears to be intrinsic to the ganglion
cells [36,37]. This location is beneficial for several reasons. First, it could enable a contrast gain control mechanism to exploit the rescaling of signals provided by prior
mean gain controls (see above). Second, residual noise in
the signal controlling gain should be reduced by the
spatial averaging provided by the ganglion cell receptive
field (see Box 1).
Box 1 Relationship between mean and contrast
The division of light into discrete photons and the resulting Poisson
fluctuations in photon arrival imply that a change in mean light level
produces an unavoidable change in the extent of fluctuations
about the mean. As a consequence, mean and contrast adaptation
are not perfectly separable. The degree of this coupling depends on
light level and spatial averaging. Thus, at the low end of cone vision
(100 absorbed photons/cone/sec), intrinsic cone noise (equivalent
to 4000 absorbed photons/cone/sec; [17]) and variability in photon
arrival cause
pffiffiffiffiffiffiffiffithe fluctuations in the cone response (SD of photon
count = 205) to exceed the response to the mean light intensity
(5 absorbed photons) in each 50 ms integration time — that is, the
effective contrast noise (SD/mean) exceeds 100%. This effective
contrast noise decreases to 5% for light levels near thepmiddle
ffiffiffiffiffiffiffiffi of
cone vision (10 000 absorbed photons/cone/sec; SD = 700,
mean = 500). Convergence to downstream cells lowers the effective
contrast noise, for example, by a factor of 20 for a peripheral
ganglion cell receiving input from 400 cones. This effective contrast
noise limits the ability to detect, and hence adapt to, real image
contrast (typically 30% for natural images). Thus, for cone vision,
independent adaptation to mean and contrast requires at least
moderate light levels and spatial averaging.
The coupling of mean and fluctuations about the mean is
much stronger for rod vision. At low to moderate rod light levels
(< 0.1 absorbed photons/rod/sec), the effective contrast noise for a
100 ms integration time exceeds 100% for rod bipolar cells that
receive input from 20–50 rods and exceeds 10% even for a ganglion
cell receiving input from 10 000 rods. Thus, over much of the range of
rod vision, mean and fluctuations about the mean are strongly
coupled. Such coupling raises the possibility that the effect of mean
light on gain works through changes in the noise of neural responses
rather than the mean response [59,60].
Rod-mediated signals
The intrinsic noise of rods is about a million times less
than that of cones [38], enabling vision to operate when
photons are scarce. Vision under these conditions relies
on amplification of the single photon responses of rods
and convergence of signals from thousands of rods onto
downstream ganglion cells [39]. The combination of
amplification and convergence poses a danger of saturating neural responses at light levels that still produce
rare photon absorptions in individual rods. The scarcity
of photons makes controlling gain rapidly on the basis
of signals from a single rod impossible over much of
the operational range of rod vision. Instead, saturation
is avoided by gain controls operating in the retinal
circuitry (Figure 4, right) [40–42]. Gain controls operating at these low light levels must contend with the
irreducible statistical fluctuations in photon absorption
(see Box 1).
Where might such gain controls be located? Rod signals
are read out by three known pathways (reviewed by [43]).
The rod bipolar pathway provides the dominant readout
at low light levels [44,45]; in this pathway rod and cone
signals are mixed late in retinal processing (Figure 4,
right). At higher light levels two other pathways contribute: one based on rod–cone electrical coupling [46], and
the other on synapses from rods to a class of OFF cone
bipolar cells [47–50]. These pathways mix rod and cone
signals immediately. Thus, the rod bipolar pathway
Current Opinion in Neurobiology 2006, 16:363–370
provides a unique opportunity for specialized processing
operating exclusively on rod signals (e.g. [51–53]), including gain controls.
Behavioral measurements show that at least one mechanism controlling the gain of rod-mediated signals extends
spatially over a region encompassing 50 rods [54,55].
This is consistent with the convergence of rod signals
onto rod bipolar cells [39,48]. Our own physiological
recordings identify the synapse between the rod bipolar
cells and AII amacrine cells as a key gain-control site at
low light levels [56]. Taken together, these studies suggest that the gain control mechanism enjoys only modest
averaging over rods, and that gain is modulated by a
handful of photons in the rod bipolar pathway. As for
the cone signaling described above, this suggests that gain
varies considerably even when the mean light level is
constant.
Conclusions
Understanding retinal gain controls will require a synthesis of four issues, which to date have been largely studied
in isolation. First, gain controls operate at multiple locations in the retina and have diverse temporal and spatial
properties. Second, these mechanisms serve multiple
functional roles: protecting neural responses from saturation, discarding unimportant or uninformative responses
www.sciencedirect.com
Retinal gain controls Dunn and Rieke 369
[57], normalizing signals according to their fidelity,
emphasizing novel stimuli [58], and so on. Third, these
mechanisms operate in the context of natural image
statistics. Fourth, gain controls impact, both positively
and negatively, other computations taking place in the
retinal circuitry. We have emphasized how photoreceptor
noise interacts with these other considerations to determine how reliably gain can be controlled and the tradeoffs
associated with different gain control locations.
Acknowledgements
We thank T Doan and G Murphy for helpful comments on the
manuscript. Support was provided by the National Institutes of Health
through grant EY-11850 (F Rieke), the Achievement Reward for College
Scientists Foundation (FA Dunn), and the Howard Hughes Medical
Institute (FA Dunn and F Rieke).
References and recommended reading
Papers of particular interest, published within the annual period of
review, have been highlighted as:
of special interest
of outstanding interest
1.
Shapley R, Enroth-Cugell C: Visual adaptation and retinal gain
controls. In Progress in Retinal Research 3. Edited by Osborne N,
Chader G. Pergamon; 1984:263-346.
2.
Walraven J, Enroth-Cugell C, Hood DC, Macleod DIA, Schnapf JL:
The control of visual sensitivity. In Visual Perception: The
Neurophysiological Foundations. Edited by Spillmann L,
Werner JS. Academic Press; 1990:53-101.
3.
Shapley R: Retinal physiology: adapting to the changing scene.
Curr Biol 1997, 7:R421-R423.
4.
Simoncelli EP, Olshausen BA: Natural image statistics and
neural representation. Annu Rev Neurosci 2001,
24:1193-1216.
5.
Simoncelli EP: Vision and the statistics of the visual
environment. Curr Opin Neurobiol 2003, 13:144-149.
6.
Bialek W: Physical limits to sensation and perception. Annu Rev
Biophys Biophys Chem 1987, 16:455-478.
7.
Viemeister NF, Bacon SP: Intensity discrimination, increment
detection, and magnitude estimation for 1-kHz tones. J Acoust
Soc Am 1988, 84:172-178.
8.
Ruderman DL, Bialek W: Statistics of natural images: scaling in
the woods. Phys Rev Lett 1994, 73:814-817.
9.
Yarbus AL: Eye Movements and Vision. Plenum; 1967.
10. Reinagel P, Zador AM: Natural scene statistics at the centre of
gaze. Network 1999, 10:341-350.
11. Mazer JA, Gallant JL: Goal-related activity in V4 during free
viewing visual search. Evidence for a ventral stream visual
salience map. Neuron 2003, 40:1241-1250.
12. Yeh T, Lee BB, Kremers J: The time course of adaptation in
macaque retinal ganglion cells. Vision Res 1996, 36:913-931.
13. Lee BB, Dacey DM, Smith VC, Pokorny J: Dynamics of sensitivity
regulation in primate outer retina: the horizontal cell network.
J Vis 2003, 3:513-526.
14. Victor JD: The dynamics of the cat retinal X cell centre. J Physiol
1987, 386:219-246.
15. Brenner N, Bialek W, de Ruyter van Steveninck R: Adaptive
rescaling maximizes information transmission. Neuron 2000,
26:695-702.
16. Chander D, Chichilnisky EJ: Adaptation to temporal contrast
in primate and salamander retina. J Neurosci 2001,
21:9904-9916.
www.sciencedirect.com
17. Schneeweis DM, Schnapf JL: The photovoltage of macaque
cone photoreceptors: adaptation, noise, and kinetics.
J Neurosci 1999, 19:1203-1216.
18. Purpura K, Tranchina D, Kaplan E, Shapley RM: Light adaptation
in the primate retina: analysis of changes in gain and dynamics
of monkey retinal ganglion cells. Vis Neurosci 1990, 4:75-93.
19. Lee BB, Pokorny J, Smith VC, Martin PR, Valberg A: Luminance
and chromatic modulation sensitivity of macaque ganglion
cells and human observers. J Opt Soc Am A 1990, 7:2223-2236.
20. Smith VC, Pokorny J, Lee BB, Dacey DM: Primate horizontal cell
dynamics: an analysis of sensitivity regulation in the outer
retina. J Neurophysiol 2001, 85:545-558.
21. MacLeod DI, Williams DR, Makous W: A visual nonlinearity fed
by single cones. Vision Res 1992, 32:347-363.
22. Lee BB, Dacey DM, Smith VC, Pokorny J: Horizontal cells reveal
cone type-specific adaptation in primate retina. Proc Natl Acad
Sci USA 1999, 96:14611-14616.
23. Chichilnisky EJ, Wandell BA: Trichromatic opponent color
classification. Vision Res 1999, 39:3444-3458.
24. Hornstein EP, Verweij J, Schnapf JL: Electrical coupling between
red and green cones in primate retina. Nat Neurosci 2004,
7:745-750.
The authors demonstrate that long- and middle-wavelength cones are
electrically coupled, mixing the signals from these cone types early in
retinal processing. Modeling suggests that such coupling decreases
chromatic sensitivity while increasing luminance sensitivity. No evidence
for short-wavelength cone coupling was found, although the sample size
was small compared with that for the other cone types. Li and DeVries [62]
provided stronger evidence for a lack of electrical coupling between
short-wavelength and middle wavelength cones in squirrel retina.
25. Verweij J, Hornstein EP, Schnapf JL: Surround antagonism
in macaque cone photoreceptors. J Neurosci 2003,
23:10249-10257.
26. Dacey DM, Lee BB: The ‘blue-on’ opponent pathway in primate
retina originates from a distinct bistratified ganglion cell type.
Nature 1994, 367:731-735.
27. Chichilnisky EJ, Baylor DA: Receptive-field microstructure of
blue-yellow ganglion cells in primate retina. Nat Neurosci 1999,
2:889-893.
28. Schein S, Sterling P, Ngo IT, Huang TM, Herr S: Evidence that
each S cone in macaque fovea drives one narrow-field and
several wide-field blue-yellow ganglion cells. J Neurosci 2004,
24:8366-8378.
29. Haverkamp S, Wassle H, Duebel J, Kuner T, Augustine GJ, Feng G,
Euler T: The primordial, blue-cone color system of the mouse
retina. J Neurosci 2005, 25:5438-5445.
This anatomical study reveals exclusive connectivity between short (S)wavelength cones and S-cone bipolar cells identified in a transgenic
mouse. Even cones containing dual opsins (short and middle wavelengths) do not contact these specialized S-cone bipolars. Similarities
with the primate S-cone circuitry could be evidence for a common
primordial color system.
30. Li W, Devries SH: Bipolar cell pathways for color and luminance
vision in a dichromatic mammalian retina. Nat Neurosci 2006,
9:669-675.
This study is the first functional demonstration of cone connectivity to
cone bipolars in the mammalian retina. Paired recordings between short
(S) and middle (M) wavelength cones and a variety of cone bipolars reveal
that one class of cone bipolars receives exclusive input from S cones,
several classes from M cones, and two classes from a mix of S and M
cones. This work suggests that the exclusivity and mixing of cone signals
are suited for different visual functions, namely chromatic and luminance
detection.
31. Diller L, Packer OS, Verweij J, McMahon MJ, Williams DR,
Dacey DM: L and M cone contributions to the midget and
parasol ganglion cell receptive fields of macaque monkey
retina. J Neurosci 2004, 24:1079-1088.
The authors’ systematic examination of long (L) and middle (M) wavelength-sensitive cone inputs to horizontal cells, midget, and parasol ganglion cells tests predictions of cone-selective and random wiring. The first
hypothesis states that L and M cone inputs are segregated in the center
and surround regions of a receptive field. The results show equal mixing of
Current Opinion in Neurobiology 2006, 16:363–370
370 Sensory systems
L and M cone inputs to all three cell types, supporting the random wiring
hypothesis in the peripheral primate retina (but see Solomon et al. [33]).
46. Schneeweis DM, Schnapf JL: Photovoltage of rods and cones in
the macaque retina. Science 1995, 268:1053-1056.
32. Jusuf PR, Martin PR, Grunert U: Random wiring in the midget
pathway of primate retina. J Neurosci 2006, 26:3908-3917.
This work systematically examines the anatomical connections between
OFF midget cone bipolars and midget ganglion cells in marmosets with two
or three cone types. The authors reach three main conclusions: first, a
midget bipolar cell can contact multiple midget ganglion cells; second,
midget ganglion cells appear to contact all midget bipolar cells within the
area spanned by their dendrites; and third, the wiring is similar with 2 or 3
cone types. Thus, the anatomy does not meet the expectations for selective
wiring that exploits the extra cone type to increase chromatic sensitivity.
47. Soucy E, Wang Y, Nirenberg S, Nathans J, Meister M: A novel
signaling pathway from rod photoreceptors to ganglion cells
in mammalian retina. Neuron 1998, 21:481-493.
33. Solomon SG, Lee BB, White AJ, Ruttiger L, Martin PR: Chromatic
organization of ganglion cell receptive fields in the peripheral
retina. J Neurosci 2005, 25:4527-4539.
34. Smirnakis SM, Berry MJ, Warland DK, Bialek W, Meister M:
Adaptation of retinal processing to image contrast and spatial
scale. Nature 1997, 386:69-73.
35. Rieke F: Temporal contrast adaptation in salamander bipolar
cells. J Neurosci 2001, 21:9445-9454.
36. Kim KJ, Rieke F: Slow Na+ inactivation and variance adaptation
in salamander retinal ganglion cells. J Neurosci 2003,
23:1506-1516.
37. Zaghloul KA, Boahen K, Demb JB: Contrast adaptation in
subthreshold and spiking responses of mammalian Y-type
retinal ganglion cells. J Neurosci 2005, 25:860-868.
This work shows that contrast adaptation in mammalian ganglion cells
involves roughly equal contributions from mechanisms operating on
synaptic inputs and ganglion cell spike generation. This is consistent
with previous work in salamander retina (Kim and Rieke, [63]).
38. Baylor DA, Nunn BJ, Schnapf JL: The photocurrent, noise and
spectral sensitivity of rods of the monkey Macaca fascicularis.
J Physiol 1984, 357:575-607.
39. Sterling P, Freed MA, Smith RG: Architecture of rod and cone
circuits to the on-beta ganglion cell. J Neurosci 1988,
8:623-642.
40. Dowling JE: The site of visual adaptation. Science 1967,
155:273-279.
41. Frishman LJ, Sieving PA: Evidence for two sites of adaptation
affecting the dark-adapted ERG of cats and primates.
Vision Res 1995, 35:435-442.
42. Naarendorp F, Sato Y, Cajdric A, Hubbard NP: Absolute and
relative sensitivity of the scotopic system of rat:
electroretinography and behavior. Vis Neurosci 2001,
18:641-656.
43. Field GD, Sampath AP, Rieke F: Retinal processing near
absolute threshold: from behavior to mechanism. Annu Rev
Physiol 2005, 67:491-514.
48. Tsukamoto Y, Morigiwa K, Ueda M, Sterling P: Microcircuits
for night vision in mouse retina. J Neurosci 2001,
21:8616-8623.
49. Fyk-Kolodziej B, Qin P, Pourcho RG: Identification of a
cone bipolar cell in cat retina which has input from
both rod and cone photoreceptors. J Comp Neurol 2003,
464:104-113.
50. Li W, Keung JW, Massey SC: Direct synaptic connections
between rods and OFF cone bipolar cells in the rabbit retina.
J Comp Neurol 2004, 474:1-12.
51. Field GD, Rieke F: Nonlinear signal transfer from mouse rods
to bipolar cells and implications for visual sensitivity.
Neuron 2002, 34:773-785.
52. Sampath AP, Rieke F: Selective transmission of single photon
responses by saturation at the rod-to-rod bipolar synapse.
Neuron 2004, 41:431-443.
53. Berntson A, Smith RG, Taylor WR: Transmission of single
photon signals through a binary synapse in the mammalian
retina. Vis Neurosci 2004, 21:693-702.
54. MacLeod DI, Chen B, Crognale M: Spatial organization of
sensitivity regulation in rod vision. Vision Res 1989,
29:965-978.
55. Cicerone CM, Hayhoe MM: The size of the pool for bleaching
adaptation in human rod vision. Vision Res 1990, 30:693-697.
56. Dunn FA, Doan T, Sampath AP, Rieke F: Controlling the gain of
rod-mediated signals in the Mammalian retina. J Neurosci
2006, 26:3959-3970.
57. Atick JJ, Redlich AN: What does the retina know about natural
scenes? Neural Comput 1992, 4:196-210.
58. Hosoya T, Baccus SA, Meister M: Dynamic predictive coding by
the retina. Nature 2005, 436:71-77.
This work shows that adaptational mechanisms in the retina are much
more diverse than previously appreciated. In particular, these mechanisms
increase sensitivity to changes from one distribution of stimuli to another
for several classes of ‘complex’ stimuli (horizontal versus vertical bars,
temporally correlated or anticorrelated intensity waveforms, and so on).
59. Donner K, Copenhagen DR, Reuter T: Weber and noise
adaptation in the retina of the toad Bufo marinus. J Gen Physiol
1990, 95:733-753.
60. Brown LG, Rudd ME: Evidence for a noise gain control
mechanism in human vision. Vision Res 1998, 38:1925-1933.
44. Deans MR, Volgyi B, Goodenough DA, Bloomfield SA, Paul DL:
Connexin36 is essential for transmission of rod-mediated
visual signals in the mammalian retina. Neuron 2002,
36:703-712.
61. Rodieck RW: The first steps in seeing. Sinauer; 1998.
45. Volgyi B, Deans MR, Paul DL, Bloomfield SA: Convergence and
segregation of the multiple rod pathways in mammalian retina.
J Neurosci 2004, 24:11182-11192.
63. Kim K, Rieke F: Temporal contrast adaptation in the input and
output signals of salamander retinal ganglion cells. J Neurosci
2001, 21:287-299.
Current Opinion in Neurobiology 2006, 16:363–370
62. Li W, DeVries SH: Separate blue and green cone networks in
the mammalian retina. Nat Neurosci 2004, 7:751-756.
www.sciencedirect.com