J Neurophysiol 90: 3594 –3607, 2003.
First published August 27, 2003; 10.1152/jn.00699.2003.
Quantitative Characterization of Visual Response Properties in the Mouse
Dorsal Lateral Geniculate Nucleus
Matthew S. Grubb and Ian D. Thompson
University Laboratory of Physiology, Oxford OX1 3PT, United Kingdom
Submitted 19 July 2003; accepted in final form 26 August 2003
Studies of the functional characteristics of individual neurons in the visual system have added greatly to our understanding of how the brain works at the systems level. However,
while most work on visual neuronal response properties concentrates on species with good vision—monkeys, cats, ferrets—mice have been largely neglected by visual neurophysiologists as a model organism. This is partly because their
visual acuity is poor (Gianfranceschi et al. 1999; Prusky et al.
2000; Sinex et al. 1979) and partly because their small size
makes in vivo cellular recordings difficult. There are currently
two very good reasons for studying the mouse visual system at
the single cell level, however. First, for asking questions concerning the brain’s development, techniques for manipulating
an animal’s genome are invaluable ways of making highly
specific alterations to growing pathways. Among mammals,
these techniques are most advanced in mice. A bank of quantitative information concerning neuronal response properties in
wild-type mice would therefore constitute a first step toward
identifying single cell malfunction in transgenic animals and
toward inferring molecular mechanisms of normal development from those malfunctions. Second, for asking questions
about the brain’s evolution, comparing and contrasting specific
functions in a range of current species is extremely useful. A
quantitative description of visual response properties in the
mouse would therefore allow rigorous comparisons of cellular
function to be made between rodents, monkeys, prosimians,
and carnivores. This should teach us more about the pressures
that drove different visual systems to become the way they are.
We chose to start quantifying the mouse visual system by
characterizing the visual response properties of neurons in the
dorsal lateral geniculate nucleus (dLGN). As part of the best
understood of the brain’s sensory systems, the dLGN is open to
investigation using extremely well developed techniques for
analyzing single cell function. As the first stage of visual
processing within the brain, it represents a major center
through which a great deal of information makes its way to
higher processing regions in the cortex. Moreover, while a few
studies have begun to describe the functional characteristics of
the mouse retina (Balkema and Pinto 1982; Rossi et al. 2001;
Stone and Pinto 1993) and visual cortex (Dräger 1975;
Mangini and Pearlman 1980; Metin et al. 1988; Porciatti et al.
1999), we will be unable to fully understand what these regions
are doing until we know exactly what information is processed
by the structure that directly links them. Apart from the fact
that all cells are monocular (Metin et al. 1983), there are no
data currently available concerning single neuron response
properties in the mouse dLGN. This point is especially important where it concerns studies of function in visual cortex in
genetically manipulated mice (e.g., Fagiolini et al. 2003). The
developmental or functional abnormalities underlying altered
cortical physiology in these mutants could occur at the cortical
level but could just as easily be located at the subcortical level.
To explain changes in cortical function in transgenic mice, we
need to know the functional properties of their thalamic input,
and the first step toward such knowledge in the visual system
is a quantitative description of response properties in the wildtype mouse dLGN.
We present such a description here. After mapping their
receptive fields, we investigate the linearity of spatial summation, spatial frequency tuning, temporal frequency tuning, and
contrast response characteristics of individual mouse dLGN
cells. We also compare these characteristics across possible
Address for reprint requests and other correspondence: I. D. Thompson,
Univ. Laboratory of Physiology, Oxford OX1 3PT, UK (E-mail:
[email protected]).
The costs of publication of this article were defrayed in part by the payment
of page charges. The article must therefore be hereby marked ‘‘advertisement’’
in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
INTRODUCTION
3594
0022-3077/03 $5.00 Copyright © 2003 The American Physiological Society
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
Grubb, Matthew S. and Ian D. Thompson. Quantitative characterization of visual response properties in the mouse dorsal lateral
geniculate nucleus. J Neurophysiol 90: 3594 –3607, 2003. First published August 27, 2003; 10.1152/jn.00699.2003. We present a quantitative analysis of the visual response properties of single neurons in
the dorsal lateral geniculate nucleus (dLGN) of wild-type C57Bl/6J
mice. Extracellular recordings were made from single dLGN cells in
mice under halothane and nitrous oxide anesthesia. After mapping the
receptive fields (RFs) of these cells using reverse correlation of
responses to flashed square stimuli, we used sinusoidal gratings to
describe their linearity of spatial summation, spatial frequency tuning,
temporal frequency tuning, and contrast response characteristics. All
cells in our sample had RFs dominated by a single, roughly circular
“center” mechanism that responded to either increases (ON-center) or
decreases (OFF-center) in stimulus luminance, and almost all cells
passed a modified null test for linearity of spatial summation. A
difference of Gaussians model was used to relate spatial frequency
tuning to the spatial properties of cells’ RFs, revealing that mouse
dLGN cells have large RFs (center diameter approximately 11°) and
correspondingly poor spatial resolution (approximately 0.2c/°). Temporally, most cells in the mouse dLGN respond best to stimuli of
approximately 4 Hz. We looked for evidence of parallel processing in
the mouse dLGN and found it only in a functional difference between
ON- and OFF-center cells: ON-center cells were more sensitive to
stimulus contrast than their OFF-center neighbors.
RESPONSE PROPERTIES IN MOUSE dLGN
functional subdivisions in the nucleus and find evidence for an
important visual processing difference between ON- and OFFcenter cells.
3595
linked to the timing of visual stimulus changes using in-house software written by Darragh Smyth.
Optical quality and eye movements
METHODS
Animals
We used adult (⬎3 mo of age) pigmented C57Bl/6J mice of both
sexes weighing 20 –36 g (Harlan Olac, Oxon, UK), housed under a
12:12-h light:dark cycle. All experiments were conducted under the
auspices of the UK Home Office project and personal licenses held by
the authors.
Electrophysiology
J Neurophysiol • VOL
TABLE 1. Summary of response properties for all cells in our
dLGN sample
Parameter
Spontaneous activity (blank spikes/s)
Spatial summation (linearity)
Spatial tuning
kc
ks
rc (°)
rs (°)
Peak (c/°)
Cutoff (c/°)
Temporal tuning
Peak (Hz)
High50 (Hz)
Latency (ms)
Contrast response
Gain (spikes/s/%)
c50 (%)
Mean ⫾
SE/median
Number
of cells
3.24
0.44 ⴞ 0.04
133
34
14.88
0.97
5.61 ⴞ 0.41
16.98
0.027
0.18 ⴞ 0.0095
92
80
92
80
80
92
3.95 ⴞ 0.24
7.26 ⴞ 0.40
91.86
47
47
50
0.47 ⴞ 0.05
32.53 ⴞ 2.18
58
58
Values in bold are means ⫾ SE and come from samples that passed a
Kolmogorov-Smirnov test of normality. Those in italics are medians and come
from samples that failed such a normality test.
90 • DECEMBER 2003 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
Mice were first anesthetized with an intraperitoneal injection of
25% fentanyl and fluanisone (Hypnorm, Janssen Animal Health,
Bucks, UK) and 25% midazolam (Hypnovel, Roche Products, Herts,
UK) in Water for Injections (Norbrook Laboratories, County Down,
UK), administered at 2.7 ␮l/g. Surgical anesthesia was maintained
throughout the set-up procedures by subsequent applications of this
induction dose. To prevent the accumulation of bronchial secretions,
we gave 5 ␮l atropine sulfate (Animalcare, Yorks, UK) subcutaneously. A tracheotomy was performed as described elsewhere
(Schwarte et al. 2000), and a small plastic tube (1 mm OD; 0.5 mm
ID; SIMS Portex, Kent, UK) was inserted and secured by ligation
around the trachea. Application of a spot of cyanoacrylate glue added
extra stability by attaching the tube to the skin of the upper thorax.
The animal was situated in a custom-built head holder and secured
using a bite bar and ear bars before the tracheal tube was connected to
a respiratory pump (MiniVent 845, Hugo Sachs Elektronik, MarchHugstetten, Germany). We adjusted pump rate (usually 120 –140
strokes/min) and stroke volume (usually 150 –200 ␮l) to maintain the
expired carbon dioxide level, monitored with a Capstar-100 CO2
analyser (CWE, Ardmore, PA), at 2.5– 4%. Once on the pump, the
mice breathed a 1:3 mixture of oxygen and nitrous oxide, along with
1–1.5% halothane for maintained anesthesia. The mice were not
paralyzed at any stage, allowing continued assessment of anesthetic
depth via the paw pinch response. Throughout the experiment, heart
rate was monitored via an ECG (approximately 5 Hz was normal).
Core body temperature, measured with a rectal probe, was maintained
at approximately 37°C by combining a high ambient temperature with
heat from a thermostatically coupled blanket (NP 50 –7061-R, Harvard Apparatus, Kent, UK).
A circular craniotomy around 0.5 mm in diameter was made unilaterally above the presumptive location of the dLGN, approximately
2.5 mm posterior and 2 mm lateral of the bregma suture (Paxinos and
Franklin 2001). A small durotomy was made in the center of this
exposed zone, and after any whiskers occluding the eyes had been
trimmed, a tungsten-in-glass recording electrode (Alan Ainsworth,
Northants, UK) was lowered vertically into the brain using a microdrive. Changes in background activity while advancing through the
brain were used to locate the electrode relative to the dLGN. Above
the dLGN, we usually traveled through two bands of extremely active
hippocampal noise: one band approximately 100 ␮m wide at a depth
of approximately 1,300 ␮m from the pial surface and a second
fragmented band approximately 400 ␮m wide at a depth of approximately 2,000 ␮m. The dLGN, characterized by brisk, reliable visual
responses, was usually found at a depth of 2,500 –3,000 ␮m. Weak
and unreliable visual responses at this depth were a good sign that the
electrode was too medial and was recording from cells in the lateral
posterior nucleus (LP). Once in the dLGN, the electrode was advanced or retracted in 5-␮m steps, and its signals were amplified,
filtered, and thresholded until we were confident we were recording
from a single spike at any given time. The timing of spike events was
Mice are prone to temporary lens opacity while under any kind of
physiological stress (Fraunfelder and Burns 1970). Any visual responses recorded during such optical clouding were excluded from the
following analyses. We found that corneal quality was best maintained if the eyes were simply left untouched, and because of the
extremely small size of the mouse pupil, depth of field was assumed
to be extremely large in our animals (Remtulla and Hallett 1985).
Contact lenses were therefore not required for protective or corrective
purposes. To optimize optical quality we did not use atropine to dilate
the pupil; this meant that it was not possible to plot the location of the
optic disc in our experiments. Crucially, the fact that the most sensitive dLGN cells in this study could resolve spatial frequencies at
around the behavioral limit for C57Bl/6J mice (see DISCUSSION) suggests that optical quality was not significantly reduced under our
experimental conditions. We also used our spatial frequency tuning
data to confirm that optical quality did not degrade over the course of
our experiments. Across all animals, we found no significant correlations between time from experiment start (ⱕ13 h) and any of the six
spatial tuning properties listed in Table 1 (Pearson or Spearman
correlations, depending on Normality, P ⬎ 0.05). Furthermore, within
animals, there were no significant differences between cells recorded
in the first versus the second half of experiments (paired t-test or
Wilcoxon paired test, depending on Normality, P ⬎ 0.05) for any of
the six spatial tuning measures.
We also believe that eye movements did not have a significant
impact on the data presented here. Previous studies of single unit
visual responses in anesthetized C57Bl/6J mice report extremely
small eye movements over extended periods of time (Dräger 1975;
Gordon and Stryker 1996; Wagor et al. 1980). Here, when we remapped the receptive fields (RFs) of cells 0.5–1 h after their initial
localization, we saw a mean shift (⫾SE) in RF position of only 3.6 ⫾
0.5° (n ⫽ 11). In addition, over a timescale of approximately 10 min,
response features such as the phase-locked raster plots and the sinusoidal dependence of cell responses on stimulus phase in Fig. 2, along
with the regular shifts in response phase with increasing temporal
frequency in Fig. 7, all suggest that RF position changed very little
during single experiments. Quite simply, we would not have been able
3596
M. S GRUBB AND I. D. THOMPSON
to gather most of the data presented here had eye movements in our
preparation been large and frequent.
Stimulus presentation
Data processing
We generated RF maps from responses to flashed square stimuli
using reverse correlation analysis (Jones and Palmer 1987). Separate
ON and OFF RF profiles were constructed for a given cell by counting
the total number of spikes elicited when a stimulus of a particular
polarity (white squares for ON and black squares for OFF) and centered
at a particular point on the screen, was presented. Locations and
polarities that did not elicit a significant response (⬎2 SD above
spontaneous) were assigned a count of zero, and the response at all
points was normalized by dividing by the response to the optimal
stimulus, regardless of position or polarity. ON and OFF profiles were
subsequently overlaid to construct the “combined” RF profile. This
was smoothed with a Gaussian filter, chosen such that, within the
distance between two adjacent points of the original profile, the
Gaussian decayed to ⬍10%.
Responses to gratings were analyzed by first binning all responses
to a given stimulus according to their poststimulus time. We used 128
bins per stimulus cycle, producing a poststimulus time histogram
(PSTH) for each stimulus. A fast Fourier transform of these data then
converted them into amplitudes (in spikes/s) and phases (in degrees or
radians) of harmonics of the stimulus’ modulation frequency (Hochstein and Shapley 1976).
Curve fitting
All curve fits were carried out using a least squares minimalization
algorithm. Specifically, the algorithm minimized the following value
冘
关R data共pi兲 ⫺ Rfit共pi兲兴2
(1)
i
where Rdata denotes the recorded response amplitude of a cell, and Rfit
denotes the fitted response amplitude of that cell to a stimulus parameter p of a particular value i. As is the case whenever an algorithmbased minimalization approach is used to fit functions to raw dLGN
response data (e.g., Derrington and Lennie 1984; Hawken et al. 1996;
Norton et al. 1988; So and Shapley 1981; White et al. 2001), we could
never be sure that our minima were global, i.e., as small as possible
given all possible combinations of continuous input parameter values.
However, for each function (but not for each cell), starting parameters
close to the expected solution were set to avoid convergence to any
inappropriate minima, and the good fits achieved with this algorithm
(explaining approximately 97% of the raw data variance, see RESULTS)
suggest that the local minima it produced were sufficient to provide us
with useful and accurate data.
Plots of stimulus spatial frequency versus cell F1 response amplitude were fitted with a difference of Gaussians curve (Rodieck 1965;
So and Shapley 1981)
R共␷兲 ⫽ b ⫹ 共kc ⫺ b兲 䡠 共e⫺共␲ 䡠 rc 䡠 ␷兲 ⫺ ks 䡠 e⫺共␲ 䡠 rs 䡠 ␷兲 兲
2
Histology
We ended each experiment with an intraperitoneal injection of 0.3
ml pentobarbitone sodium (Sagatal, Rhône Mérieux, Essex, UK). The
mouse was transcardially perfused with phosphate buffered saline
(Sigma, Dorset, UK) followed by 4% paraformaldehyde (TAAB
Laboratories, Berks, UK) in 0.1 M phosphate buffer. The brain was
removed, sunk in 30% sucrose for cryoprotection, and sectioned
coronally at 30 –50 ␮m on a freezing microtome. We mounted sections from distilled water onto gelatinized slides. Subsequent staining
for cell bodies with cresyl violet allowed the identification of small (6
J Neurophysiol • VOL
2
(2)
where R is F1 response amplitude, ␷ is spatial frequency, b is baseline
response, kc is the area under the RF center’s Gaussian function, ks is
the relative area under the RF surround’s Gaussian function, and rc
and rs are the radii of the center and surround Gaussian functions,
respectively, at the point each mechanism has reached 1/e of its peak.
Unsmoothed maps of RF centers generated by reverse correlation
were fitted using a two-dimensional Gaussian curve of the form
90 • DECEMBER 2003 •
W共x, y兲 ⫽ a 䡠 e⫺共x/rx兲 䡠 e⫺共y/ry兲
2
www.jn.org
2
(3)
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
Visually responsive neurons were first identified using a hand-held
ophthalmoscope. For quantitative receptive field characterization,
stimuli were then presented on a CRT monitor controlled by a Cambridge Research Systems VSG 2/4 graphics card. The display was
initially placed with a given cell’s presumptive RF at its center;
mapping with reverse correlation then confirmed this location, or
directed movement of the monitor in order that the display covered the
entire RF area. Only cells whose RFs could be localized in this way
were subjected to subsequent analysis of response characteristics. The
display comprised 800 ⫻ 600 pixels, and at known viewing distances
between 135 and 400 mm, subtended 53–111° ⫻ 41–95°. With
maximum and minimum luminances of 99 and 2 cd/m2, respectively,
the display was the main source of illumination in an otherwise dim
room.
For quantitative receptive field mapping, we presented white (99
cd/m2) or black (2 cd/m2) square stimuli at various positions on a gray
background of mean luminance (51 cd/m2). Maps were obtained at
three resolutions. At the coarsest resolution, squares occupied 300 ⫻
300 pixels (21–58° square); for finer maps, they occupied 150 ⫻ 150
pixels (11–31° square); and for the finest maps, they occupied 75 ⫻
75 pixels (5–16° square). Squares were presented for 200 ms each at
every position on a 8 ⫻ 6 or 16 ⫻ 12 grid that comprised the entire
display. Each square was centered on one grid position but occupied
nine of them such that stimuli were considerably spatially overlapped.
Each stimulus run, consisting of one black and one white stimulus at
all 48 or 192 grid positions, was preceded by 10 s of a blank screen
at mean luminance from which spontaneous activity levels were
calculated.
For quantitative visual response characterization, we presented
sinusoidal monochromatic vertical gratings covering the full extent of
the display. These gratings were drifting in all experiments except the
null test, in which they were stationary and sinusoidally contrastmodulated. In all experiments, we kept all grating parameters constant
except the one under study; gratings varying in this parameter were
presented in a pseudorandom sequence. Every such stimulus repeat
also contained a blank stimulus at mean luminance from which
spontaneous activity levels were calculated. Each presentation of a
grating stimulus in a given repeat occurred for seven cycles; of these,
we used only responses from the last five cycles to discard any flash
responses to the initial presentation of the stimulus. The exception was
assessing temporal frequency (TF): in these experiments, to ensure we
captured enough responses to high TF stimuli, stimuli over 1 Hz were
presented for 7 s, with the first 2 s being discarded to allow for flash
responses.
To classify neuronal responses as sustained or transient, we flashed
full-screen white (99 cd/m2, for ON-center cells) or black (2 cd/m2, for
OFF-center cells) blank stimuli whose duration varied pseudorandomly
between 100 and 1,000 ms. There was a 4,000-ms black (for ON-center
cells) or white (for OFF-center cells) inter-stimulus interval, and ⱖ10
repeats of each stimulus duration were presented to each cell.
␮A for 6 s) electrolytic lesions made at precise locations on successful
penetrations. Locating these lesions made it possible to reconstruct
electrode tracks and localize recorded units. Only cells unequivocally
located within the dLGN were included in subsequent physiological
analyses.
RESPONSE PROPERTIES IN MOUSE dLGN
where W is a weighting function describing the RF center, x and y are
positions in the RF, a is amplitude, and rx and ry are the radii of the
function where it falls to 1/e of its maximum in the x and y directions,
respectively. The response of a linear cell to any stimulus can be
predicted by integrating, over the RF, the product of such a weighting
function and the stimulus. By applying Fourier transforms of this
response, a RF center mechanism’s F1 response amplitude to a vertical sinusoidal grating of a particular spatial frequency is given by the
center function in Eq. 2. In a cell that sums spatial inputs in a linear
manner, the two independently measured values rc and rx should be
equal.
Plots of stimulus temporal frequency versus cell F1 response amplitude were fitted with an atheoretical function comprising two
half-Gaussians
R共␻兲 ⫽ b1 ⫹ 共a ⫺ b1兲 䡠 e⫺关共p⫺␻兲/s兴
R共␻兲 ⫽ b2 ⫹ 共a ⫺ b2兲 䡠 e
3597
et al. 1979; Gabriel et al. 1996; Hale et al. 1979; Lennie and
Perry 1981).
The quantitative analyses that follow were carried out only
on those cells for which a RF had already been mapped in the
preceding manner. Our description of the mouse dLGN is
therefore incomplete, excluding the 33% of cells for which a
RF could not be obtained. However, while bearing in mind that
the descriptions below apply only to a subset of all mouse
dLGN neurons, also note that this subset contains the majority
of cells we encountered and contains the type of cells on which
most previous studies of the dLGN in mammals have concentrated.
for ␻ ⬍ p
(4a)
Linearity of spatial summation
⫺关共p⫺␻兲/s兴2
for ␻ ⬎ p
(4b)
We assessed linearity of summation in mouse dLGN cells
using a modified null test (Enroth-Cugell and Robson 1966;
Hochstein and Shapley 1976). Stimuli were stationary sinusoidal gratings whose contrast was counterphased sinusoidally.
These gratings were presented at a number of different spatial
phases and had spatial frequencies (SFs) at, or at multiples of,
the cell’s preferred SF. All 34 tested cells were presented with
gratings at optimal SF; 18 of these cells were also presented
with gratings at twice optimal SF, and 5 of those cells were
also presented with gratings at three times optimal SF. TF was
either 1 Hz or optimal; maximum contrast was 70%. In a linear
neuron, it should be possible to find two spatial phases at which
the introduction of such a grating does not produce any change
in cell firing. These “null” positions occur when ON and OFF
influences across all RF domains are perfectly balanced (Enroth-Cugell and Robson 1966). In all of our tested cells, at all
spatial frequencies, we found evidence for null positions. As
shown in Fig. 2, maximal responses to gratings at one phase
were contrasted with a complete lack of responses when the
phase was shifted by ⫾90°.
An extension to the null test states that, in linear cells,
responses should be dominated by their fundamental (F1)
harmonic component, since such cells respond to only one-half
of a given grating cycle (Shapley and Lennie 1985). Furthermore, the amplitude of this F1 component should vary systematically and sinusoidally with changes in stimulus phase
(Hochstein and Shapley 1976). Figure 2A shows exactly such
F1 dominance and sinusoidal variation in a mouse dLGN cell;
the two null positions are revealed as crossings of the zero line.
Using these data, linearity can be quantified by dividing the
mean F2 response amplitude by the maximum F1 response
amplitude across all spatial phases at any given SF and taking
the maximum value for a given cell across all SFs (Hochstein
and Shapley 1976). Cells with a linearity value ⬍1, such as the
cell in Fig. 2A (linearity ⫽ 0.11), are considered linear, while
those with values ⬎1 can be classified as nonlinear. While
many cell responses included large F2 components (Fig. 2B),
these were usually not indicative of ON-OFF frequency doubled
responses. Instead, they were caused by strong response rectification: low levels of spontaneous activity coupled with strong
transient firing acted to raise the strength of all non-fundamental response harmonics. As a result, the F2, and indeed F3 and
F4, response components were large and varied sinusoidally in
step with the F1 responses. The F1 component was always
strongest in these neurons, however, and as such, their linearity
values were ⬍1.
where R is F1 response amplitude, ␻ is temporal frequency, p is peak
temporal frequency, a is the response amplitude at optimum temporal
frequency, s is the Gaussian spread, b1 is the baseline on the lowfrequency side of the curve, and b2 is the baseline on the highfrequency side of the curve. Plots of stimulus temporal frequency
versus cell F1 response phase were simply fitted with a straight line.
Plots of stimulus Michelson contrast versus cell F1 response amplitude were fitted using a hyperbolic function (Albrecht and Hamilton 1982) of the form
R共c兲 ⫽ b ⫹ 共Rmax ⫺ b兲 䡠
cn
cn ⫹ hn
(5)
where R is F1 response amplitude, c is contrast, Rmax is maximum
response, h is the contrast at which R reaches one-half of Rmax, and n
is rate of change.
Statistical analysis
Sample sizes in this study were generally large enough (n ⬎ 30) to
reliably assess their Normality using a Kolmogorov-Smirnov test.
Those deemed Normal were described using mean ⫾ SE and were
compared using parametric tests. Those deemed non-Normal were
described using the median and were compared using nonparametric
tests. All comparison tests were two-tailed, with the level of significance set at 0.05 unless stated otherwise in the text.
RESULTS
Receptive field structure
Using reverse correlation of responses to flashed square
stimuli, we attempted to map the RFs of 199 well-isolated
dLGN neurons in 33 mice. In 133 (67%) of these cells, this
mapping was successful, revealing clearly localized regions of
visual space in which stimulus changes produced robust
changes in cell firing. In all 133 of these cases, the RFs were
characterized by the presence of a single, roughly circular
domain that responded to either increases (ON-center, n ⫽ 82)
or decreases (OFF-center, n ⫽ 51) in stimulus luminance (Fig.
1). We often also saw weaker responses to stimuli of the
opposite polarity in the region immediately surrounding these
center domains (Fig. 1A), suggesting that a large proportion of
neurons in the mouse dLGN behave like classical geniculate
center-surround filters (e.g., Wiesel and Hubel 1966). We
never saw any evidence of ON-OFF or simple cell-like RFs in our
sample, even though our methods should be sensitive to them
and despite reports of their presence in the rat dLGN (Fukuda
J Neurophysiol • VOL
90 • DECEMBER 2003 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
2
3598
M. S GRUBB AND I. D. THOMPSON
There was a single neuron with a linearity value above unity,
however. In this cell, we observed a dominant F2 component
reflecting doubled responses to a single stimulus cycle (see Fig.
2C, linearity ⫽ 1.63). Unlike classic retinal or geniculate
Y-cells (Hochstein and Shapley 1976; So and Shapley 1979),
this F2 response was not constant across all stimulus spatial
phases but showed sinusoidal variation with evidence of null
locations. This is exactly the type of behavior expected from an
ON-OFF RF possessing spatially overlapping, nonantagonistic ON
and OFF subfields (Lennie and Perry 1981). However, not only
did this cell’s mapped RF reveal no evidence for such an
ON-OFF organization (it was ON-center, with a clear OFF surround), we were extremely puzzled to see that its nonlinear
behavior was evident only at twice its optimal SF. When tested
at its optimal SF and at three times optimal, this cell displayed
entirely linear behavior reflected in a dominant, sinusoidally
modulated F1 component. We also saw exactly the same
pattern of results in another ON-center cell, although in this cell
a high maximal F1 response at twice the optimal SF produced
a high, but “linear,” linearity value of 0.88. We cannot explain
the behavior of these two cells—neither of them were abnormal in terms of any other response parameters—and describe
them in full in the hope that an imaginative reader may be able
to.
J Neurophysiol • VOL
Spatial summation population data are summarized in Fig. 3.
Here, linearity values are seen clustered around the mean of
0.44, and the two cells with partial evidence of ON-OFF activity
are clearly visible at 0.88 and 1.63.
Spatial frequency tuning
We assessed the spatial tuning characteristics of our dLGN
sample by presenting cells with drifting sinusoidal gratings of
various SFs at a temporal frequency of 1 Hz and a contrast of
70%. As well as directly revealing important aspects of a cell’s
spatial processing capabilities, responses of linear cells to
gratings of varying SF can be used to predict RF structure and
vice versa. Fitting a SF tuning plot with a curve generated from
a difference of Gaussians (DoG) equation (Rodieck 1965; So
and Shapley 1981) models the center and surround mechanisms of a cell as symmetrical antagonistic Gaussian functions
and can provide measures of the strengths and sizes of a given
cell’s center and surround RF regions (see METHODS). Because
a DoG curve also provides a smooth fit to raw data, it can also
be used to precisely calculate a cell’s SF optimum (or “peak”)
and cutoff (taken as the high SF at which response amplitude
decayed to 1% of its maximum).
We obtained good DoG fits to SF tuning plots in 92 mouse
90 • DECEMBER 2003 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
FIG. 1. Examples of ON- and OFF-center receptive fields in the mouse dorsal lateral geniculate nucleus (dLGN). A: ON-center
receptive field. B: OFF-center receptive field. Small panels show raw response-weighted profiles for white (ON, top) and black (OFF,
bottom) flashed squares (10.6° ⫻ 10.6° in A; 11.1° ⫻ 11.1° in B). The intensity of each pixel reflects the strength of the cell’s
response, as indicated by the greyscale key, to a flashed square centered on that location. Each large panel shows a combined,
smoothed receptive field profile. To produce these, each raw OFF profile was subtracted from the corresponding raw ON profile, and
the resulting map was smoothed with a Gaussian filter (see METHODS). Scale bars represent 14.1° in A and 14.8° in B.
RESPONSE PROPERTIES IN MOUSE dLGN
3599
dLGN cells: over this sample, the DoG function accounted for
a median of 98% of a cell’s raw data variance. The cell shown
in Fig. 4A1 responded to the highest SFs of any of the cells in
our sample, with a cutoff of 0.53c/°, a peak of 0.06c/°, and a
correspondingly small center mechanism radius, or rc, of 1.29°.
Most cells, however, responded very much like the cell in Fig.
4A2, with a cutoff at 0.18c/°, peak at 0.023c/°, and rc of 3.76°.
While the responses of such cells began to roll off at SFs below
the peak, assessing the true extent of this roll-off—vital for
estimating RF surround properties—required the presentation
of extremely low RFs that simply could not be handled by the
size of stimulus display we had available. We still present
population data for RF surround parameters in Table 1, but
these data are likely less reliable than our estimates of RF
center mechanism performance. In other cells, responses became progressively greater with decreasing SF, with no lowfrequency roll-off (Fig. 4A3). Whether these cells (n ⫽ 12) are
truly low-pass or whether their responses would decrease with
the presentation of extremely low SFs, we cannot tell. Surround and peak measures were thus not calculated for these
J Neurophysiol • VOL
cells, but their cutoffs still gave a very good indication of RF
center size.
Summary data for our sample are shown in Fig. 4, B–D.
Receptive field center radii were large, mainly in the order of
2–10°, with a few outlying cells possessing very large radii
ⱕ27° (Fig. 4B). Spatial frequency cutoffs were correspondingly low—most mouse dLGN cells can only resolve SFs up to
approximately 0.2c/°, although a couple of cells had cutoffs
around 0.5c/° (Fig. 4C). Peak spatial performance for most of
our dLGN sample was at around 0.03c/° (Fig. 4D).
Spatial tuning was assessed in cells whose RFs spanned a
broad range of visual field eccentricities. In a subset of cells
(n ⫽ 72), eccentricities in visual space were calculated by
combining information about a given cell’s RF position on the
stimulus screen, taken from two-dimensional Gaussian fits to
reverse-correlated response profiles (see METHODS), with information about the absolute location of the screen in the visual
field. In this frame of reference, the zero vertical meridian (0°
eccentricity) was aligned with the center of the animal’s head,
a far from perfect state of affairs. Even though eye position is
90 • DECEMBER 2003 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
FIG. 2. Assessing linearity of spatial summation in
mouse dLGN neurons using a modified null test.
Stimuli were sinusoidally modulating stationary sinusoidal gratings, at or above a cell’s preferred spatial
frequency, presented at a number of different spatial
phases. Temporal frequency was 1 Hz in A, 6 Hz in B,
and 3.6 Hz in C. Maximum stimulus contrast was
70%. A: linear, or X-like cell. Raster plots (each dot
represents a single spike) and poststimulus time histograms (PSTHs) show that the null stimulus (–150°)
produced no change in cell firing at all, while stimuli
90° away in either direction produced strong responses over one-half the stimulus cycle. Note the
double peak in the 120° and ⫺60° PSTHs, caused
here by the presence of both burst (1st peak) and tonic
(2nd peak) firing (e.g., Guido et al. 1992). The large
plot shows the sinusoidal dependence of this cell’s F1
responses on the spatial phase of the stimulus; 2 null
positions, 180° apart, occur where the F1 plot crosses
0. Since its responses are dominated by the fundamental harmonic, cell A has a low linearity value of 0.11
(see RESULTS). B: another linear unit. Like cell A, cell
B shows 2 null positions and a sinusoidal dependence
of its F1 responses on stimulus phase. Raster plots and
PSTHs show that the large F2 response component in
this cell was due to strong rectification rather than
frequency doubled ON-OFF responses (see RESULTS).
As such, this cell also has a low linearity value of
0.47. C: example of nonlinear behavior. Although
displaying a null response at ⫺30°, this cell shows
clear frequency doubled ON-OFF responses to the stimulus at phases 90° away. Indeed, this cell shows null
responding and sinusoidal variation in its F2 response
component, but neither of these in its F1 component.
Such nonlinear responses were observed only at a
single spatial frequency (see RESULTS); even so, this
cell’s linearity index is high, at 1.63.
3600
M. S GRUBB AND I. D. THOMPSON
of center radius taken from our reverse-correlated RF maps.
This measure, rx, was taken as the horizontal radius of a
two-dimensional Gaussian fit to a given cell’s center profile
(see METHODS), where the fit reached 1/e of its peak. We took a
measure of horizontal RF width alone because our spatial
frequency analysis, based on responses to vertical gratings,
assessed tuning only in the horizontal dimension. Figure 6
shows that the agreement between the two measures is very
good: although only two cells provide points that lie on the
(dashed) line of unity, rc and rx are very well correlated for
such a small sample size (r ⫽ 0.62, P ⫽ 0.04). Although more
time-consuming mapping using smaller squares and a finer
positioning grid would presumably tighten the relationship
shown here, this analysis shows that two independent measures
of RF center radius agree very well. We were thus reassured
reported to vary very little between mice placed securely in a
stereotaxic frame (Schuett et al. 2002) and the wide range of
stimulus positions employed here would minimize the effects
of such variation, differences in eye position across animals
could have masked differences in spatial tuning with eccentricity. Ideally, eccentricity should have been calculated in
retinal rather than spatial coordinates; however, determining
the optic disc projection in mice involves dilating the pupil
(e.g., Gordon and Stryker 1996; Schuett et al. 2002), a process
that would almost certainly alter the spatial tuning properties of
cells in the dLGN. We chose to maintain optical quality as high
as possible by leaving the eyes untouched (see METHODS), at the
expense of accurate information concerning eye position. Despite this important drawback, we still thought it worthwhile to
address whether spatial tuning properties varied with visual
eccentricity in the mouse dLGN (Fig. 5). Indeed, center radius
was found to be correlated positively with eccentricity (r ⫽
0.27, P ⫽ 0.02; Fig. 5A). However, RFs at 90° are only 3.7°
wider on average than their counterparts on the zero vertical
meridian (regression slope ⫽ 0.041°/°). Furthermore, this relationship is rendered insignificant if the three cells with largest
rc, all recorded at very eccentric locations, are removed from
the analysis. Spatial frequency cutoff values are negatively
correlated with eccentricity, but this relationship is not quite
significant (r ⫽ ⫺0.23, P ⫽ 0.052; Fig. 5B), and the slope of
the regression line is very small at ⫺0.00084c/°/°. The spatial
performance of mouse dLGN cells therefore becomes poorer
with increasing visual field eccentricity but only by an extremely small margin. For this reason, none of the analyses
presented here take RF eccentricity into account.
Since almost all cells in our mouse dLGN sample display
linear spatial summation, measurements of their spatial RF
properties via DoG fits to SF tuning plots should agree with
independent estimates of their spatial RF characteristics. In 11
cells whose RFs we had mapped at highest resolution, we
tested this prediction by comparing rc with a separate measure
J Neurophysiol • VOL
FIG. 4. Spatial frequency tuning properties of mouse dLGN neurons. A:
examples of spatial frequency (SF) tuning in 3 dLGN cells. Filled circles show
F1 response amplitudes to drifting sinusoidal gratings of various SFs; solid
lines show the best fit to these data of a difference of Gaussians (DoG)
receptive field (RF) model. Note the different y axis scales for the 3 examples
and the logarithmic x axis. Mouse dLGN cells respond to very low SFs. Cell
1 had the highest cutoff SF of any in our sample at 0.53c/°, while most cells
in our sample responded very much like cell 2, which peaks at 0.023c/° and
cuts off at 0.18c/°. Cell 3 is a low-pass cell, in which progressively decreasing
SFs were met with progressively increasing responses. Spatial tuning parameters for all 3 units: cell 1: kc 18.30, ks 0.51, rc 1.29°, rs 9.96°, peak 0.06c/°,
cutoff 0.53c/°; cell 2: kc 14.65, ks 0.78, rc 3.76°, rs 26.3°, peak 0.023c/°, cutoff
0.18c/°; cell 3: kc 18.05, rc 4.43°, cutoff 0.16c/°. B: distribution of RF center
radii, calculated from DoG fits to SF responses, of our full dLGN sample (n ⫽
92). Most radii are clustered around the mean at 5.6°, indicated by the solid
arrow, with a few large radius outliers ⱕ27.3°. C: distribution of SF cutoff
values across our dLGN sample (n ⫽ 92). These data were rather tightly
clustered around the mean at 0.18c/°, indicated by the solid arrow. D: distribution of peak SFs across all non–low-pass units (n ⫽ 80). These data did not
pass a Kolmogorov-Smirmov test for normality; the median at 0.027c/° is thus
marked with an open arrow.
90 • DECEMBER 2003 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
FIG. 3. Linearity of spatial summation across our mouse dLGN sample
(n ⫽ 34). The linearity measure should reflect linear spatial summation when
⬍1, and substantially nonlinear summation when ⬎1 (Hochstein and Shapley
1976; see RESULTS). Almost all cells were completely linear: the mean at 0.44
is marked with the arrow, and only 1 cell with spatial frequency-dependent
nonlinear summation (cell C in Fig. 2, see RESULTS) had a linearity value ⬎1.
RESPONSE PROPERTIES IN MOUSE dLGN
3601
FIG. 5. Relationships between receptive field center size,
spatial resolution, and visual eccentricity in the mouse dLGN.
A: each point represents a single unit for which visual eccentricity and receptive field center radius (rc) were both measured (n ⫽ 72). These 2 parameters are weakly, but significantly correlated (Pearson r ⫽ 0.27, P ⫽ 0.02). The solid line
(slope ⫽ 0.041°/°, intercept ⫽ 3.91°) shows the result of
linear regression on these 2 values. It should be noted that the
correlation between these 2 parameters becomes statistically
insignificant, although still positive, when the 3 outlying
points with largest rc are ignored (r ⫽ 0.2, P ⫽ 0.16). B: same
sample, here with visual eccentricity plotted against SF cutoff. These parameters are negatively correlated, although not
quite significantly (r ⫽ ⫺0.23, P ⫽ 0.052). The solid line
(slope ⫽ ⫺0.00084c/°/°, intercept ⫽ 0.21c/°) shows the best
linear regression fit to these data.
that measuring spatial RF center parameters through DoG fits
to SF tuning plots provided accurate and useful data.
We examined the TF tuning of mouse dLGN cells using
drifting gratings of various TFs at optimal SF and 70% contrast. As illustrated in Fig. 7A, 1–3, mouse dLGN cells were
band-pass tuned for TF: an optimal TF at which response
amplitude was maximal was accompanied by decreased responses at lower TFs and a somewhat sharper decline in
response at higher TFs. TF tuning plots were successfully fitted
with an atheoretical two half-Gaussians function (see METHODS)
in 47 cells. Over this sample, the function accounted for a
median of 97% of a cell’s raw data variance. From the resulting
curves, we calculated the TF peak along with the high50, the
high TF at which response amplitude fell to 50% of its maximum value. As shown in Fig. 7, B and C, most mouse dLGN
cells responded best to TFs around 4 Hz and were still responding at half-maximum response amplitudes at 7 Hz. While some
cells preferred rather low TFs of approximately 1 Hz (Fig.
FIG. 6. Agreement of independent measures of mouse dLGN receptive
field center size. In 11 neurons, we obtained measures of RF center size both
from DoG fits to spatial frequency tuning curves (rc) and from Gaussian fits to
RF maps generated by reverse correlation of responses to small flashed squares
(rx). Although only 2 units lie directly on the dashed line representing a 1:1
relationship, over all 11 cells, these 2 parameters were well correlated (Pearson
r ⫽ 0.62, P ⫽ 0.04).
J Neurophysiol • VOL
Contrast response characteristics
Stimulus contrast was the final domain in which we analyzed
the response characteristics of mouse dLGN neurons. Stimuli
were drifting sinusoidal gratings of various Michelson contrasts at peak SF and either peak TF or 1 Hz. All mouse dLGN
cells displayed monotonically increasing responses with increasing contrast—we saw no supersaturation of responses at
high contrasts (DeBusk et al. 1997; Li and Creutzfeldt 1984).
However, as shown in Fig. 8A, 1–3, there was considerable
variation in the nature of these increases. Some cells increased
their responses almost linearly with increasing contrast (Fig.
8A1), others had sigmoidal response profiles (Fig. 8A2), and
others increased their responses rapidly at even the lowest
contrasts (Fig. 8A3).
Contrast response data were successfully fitted with a hyperbolic function (see METHODS) in 58 cells. Over this sample,
the function accounted for a median of 97% of a cell’s raw data
variance. Despite the excellent fit of this function to the raw
data, however, the fitted parameters themselves were not useful
for describing the contrast response characteristics of mouse
dLGN cells. This was due to the large number of cells whose
responses did not saturate even at the highest contrasts (e.g.,
Fig. 8A, 1 and 3). In these cells, the fitted h parameter, the
value at which the cell’s response reaches one-half of Rmax, its
fitted maximum, was often ⬎100%, and Rmax itself often
reached unrealistically high values (ⱕ3,546 spikes/s). In other
words, for many cells, the 0 –100% range of “real” contrasts
occupied only the very lowest portion of the best fitting hyperbolic function. This is not a problem unique to the mouse
dLGN: Xu et al. (2001) encountered similar difficulties in
fitting a hyperbolic function to the contrast responses of non-
90 • DECEMBER 2003 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
Temporal frequency tuning
7A1), one responded best at over 10 Hz (Fig. 7A3), and high50
values ranged between 3 and 15 Hz.
We also used plots of TF versus response phase to calculate
a further temporal response property. If TF, in hertz, is converted to cycles per millisecond, and response phase, in radians, is converted into cycles, the slope of the best fitting line
relating the two is expressed in milliseconds and can be used as
an estimate of a cell’s visual latency (e.g., Hawken et al. 1996).
In our sample, this was possible in 50 cells. The cell in Fig.
7A1 had by far the longest latency of any in our sample at 256
ms but showed no other abnormal RF characteristics. Most
mouse dLGN cells had latencies like those of the cells shown
in Fig. 7A, 2 and 3, at around 90 ms (Fig. 7D).
3602
M. S GRUBB AND I. D. THOMPSON
saturating dLGN neurons in the owl monkey. Instead of using
wildly skewed fitted parameters to describe the contrast response characteristics of mouse dLGN cells, we chose to use
the fitted function over the “real” range of 0 –100% contrast to
calculate two more useful contrast response values. Contrast
gain was calculated as the slope of a tangent to the curve where
response amplitude was 20% of its value at 100% contrast. c50
represented the contrast at which response amplitude reached
50% of its value at 100% contrast.
Distributions of contrast gain and c50 values across our
mouse dLGN sample are shown in Fig. 8, B and C, respectively. Most gain values were rather low (Fig. 8B) and below
the mean of 0.47 spikes/s/%. At most, mouse dLGN cells at the
early, linear portion of their contrast response curve increased
their firing by 1.58 spikes/s/%. c50 values were spread over a
large range around the mean at 32.5% (Fig. 8C).
Parallel processing in the mouse dLGN?
To look for evidence of parallel processing in the mouse
dLGN, we used our sample to investigate four possible functional subdivisions of the nucleus. Since our mouse dLGN
neurons are almost all linear (Fig. 3), we could not assess
differences between linear and nonlinear cells (e.g., So and
Shapley 1979; Price and Morgan 1987). We could, however,
look for differences based on another functional feature that
has been used to distinguish cell groups in the carnivore dLGN.
In cats, cells that show sustained or transient responses to
constant stimuli have been shown to display different visual
processing characteristics (Cleland et al. 1971). We classified a
J Neurophysiol • VOL
subset of our neurons as “sustained” or “transient” based on
their responses to full-screen black (for OFF-center cells) or
white (for ON-center cells) flashes of varying duration (Fig.
9A). Responses over many repeats of each flash stimulus were
pooled, and this response strength was correlated with stimulus
duration. Each cell’s “sustainedness” value was the Pearson
coefficient, r, resulting from this correlation. Cells with high
sustainedness values thus fired increasingly more spikes to
stimuli of increasingly longer durations. The distribution of
sustainedness values across 46 cells was clearly bimodal and
was split at the median: cells with sustainedness values above
0.3 were classified as “sustained” (n ⫽ 23), while those with
sustainedness values below 0.3 were classified as “transient”
(n ⫽ 23; Fig. 9B).
We could also look for differences based on the position of
each of our cells in the dLGN. Although the mouse dLGN is
not obviously laminated, by analogy to the rat it may display
“hidden” lamination (Reese 1988). Based on structural and
anatomical characteristics, the portion of the rat dLGN that lies
closest to the optic tract on the dorsolateral edge of the nucleus
may represent a visual processing component separate from the
rest of the dLGN (Reese 1988). We have suggestive biochemical and anatomical evidence that a similar subdivision may
also occur in the mouse dLGN (data not shown), so we compared the visual response characteristics of cells in our sample
that lay close to the optic tract (within 100 ␮m of the dLGN⬘s
dorsolateral edge) with those lying within the core of the
nucleus. We also compared the functional properties of ON- and
OFF-center cells, and used the same sample to compare the
90 • DECEMBER 2003 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
FIG. 7. Temporal frequency tuning properties of
mouse dLGN neurons. A: examples of temporal frequency (TF) tuning in 3 dLGN cells. Left: filled circles
show F1 response amplitudes to drifting sinusoidal
gratings of various TFs, while solid lines show the best
fit to these data of a 2-Gaussian-halves equation. Note
the different y axis scales for each example and the
logarithmic x axis. Right: filled circles show F1 phases
of the same responses to the same TFs, while solid lines
show the best fitting linear regression to these data. The
slope of this line can be taken as a measure of cell
latency (e.g., Hawken et al. 1996). Note the linear x axis
here. Cell 1 had one of the lowest TF preferences
(peak ⫽ 1.29 Hz) and the longest latency (256 ms) of
any of the cells in our sample, while cell 3 had the
highest peak TF we observed, at 10.73 Hz. Most tuning
curves were like that of cell 2, which displayed a rather
steady rise to peak (here at 5.63 Hz) followed by a sharp
drop-off in response amplitude at higher TFs. All cells
in our sample (n ⫽ 47) showed band-pass tuning like
that of these 3 example cells. Temporal tuning parameters for all 3 units: cell 1: peak 1.29 Hz, high50 2.77
Hz, latency 256 ms; cell 2: peak 5.63 Hz, high50 9.24
Hz, latency 94 ms; cell 3: peak 10.73 Hz, high50 14.95
Hz, latency 86 ms. B: distribution of peak TFs across
our dLGN sample. Only cell 3 stands out from a population heavily clustered around the mean at 3.9 Hz,
marked by the solid arrow. C: high TFs at half-maximum curve height (high50) for all cells in our dLGN
sample. The mean at 7.26 Hz is indicated by the solid
arrow. D: histogram of latencies. While most values are
clustered tightly in the 75- to 125-ms range, cell 1’s
latency of 256 ms renders this sample non-Normal as
assessed with a Kolmogorov-Smirnov test. The median
of 92 ms is therefore indicated with the clear arrow.
RESPONSE PROPERTIES IN MOUSE dLGN
3603
between sustained and transient cells (P ⫽ 1), between neurons
situated near the optic tract and neurons located in the core of
the dLGN (P ⫽ 0.99), between ON- and OFF-center cells (P ⫽
0.89), or between cells recorded in male versus female animals
(P ⫽ 0.99). Taking all 13 response parameters described in this
paper together, we have no evidence for functional subgroups
of cells within the mouse dLGN.
We did find some evidence for parallel processing, however,
when we compared groups within single response parameters.
Individual t-test or Mann-Whitney tests (depending on Normality) for each of the 13 parameters listed in Table 1 revealed
no significant differences between sustained and transient cells
or between cells located in the dorsoventral “shell” of the
dLGN and neurons situated in the core of the nucleus. We did
find one significant difference between cells recorded in male
and female animals: in males, rs, the diameter of the RF’s
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
FIG. 8. Contrast response characteristics of mouse dLGN cells. A: examples of responses to drifting sinusoidal gratings of various Michelson contrasts
between 0 and 100%. Filled circles represent F1 response amplitudes to each
contrast; solid lines show the best hyperbolic fit to these data. Note the
different y axis scales. While some neurons like cell 1 displayed an almost
linear increase in response amplitude with increasing contrast, most showed at
least some saturation near 100%. Many, like cell 2, had distinctly sigmoid
response curves, rising sharply only after a period of slow increase at low
contrasts. Others, like cell 3, showed rapid increases in response amplitude at
even the lowest contrasts and began to saturate sooner. Contrast response
parameters for these 3 units: cell 1: gain 0.38 spikes/s/%, c50 43.3%; cell 2:
gain 0.97 spikes/s/%, c50 35.3%; cell 3: gain 0.24 spikes/s/%, c50 25.9%. B:
contrast gain distribution across our dLGN sample. Although skewed toward
low values, the sample passed a test of Normality (see METHODS), and the mean
at 0.47 spikes/s/% is marked with the solid arrow. C: c50 values across our
dLGN sample were spread over a large range. The solid arrow marks the mean
at 32.5%.
response characteristics of cells recorded in male and female
mice. Since for each response parameter shown in Table 1 the
same data set was subjected to four separate bivariate comparisons, we used Bonferroni correction of statistical significance
levels to allow for the increased chance, with repeated tests, of
finding significant differences between populations. The level
of significance for all following tests was therefore 0.0125.
We first looked for evidence of parallel processing in the
mouse dLGN by comparing candidate subpopulations across
the combination of all 13 response parameters listed in Table 1.
Since these parameters did not satisfy the assumptions of
Normality and equal variance required for a traditional multivariate ANOVA, we followed Van Hooser et al. (2003) in
employing a multivariate rank test (Choi and Marden 1997)
instead. Using this test, we found no significant differences
J Neurophysiol • VOL
FIG. 9. Classifying mouse dLGN cells as “sustained” or “transient.” A:
examples of sustained (cell 1) and transient (cell 2) responses to full-screen
high contrast flashes. Cell 1’s responses continued as long as the stimulus was
present, resulting in a high sustainedness (s) value of 0.997. The value represents
the Pearson correlation coefficient for the relationship between stimulus duration
and cell response (see RESULTS). In contrast, the duration of cell 2’s responses did
not vary with the duration of the stimulus, producing a low sustainedness value of
0.061. Plots show all spikes fired over several repeats of each stimulus; each dot
represents a single action potential. The dotted line shows the time of stimulus
termination. B: sustainedness values for all 46 neurons tested with full-screen
flashes. The distribution is clearly bimodal. The dotted line shows the median
sustainedness value of 0.3, at which the sample was split: all cells with a
sustainedness value ⬍0.3 were classified as “transient,” while all those with a
sustainedness value more than 0.3 were classified as “sustained.”
90 • DECEMBER 2003 •
www.jn.org
3604
M. S GRUBB AND I. D. THOMPSON
FIG. 10. Contrast response differences between ON- and OFF-center mouse
dLGN neurons. A: ON-center cells have significantly higher mean contrast
gains than their OFF-center counterparts (ON-center mean ⫾ SE ⫽ 0.6 ⫾ 0.07
spikes/s/%; OFF-center ⫽ 0.34 ⫾ 0.06 spikes/s/%; t-test, P ⫽ 0.0061). Error
bars show SE. B: ON-center cells have significantly lower mean c50s than
OFF-center cells (ON-center mean ⫾ SE ⫽ 24.9⫾ 2.4%; OFF-center ⫽ 40.7 ⫾
3.1%; t-test, P ⫽ 0.0001). Error bars show SE.
DISCUSSION
We have recorded visual responses from a sample of single
neurons in the mouse dLGN. In the subset of cells for which
we could obtain receptive field maps through reverse correlation, we saw classic geniculate center–surround organization.
We used sinusoidal grating stimuli to describe quantitatively
these cells’ linearity of spatial summation, spatial tuning, temporal tuning, and contrast response properties. Comparisons of
these properties across different subpopulations within the
mouse dLGN revealed significant differences only between ONand OFF-center neurons: ON-center cells are more spontaneously
active, and more sensitive to stimulus contrast than their OFFcenter neighbors.
Comparison with other species
Any cross-species homologies drawn on the basis of functional characteristics should be treated with extreme caution.
The natural world is full of structures that perform the same
function but have evolved from completely different starting
points. However, it seems clear from our sample that, in terms
of linearity of spatial summation, the mouse dLGN is not like
that of the carnivore. In cats and ferrets, ⱖ20% of dLGN cells
J Neurophysiol • VOL
90 • DECEMBER 2003 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
surround mechanism, was larger than in females (male median
22°, female 13°, Mann-Whitney test, P ⫽ 0.01). However,
since this is a single isolated difference and concerns a measure
that we do not believe to be particularly accurate, we do not
claim that spatial tuning differs between neurons in the male
and female mouse dLGN. We have much stronger evidence,
however, that a functional subdivision of the mouse dLGN
occurs between ON- and OFF-center cells. While the linearity,
spatial tuning, and temporal tuning of these two cell types were
indistinguishable, they differed significantly in terms of their
spontaneous firing rates and contrast response characteristics
(Fig. 10). Spontaneous firing rates, measured from responses to
a blank screen at mean luminance (51 cd/m2), were significantly higher in ON-center cells (median ⫽ 4.1 spikes/s) than in
OFF-center cells (median ⫽ 1.7 spikes/s; Mann-Whitney test,
P ⬍ 0.0001). In addition, over both contrast parameters, ONcenter cells were more sensitive than their OFF-center neighbors, possessing significantly higher gain (ON-center ⫽ 0.6 ⫾
0.07 spikes/s/%; OFF-center ⫽ 0.34 ⫾ 0.06 spikes/s/%; t-test,
P ⫽ 0.006; Fig. 10A) and lower c50 values (ON-center ⫽ 24.9 ⫾
2.4%; OFF-center ⫽ 40.7 ⫾ 3.1%; t-test, P ⫽ 0.0001; Fig. 10B).
in both A and C layers display Y-like nonlinearities (Price and
Morgan 1987; So and Shapley 1979; Sur and Sherman 1982).
We see no such cells in our sample. Instead the mouse dLGN
appears much more like that of the primate in that almost all
cells show wholly linear spatial summation (Blakemore and
Vital-Durand 1986; Derrington and Lennie 1984; Kaplan and
Shapley 1982; Levitt et al. 2001; Usrey and Reid 2000; White
et al. 2001; Xu et al. 2001); squirrel dLGN cells are almost all
linear too (Van Hooser et al. 2003). This should not be too
surprising—mice are more closely related to primates than are
cats or ferrets (Arnason et al. 2002), and as such, may offer a
better model for understanding our own visual system. That
mouse dLGN cells are mostly linear does, however, argue
against recent hypotheses that such linearity may be linked to
diurnal or tree-dwelling behavior (Van Hooser et al. 2003).
Rats and mice are very closely related (Arnason et al. 2002),
so we were surprised to observe an apparent difference in
dLGN cell types between the two species. While we observed
no cells with Y-like nonlinearities of spatial summation in the
mouse dLGN, in the rat dLGN, they outnumber the X-like,
linear neurons (Gabriel et al. 1996; Lennie and Perry 1981). It
seems unlikely that the mouse dLGN does contain neurons
with Y-like nonlinearities and that it was our methods that
caused us to miss them: tungsten electrodes bias sampling
toward Y-cells (So and Shapley 1979), reverse correlation
mapping reveals the RFs of Y-cells just as readily as it does
those of X-cells (e.g., Usrey et al. 1999), and in many cells, we
carried out null testing at higher than optimal SFs, where
Y-like nonlinearities are most obvious (e.g., Hochstein and
Shapley 1976). In addition, we recorded one neuron with
classic Y-like nonlinear behavior as assessed with a modified
null test (at twice optimal SF), but on electrode track reconstruction this cell was found to lie outside the dLGN, in the
region of the optic tract. It is certainly still possible that cells
with Y-like nonlinearities exist in the mouse dLGN and will be
found in future investigations; however, the present data suggest that such cells occupy at most a minor portion of the
dLGN population, and at least raise the possibility that the
mouse and rat dLGN could be fundamentally functionally
different.
Of course, it could be the case that in searching for crossspecies homologies of function, linearity is an extremely misleading benchmark. Indeed, despite the fact that carnivore
Y-cells are nonlinear and primate M-cells predominantly linear, it has been proposed on the basis of other functional and
anatomical data that these two dLGN sub-populations are
homologous (e.g., Dreher et al. 1976; Levitt et al. 2001; Sherman et al. 1976). Could a homologous (linear) Y- or M-cell–
like subpopulation exist within the mouse dLGN? If so, by
analogy with carnivores and primates, we might expect such
Y-like cells to represent a functionally distinct subpopulation.
We found no evidence for functional subdivision in our sample
of mouse dLGN neurons, and on the single measure investigated—peak SF—Lennie and Perry (1981) found no functional
difference between rat X- and Y-like neurons either. What
would be of great interest is information concerning the dendritic arbors of the subset of retinal ganglion cells that project
to the mouse dLGN. ␣-type retinal ganglion cells, which have
Y-like functional characteristics in the cat (e.g., Saito 1983)
and provide the predominant retinal input to M-cells in the
monkey dLGN (e.g., Perry et al. 1984), are known to exist in
RESPONSE PROPERTIES IN MOUSE dLGN
J Neurophysiol • VOL
spatial frequencies, it is still reasonably sensitive to differences
in stimulus contrast.
Comparison with other aspects of mouse vision
In all mammalian species studied so far, very little transformation of visual information occurs at the retinogeniculate
synapse. This trend is continued in the mouse, in which the
degree of retinogeniculate convergence is known to be low
(Chen and Regehr 2000). The RFs of mouse retinal ganglion
cells (RGCs), like the dLGN cells they project to, have a
center-surround organization (e.g., Balkema and Pinto 1982)
and are around 7–10° in diameter (Balkema and Pinto 1982;
Stone and Pinto 1993). In terms of spatial resolution, single
cells in mouse retina have a mean cutoff of approximately 0.2
c/° (Stone and Pinto 1993), while the electroretinogram (ERG)
response of the retina as a whole has an acuity of approximately 0.6 c/° (Rossi et al. 2001). Our dLGN sample here has
an average cutoff of 0.18 c/° and maximum acuity values
around 0.5 c/°. One aspect of visual processing that does
change in the transition from retina to thalamus, however, is
the strength of surround inhibition: it is stronger in cat dLGN
cells than in the RGCs that project to them (Usrey et al. 1999).
This certainly seems to hold true in the mouse as well. We saw
significant surround inhibition, as evidenced by decreased responses at low SFs, in almost all of our dLGN cells (only 12/92
cells appeared low-pass). In an isolated preparation of mouse
RGCs, however, such effects were rarely observed (Stone and
Pinto 1993), although comparisons between in vivo and in
vitro preparations should be taken with caution. The other
major difference between retinal and geniculate visual processing in the mouse relates to linearity: our sample contained no
Y-like cells, but Stone and Pinto (1993) observed four such
cells in a sample of 15 mouse RGCs. Although we believe it to
be unlikely, we may have simply missed such Y-like cells in
the mouse dLGN. Alternatively, nonlinear mouse RGCs could
project to nongeniculate targets or could synapse onto linear
dLGN neurons (Usrey et al. 1999: 2/12 retinogeniculate cell
pairs did exactly this).
The representation of the visual world changes a great deal
more over the thalamocortical pathway. Despite this, the
mouse primary visual cortex (V1) retains a number of spatial
processing features present in its thalamic input. Mean RF
sizes in mouse V1 range from 6° (Gordon et al. 1996) to 14°
(Metin et al. 1988)—roughly in keeping with the approximately 11° diameter in the dLGN. In addition, VEP data puts
cortical acuity at approximately 0.6 c/° (Porciatti et al. 1999),
and the best cell in our sample had a SF cutoff of 0.53 c/°. In
the temporal contrast domain, however, information present in
the thalamus is discarded by cortical processing. While 38/51
(75%) of cells in our dLGN sample could signal a TF of 16 Hz,
the TF cutoff for the entirety of V1 is approximately 12 Hz
(Porciatti et al. 1999). Such a loss of temporal resolution over
the geniculocortical synapse is also seen in the macaque
(Hawken et al. 1996). To be sure of these comparisons, however, the field needs quantitative data concerning the response
properties of single units in mouse V1.
At the behavioral level, mouse vision has only been assessed
in terms of its spatial acuity (Gianfranceschi et al. 1999; Prusky
et al. 2000; Sinex et al. 1979). In primates, behavioral acuity
has been shown to match the performance of the best thalamic
90 • DECEMBER 2003 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
the mouse retina (Peichl et al. 1987). If these ␣ cells project to
the dLGN in mice, do the cells they contact display distinct
functional characteristics? Until such evidence is available, we
must leave open the important possibility that a subpopulation
of Y- or M-like cells could exist in the mouse dLGN.
In terms of spatial tuning properties, comparing the mouse
dLGN with all other species studied thus far is simple: mouse
spatial vision is extremely poor. While primate and cat dLGN
cell RFs in the central visual field are usually ⬍1° in diameter
(e.g., Derrington and Lennie 1984; Levitt et al. 2001; Norton et
al. 1988; So and Shapley 1981; Usrey and Reid 2000; Usrey et
al. 1999; White et al. 2001; Xu et al. 2001), the average RF
diameter of the cells in our mouse dLGN sample is, at approximately 11°, even larger than the dLGN RFs found in the rat
(6 –11°: Hale et al. 1979; 8°: Fukuda et al. 1979) or squirrel
(1.5– 4°: Van Hooser et al. 2003). These large RFs are reflected
in poor spatial resolution. The average peak SF in our sample,
at 0.027 c/°, is much lower than those reported for rats (0.05–
0.09 c/°: Lennie and Perry 1981), squirrels (0.1– 0.3 c/°: Van
Hooser et al. 2003), carnivores (ferrets 0.1–1.3 c/°: Price and
Morgan 1987; cat X-cells 0.5–1 c/°: Lehmkuhle et al. 1980),
and primates (owl monkey P-cells approximately 0.7 c/°: Xu et
al. 2001; galago P-cells 0.8 c/°: Norton et al. 1988).
Temporally, however, mouse dLGN cells do not perform as
poorly. The mean peak TF in our sample was 3.8 Hz, and while
much lower than the same measure in macaque dLGN neurons
(10 –16 Hz: e.g., Derrington and Lennie 1984; Hawken et al.
1996), this compares very well with peak TFs in owl monkeys
(approximately 2 Hz: Xu et al. 2001; although see O’Keefe et
al. 1998), galagos (approximately 2– 4 Hz: Norton et al. 1988),
squirrels (approximately 5–9 Hz: Van Hooser et al. 2003), and
cats (4.75 Hz: Mukherjee and Kaplan 1995). Few other studies
have used TF versus F1 phase plots to calculate dLGN cell
latencies: values of around 75 ms in adult cats (Saul and
Feidler 2002) are slightly faster than the 94-ms median seen
here, but much quicker latencies have been observed in macaque monkeys (approximately 53 ms: Hawken et al. 1996;
approximately 40 ms: Levitt et al. 2001). Despite the short
distance between the mouse retina and dLGN, latencies may be
relatively slow due to differences in retinal processing, retinogeniculate synaptic efficiency, or retinal ganglion cell axon
diameter.
In terms of contrast response properties, the mouse dLGN
appears intermediate between primate M- and P-cells. Mean
contrast gain here (ON-center cells 0.6 spikes/s/%, OFF-center
cells 0.3 spikes/s/, all cells 0.47 spikes/s/%) generally lies
between average values for the two main dLGN subgroups in
macaque (M-cells 0.9 spikes/s/%, P-cells 0.2 spikes/s/%, Levitt
et al. 2001), squirrel monkey (M-cells 1.4 spikes/s/%, P-cells
0.2 spikes/s/%, Usrey and Reid 2000), and owl monkey (Mcells 0.57 spikes/s/%, P-cells 0.15 spikes/s/%, Usrey and Reid
2000), but is lower than the average gain for X- and Y-like
cells in the squirrel (approximately 0.8 spikes/s/%, Van Hooser
et al. 2003). Mean mouse c50 (ON-center cells 25%, OFF-center
cells 40%, all cells 33%) is very similar to that in squirrel
dLGN (approximately 35%, Van Hooser et al. 2003) and is
also intermediate between values reported for primate M- and
P-cells (squirrel monkey: M-cells 18%, P-cells 53%; owl monkey: M-cells 21%, P-cells 50%, Usrey and Reid 2000). Thus
although the mouse visual system operates over very low
3605
3606
M. S GRUBB AND I. D. THOMPSON
cells (e.g., Blakemore and Vital-Durand 1986), and this seems
to be mirrored in the mouse. Behavioral estimates put mouse
visual acuity at 0.5– 0.6 c/°, while the cell with the best
resolution in our sample had a cutoff SF of 0.53 c/°.
neurons do represent a functionally distinct subpopulation
within the mouse dLGN and that visual information in the
mouse brain is processed in a parallel manner.
The authors thank D. Smyth for computing assistance, T. Curnow for help
with data processing, and C. Akerman for helpful discussions.
Parallel processing in the mouse dLGN?
J Neurophysiol • VOL
DISCLOSURES
M. S. Grubb is supported by The Wellcome Trust.
REFERENCES
Albrecht DG and Hamilton DB. Striate cortex of monkey and cat: contrast
response function. J Neurophysiol 48: 217–237, 1982.
Arnason U, Adegoke JA, Bodin K, Born EW, Esa YB, Gullberg A, Nilsson
M, Short RV, Xu X, and Janke A. Mammalian mitogenomic relationships
and the root of the eutherian tree. Proc Natl Acad Sci USA 99: 8151– 8156,
2002.
Balkema GW Jr and Pinto LH. Electrophysiology of retinal ganglion cells in
the mouse: a study of a normally pigmented mouse and a congenic hypopigmentation mutant, pearl. J Neurophysiol 48: 968 –980, 1982.
Blakemore C and Vital-Durand F. Organization and post-natal development
of the monkey’s lateral geniculate nucleus. J Physiol 380: 453– 491, 1986.
Chen C and Regehr WG. Developmental remodeling of the retinogeniculate
synapse. Neuron 28: 955–966, 2000.
Chichilnisky EJ and Kalmar RS. Functional asymmetries in ON and OFF
ganglion cells of primate retina. J Neurosci 22: 2737–2747, 2002.
Choi K and Marden J. An approach to multivariate rank tests in multivariate
analysis of variance. J Am Stat Assoc 92: 1581–1590, 1997.
Cleland BG, Dubin MW, and Levick WR. Sustained and transient neurones
in cat’s retina and lateral geniculate nucleus. J Physiol 217: 473– 496, 1971.
DeBusk BC, DeBruyn EJ, Snider RK, Kabara JF, and Bonds AB. Stimulus-dependent modulation of spike burst length in cat striate cortical cells.
J Neurophysiol 78: 199 –213, 1997.
Derrington AM and Lennie P. Spatial and temporal contrast sensitivities of
neurones in lateral geniculate nucleus of macaque. J Physiol 357: 219 –240,
1984.
Dräger UC. Receptive fields of single cells and topography in mouse visual
cortex. J Comp Neurol 160: 269 –290, 1975.
Dreher B, Fukada Y, and Rodieck RW. Identification, classification and
anatomical segregation of cells with X-like and Y-like properties in the
lateral geniculate nucleus of old-world primates. J Physiol 258: 433– 452,
1976.
Enroth-Cugell C and Robson JG. The contrast sensitivity of retinal ganglion
cells of the cat. J Physiol 187: 517–552, 1966.
Fagiolini M, Katagiri H, Miyamoto H, Mori H, Grant SG, Mishina M, and
Hensch TK. Separable features of visual cortical plasticity revealed by
N-methyl-D-aspartate receptor 2A signaling. Proc Natl Acad Sci USA 100:
2854 –2859, 2003.
Fraunfelder FT and Burns RP. Acute reversible lens opacity: caused by
drugs, cold, anoxia, asphyxia, stress, death and dehydration. Exp Eye Res 10:
19 –30, 1970.
Fukuda Y, Sumitomo I, Sugitani M, and Iwama K. Receptive-field properties of cells in the dorsal part of the albino rat’s lateral geniculate nucleus.
Jpn J Physiol 29: 283–307, 1979.
Gabriel S, Gabriel HJ, Grutzmann R, Berlin K, and Davidowa H. Effects
of cholecystokinin on Y, X, and W cells in the dorsal lateral geniculate
nucleus of rats. Exp Brain Res 109: 43–55, 1996.
Gianfranceschi L, Fiorentini A, and Maffei L. Behavioural visual acuity of
wild type and bcl2 transgenic mouse. Vision Res 39: 569 –574, 1999.
Gordon JA, Cioffi D, Silva AJ, and Stryker MP. Deficient plasticity in the
primary visual cortex of alpha-calcium/calmodulin-dependent protein kinase
II mutant mice. Neuron 17: 491– 499, 1996.
Gordon JA and Stryker MP. Experience-dependent plasticity of binocular
responses in the primary visual cortex of the mouse. J Neurosci 16: 3274 –
3286, 1996.
Hale PT, Sefton AJ, and Dreher B. A correlation of receptive field properties
with conduction velocity of cells in the rat’s retino-geniculo-cortical pathway. Exp Brain Res 35: 425– 442, 1979.
Hawken MJ, Shapley RM, and Grosof DH. Temporal-frequency selectivity
in monkey visual cortex. Vis Neurosci 13: 477– 492, 1996.
90 • DECEMBER 2003 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
Studies of parallel processing pathways through the dLGN
have tended to concentrate on differences between linear and
nonlinear cells (e.g., So and Shapley 1979) or differences
between cells located in separate geniculate laminae (e.g.,
Kaplan and Shapley 1982). However, the mouse dLGN is not
laminated, and almost all of its cells sum spatial inputs linearly,
so we were forced to look for other forms of functional
subdivision within the nucleus. The cells in our sample did
appear to be differentiated on the basis of their response to a
constant stimulus—Fig. 9 shows that mouse dLGN neurons
tended to fire in either a “sustained” or a “transient” manner to
a maintained step in stimulus luminance. However, sustained
and transient cells were inseparable on any of the other spatial,
temporal, or contrast response parameters we measured.
We did find some evidence for parallel processing, however,
between ON- and OFF-center neurons. Taking all response parameters into account, these cell types were not significantly
different, but on individual measures we found ON-center cells
to be more sensitive to stimulus contrast and more spontaneously active than their OFF-center neighbors. Similar differences have been observed in the primate retina (Chichilnisky
and Kalmar 2002), suggesting that parallel processing through
ON- and OFF-center pathways is a visual system feature conserved across species and brain regions. In the dLGN, only one
previous study reports any ON- versus OFF-center differences:
primate ON-center cells have faster response kinetics (Lankheet
et al. 1998). Our mouse dLGN sample, however, displayed no
significant differences in temporal tuning between ON- and
OFF-center neurons. This may reflect a species difference or,
more likely, a lack of statistical sensitivity in our study: there
were trends toward higher peak TFs, higher high50s, and lower
latencies in our ON-center cells. It will be interesting to see
whether differences between ON- and OFF-center cells at the
thalamic level have any influence on cortical responses or
visual behavior.
Although our results suggest that, ON- and OFF-center differences excluded, there are no major functional subdivisions
within the mouse dLGN, they do not prove this. If there were
just a single (ON- or OFF-center) functional cell type within the
nucleus, the mouse visual system would differ greatly from
that of primates and carnivores and would appear to offer little
as a model for our own brain. However, it is impossible to
prove a lack of subdivision—although we tested the candidate
groupings available and most obvious to us, the distinct possibility remains that parallel processing occurs in the mouse
dLGN according to hitherto hidden subdivisions. For example,
as discussed above, Y-like (linear) cells may yet be uncovered
in the mouse dLGN and may possess a distinct set of visual
processing characteristics. Alternatively, an important functional distinction may be present between the cells described in
detail here, whose RFs could be obtained via reverse correlation techniques, and the 33% of cells we encountered whose
RFs could not be mapped quantitatively and objectively in this
way. Perhaps future investigations will show that these latter
RESPONSE PROPERTIES IN MOUSE dLGN
J Neurophysiol • VOL
Remtulla S and Hallett PE. A schematic eye for the mouse, and comparisons
with the rat. Vision Res 25: 21–31, 1985.
Rodieck RW. Quantitative analysis of cat retinal ganglion cell response to
visual stimuli. Vision Res 5: 583– 601, 1965.
Rossi FM, Pizzorusso T, Porciatti V, Marubio LM, Maffei L, and Changeux JP. Requirement of the nicotinic acetylcholine receptor beta 2 subunit
for the anatomical and functional development of the visual system. Proc
Natl Acad Sci USA 98: 6453– 6458, 2001.
Saito HA. Morphology of physiologically identified X-, Y-, and W-type retinal
ganglion cells of the cat. J Comp Neurol 221: 279 –288, 1983.
Saul AB and Feidler JC. Development of response timing and direction
selectivity in cat visual thalamus and cortex. J Neurosci 22: 2945–2955,
2002.
Schuett S, Bonhoeffer T, and Hubener M. Mapping retinotopic structure in
mouse visual cortex with optical imaging. J Neurosci 22: 6549 – 6559, 2002.
Schwarte LA, Zuurbier CJ, and Ince C. Mechanical ventilation of mice.
Basic Res Cardiol 95: 510 –520, 2000.
Shapley R and Lennie P. Spatial frequency analysis in the visual system.
Annu Rev Neurosci 8: 547–583, 1985.
Sherman SM, Wilson JR, Kaas JH, and Webb SV. X- and Y-cells in the
dorsal lateral geniculate nucleus of the owl monkey (Aotus trivirgatus).
Science 192: 475– 477, 1976.
Sinex DG, Burdette LJ, and Pearlman AL. A psychophysical investigation
of spatial vision in the normal and reeler mutant mouse. Vision Res 19:
853– 857, 1979.
So YT and Shapley R. Spatial properties of X and Y cells in the lateral
geniculate nucleus of the cat and conduction velocities of their inputs. Exp
Brain Res 36: 533–550, 1979.
So YT and Shapley R. Spatial tuning of cells in and around lateral geniculate
nucleus of the cat: X and Y relay cells and perigeniculate interneurons.
J Neurophysiol 45: 107–120, 1981.
Stone C and Pinto LH. Response properties of ganglion cells in the isolated
mouse retina. Vis Neurosci 10: 31–39, 1993.
Sur M and Sherman SM. Linear and nonlinear W-cells in C-laminae of the
cat’s lateral geniculate nucleus. J Neurophysiol 47: 869 – 884, 1982.
Usrey WM and Reid RC. Visual physiology of the lateral geniculate nucleus
in two species of new world monkey: Saimiri sciureus and Aotus trivirgatis.
J Physiol 523: 755–769, 2000.
Usrey WM, Reppas JB, and Reid RC. Specificity and strength of retinogeniculate connections. J Neurophysiol 82: 3527–3540, 1999.
Van Hooser SD, Heimel JA, and Nelson SB. Receptive field properties and
laminar organization of lateral geniculate nucleus in the gray squirrel (Sciurus carolinensis). J Neurophysiol 90: 3398 –3418, 2003.
Wagor E, Mangini NJ, and Pearlman AL. Retinotopic organization of striate
and extrastriate visual cortex in the mouse. J Comp Neurol 193: 187–202,
1980.
White AJR, Solomon SG, and Martin PR. Spatial properties of koniocellular
cells in the lateral geniculate nucleus of the marmoset Callithrix jacchus.
J Physiol 533: 519 –535, 2001.
Wiesel TN and Hubel DH. Spatial and chromatic interactions in the lateral
geniculate body of the rhesus monkey. J Neurophysiol 29: 1115–1156, 1966.
Xu X, Ichida JM, Allison JD, Boyd JD, Bonds AB, and Va VA. A
comparison of koniocellular, magnocellular and parvocellular receptive field
properties in the lateral geniculate nucleus of the owl monkey (Aotus
trivirgatus). J Physiol 531: 203–218, 2001.
90 • DECEMBER 2003 •
www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
Hochstein S and Shapley RM. Quantitative analysis of retinal ganglion cell
classifications. J Physiol 262: 237–264, 1976.
Jeon CJ, Strettoi E, and Masland RH. The major cell populations of the
mouse retina. J Neurosci 18: 8936 – 8946, 1998.
Jones JP and Palmer LA. The two-dimensional spatial structure of simple
receptive fields in cat striate cortex. J Neurophysiol 58: 1187–1211, 1987.
Kaplan E and Shapley RM. X and Y cells in the lateral geniculate nucleus of
macaque monkeys. J Physiol 330: 125–143, 1982.
Lankheet MJ, Lennie P, and Krauskopf J. Distinctive characteristics of
subclasses of red-green P-cells in LGN of macaque. Vis Neurosci 15: 37– 46,
1998.
Lehmkuhle S, Kratz KE, Mangel SC, and Sherman SM. Spatial and
temporal sensitivity of X- and Y-cells in dorsal lateral geniculate nucleus of
the cat. J Neurophysiol 43: 520 –541, 1980.
Lennie P and Perry VH. Spatial contrast sensitivity of cells in the lateral
geniculate nucleus of the rat. J Physiol 315: 69 –79, 1981.
Levitt JB, Schumer RA, Sherman SM, Spear PD, and Movshon JA. Visual
response properties of neurons in the LGN of normally reared and visually
deprived macaque monkeys. J Neurophysiol 85: 2111–2129, 2001.
Li CY and Creutzfeldt O. The representation of contrast and other stimulus
parameters by single neurons in area 17 of the cat. Pfluegers 401: 304 –314,
1984.
Mangini NJ and Pearlman AL. Laminar distribution of receptive field
properties in the primary visual cortex of the mouse. J Comp Neurol 193:
203–222, 1980.
Metin C, Godement P, and Imbert M. The primary visual cortex in the
mouse: receptive field properties and functional organization. Exp Brain Res
69: 594 – 612, 1988.
Metin C, Godement P, Saillour P, and Imbert M. Physiological and anatomical study of the retinogeniculate projections in the mouse. C R Seances
Acad Sci III 296: 157–162, 1983.
Mukherjee P and Kaplan E. Dynamics of neurons in the cat lateral geniculate
nucleus: in vivo electrophysiology and computational modeling. J Neurophysiol 74: 1222–1243, 1995.
Norton TT, Casagrande VA, Irvin GE, Sesma MA, and Petry HM.
Contrast-sensitivity functions of W-, X-, and Y-like relay cells in the lateral
geniculate nucleus of bush baby, Galago crassicaudatus. J Neurophysiol 59:
1639 –1656, 1988.
O’Keefe LP, Levitt JB, Kiper DC, Shapley RM, and Movshon JA. Functional organization of owl monkey lateral geniculate nucleus and visual
cortex. J Neurophysiol 80: 594 – 609, 1998.
Paxinos G and Franklin K. The Mouse Brain in Stereotaxic Coordinates.
London: Academic Press, 2001.
Peichl L, Ott H, and Boycott BB. Alpha ganglion cells in mammalian retinae.
Proc R Soc Lond B Biol Sci 231: 169 –197, 1987.
Perry VH, Oehler R, and Cowey A. Retinal ganglion cells that project to the
dorsal lateral geniculate nucleus in the macaque monkey. Neuroscience 12:
1101–1123, 1984.
Porciatti V, Pizzorusso T, and Maffei L. The visual physiology of the wild
type mouse determined with pattern VEPs. Vision Res 39: 3071–3081, 1999.
Price DJ and Morgan JE. Spatial properties of neurones in the lateral
geniculate nucleus of the pigmented ferret. Exp Brain Res 68: 28 –36, 1987.
Prusky GT, West PWR, and Douglas RM. Behavioral assessment of visual
acuity in mice and rats. Vision Res 40: 2201–2209, 2000.
Reese BE. ‘Hidden lamination’ in the dorsal lateral geniculate nucleus: the
functional organization of this thalamic region in the rat. Brain Res 472:
119 –137, 1988.
3607