Download Ordinal Position of Neurons in Cat Striate Cortex

J~URNALOINEUROPHYSIOLOGY
Vol. 42, No. 5, September
1979. Yrinted
in U.S.A.
Ordinal Position of Neurons in Cat Striate Cortex
JEAN
BULLIER
AND
GEOFFREY
H. HENRY
Dqm-tment
oj*Physiology,
John Curtin School oj’ ikkdicul
Nutionul University, Cunhurru, A. C. T. 2601, Austruliu
SUMMARY
AND
CONCLUSIONS
I. The ordinal, or serial, position of different types of striate neurons was assessed
from their latencies to electrical stimulation
of the optic radiations in paralyzed cats
anesthetized with N,O/O, (70/30%).
2. The distribution histogram for response latencies of striate neurons to electrical stimulation in the optic radiations
shows three broad peaks. These peaks become sharper if the stimulating site is moved
from a low point (OR,) to a high point (OR,)
in the optic radiations. It is argued that each
peak represents a population of cells lying
at a different ordinal, or serial, position
within the cortical pathway, i.e., receiving
a mono-, di-, or polysynaptic excitatory input from the thalamus.
3. The duration separating the peaks, of
the order of 1.0 ms, is then taken as the
transmission time for conduction of the
nervous impulse from one cortical cell to the
next. Confirmatory evidence for this value
of the cell-to-cell transmission time comes
from the measurement of interspike intervals in cortical cells that responded with
multiple spikes to electrical stimulation.
4. From the peaks in the distribution of
OR, latencies, boundaries have been established to arrive at an ordinal position for
131 cells classified as S, Sh, C, B, or cells
with nonoriented or concentric receptive
field (N-O and cone), following the terminology of Henry (14). Four ordinal groups were
distinguished: group M comprised cells receiving an exclusive monosynaptic input
from the thalamus, group C1+,iconsisted of
cells exhibiting a convergence of excitatory
inputs, one of which was monosynaptic;
group D was composed of cells in which
the lowest order input was disynaptic. Cells
were placed in group P if they were driven
from the thalamus through three or more
0022-3077/79/0000-OOOO$O
1.25 Copyright
0 1979 The
Rcscurc~h, Austruliun
synapses. Examination of the distribution of
different cell types for these various ordinal
groups revealed that no particular cell type
was exclusively associated with a single ordinal position. In particular, a large proportion of S-, Sh-, C-, and N-O and cone cells
showed a monosynaptic input from the thalamus (groups M and Cl+,)).
5. For the three classes of cell, S, Sh,
and C, with sufficient numbers in groups
MY cl+n~ and D, a search was made for differences in response properties associated
with the various ordinal groups. The properties examined included ocular dominance,
direction selectivity, orientation specificity,
and receptive-field size. With the exception
of the C-cell, which showed a difference of
direction selectivity in groups M and Cl+,),
we observed no systematic changes in the
response properties with differences in ordinal position.
6. For most cell types in cat striate cortex
there are examples of neurons receiving a
monosynaptic input. It is difficult to reconcile these findings with a hierarchical model
for the processing of visual information. It
is proposed that these cells are first-order
elements of different streams running in
parallel within the cortex and that this multiplicity of streams may be associated with
the diversity of outputs leaving the striate
cortex.
INTRODUCTION
By 1965, the pioneering studies of Hubel
and Wiesel (20-22) had led to the formulation of a model, later described as a hierarchical scheme, to explain the stages followed in the processing of information in
the visual cortex of the cat. In its original
form, simple cells of area 17, representing
the first stage in the processing sequence,
received a direct input from the dorsal lat-
American
Physiological
Society
1251
1252
J. BULLIER
AND
era1 geniculate nucleus (dLGN) and, in turn,
provided the input to complex cells. From
area 17 the axons of complex cells carried
the information into areas 18 and 19, where
they terminated on hypercomplex
cells.
Since 1965, additional
classes or subclasses have been recognized in the functional cell types found in area 17 of the cat.
Current attempts at modeling should allow
for the possibility
that these new classes
contribute to the function of area 17. In the
briefest
summary,
new subclasses
have
been suggested for simple cells (3, 13) and
complex cells (13, 16, 28), a new variety of
hypercomplex
cell has been found in area 17
(3, S), and cells with nonoriented
or concentric receptive fields have also been described in area 17 (15).
In the present study we have reexamined
the sequence of information
processing
in
the cat striate cortex. After allowing for a
greater diversity of cell types than envisaged
in the hierarchical scheme, an attempt was
made to class striate neurons according to
their ordinal or serial position in the processing sequence. The latency of orthodromic
activation by electrical stimulation at various sites along the retinostriate
pathway
was employed to assess the ordinal position
of striate neurons. This method has been
used previously
by a number of authors,
but generally their analysis was restricted to
a decision on whether the cell under examination was receiving a direct or indirect
input from the optic radiations.
Even so,
there is diversity in the conclusions
drawn
from their results; Denney, Baumgartner,
and Adorjani (7) saw their results as supporting the hierarchical
model, whereas
others (15, 19, 29, 30, 32) have shown that
a proportion
of both simple and complex
cells are potential recipients of a monosynaptic input.
The restriction
of the earlier studies to
the examination of monosynaptically
driven
cells came from the difficulty in assigning
ordinal positions from latency distributions
where the peaks representing
each ordinal
group were broad and overlapping. Only the
monosynaptically
driven cells responding
with short latency to electrical stimulation
could be separated, while groups of cells of
higher ordinal position could not be identified with confidence. We have attempted to
G. H. HENRY
overcome this problem through a reduction
of the scatter in the latency distribution.
This was done by shortening the signal-conduction time in the optic radiations by placing the stimulating electrode as close as possible to the cortex.
METHODS
General preparution
visual stimulation
und
A detailed account of the methods have been
published
elsewhere (15) and only a brief description is presented here. Experiments
were
performed on adult cats. Under halothane anesthesia, cannulas were inserted into the trachea,
the cephalic vein, and the femoral artery. Bilateral cervical sympathectomy
was performed
to diminish eye movements (25). At the end of
the surgical preparation
and after injecting the
wounds with long-lasting
local anesthetic, the
animals were paralyzed and artificially respired
with a mixture of nitrous oxide and oxygen (70/
30). The blood gasses were monitored during the
experiment and sodium bicarbonate was given to
compensate for metabolic acidosis, which was
frequently found to be present. The level of ventilation was adjusted to keep the partial pressure
of Co, in the blood at 30 mm Hg (10, 17).
The corneas were protected with plastic contact lenses and correcting lenses, determined by
retinoscopy,
were added to focus the eyes on a
tangent screen at 1 m. Artificial pupils were centered on the visual axis with the aid of an indirect
ophthalmoscope.
Single units were isolated with
tungsten-in-glass
microelectrodes
(23). The properties of the receptive field of each isolated unit
were qualitatively assessed by using bars or spots
of light of variable size and a variety of black
targets. For each unit, the minimum discharge
field (1, 3) was plotted from the responses to
light and dark edges moved in either direction
and the response to flashing light was studied
by flashing small spots or long bars of light in
different regions of the receptive field. The size
of the receptive field was measured by taking the
total area of the minimum discharge field. The
ocular-dominance
grouping followed the ratings
established by Hubel and Wiesel (21); direction
selectivity was estimated by listening to the response of the cell to an optimal stimulus moved
in both directions and was rated from 50 (nondirection selective) to 100 (direction
selective).
The orientation
selectivity was recorded as the
range of effective stimulus orientations
for moving stimuli. A pulse counter measured the spontaneous activity and the upper stimulus velocity
at which the cell ceased to fire (cutoff point) was
graded on a scale from 1 to 5.
ORDINAL
POSITION
Neurons were classified as S, Sh, C, B, or as
cells with nonoriented
(N-O) or concentric (cone)
receptive fields according to criteria laid down
previously
(14, 15). In summary,
a cell was
classed as possessing a nonoriented
receptive
field if light and dark bars, set at any orientation, produced similar responses when moved
across the receptive field. These cells also responded with an on/off discharge to small flashing spots positioned anywhere within the receptive field. Cortical
cells with concentric
receptive fields had similar response properties to
cells with concentric receptive fields found in the
dLGN and the retina. The cortical cells were differentiated from dLGN axons only by the fact
that they could be transynaptically
activated
from stimulating
electrodes in the optic radiations. The classification of neurons showing orientation specificity (S, Sh, C, B) was based primarily on the spatial disposition of discharge regions to moving and flashing stimuli. For S-cells
the discharge regions for light and dark moving
edges did not coincide spatially, although some
degree of overlap was’ often observed. If the cell
was not direction selective this feature held for
both directions of movement. When the flashing
light response was reliable enough, a similar displacement was also observed in the region of onand off-discharge.
By contrast, the discharge regions for moving edges in the receptive fields of
B- and C-cells overlapped
exactly and the
on- and off-discharge regions to stationary flashing bars also coincided spatially. B-cells were
distinguished from C-cells by their smaller receptive fields, lower spontaneous
activity, more
regular discharge to moving bars, and preference
for more slowly moving stimuli (for further details see Ref. 16). Cells with the suffix h displayed
the hypercomplex
property so that &-cells had
the same basic properties as S-cells but differed
in that the elongation of a moving bar stimulus
beyond its optimal length resulted, on most presentations, in the total suppression of the response.
Electrical stimulation
Three pairs of stimulating electrodes were positioned, one in the optic chiasm and two in the
optic radiations.
The electrical activation
was
provided by current pulses (O-10 mA) of variable duration (50-200 ps). The stimulating electrodes were made of tungsten wire isolated with
glass, with approximately
1 mm of wire protruding from the glass. The optic chiasm electrode
was positioned at Horsley-Clarke
AP 13.0- 15.0;
ML 1.0 (approximate
depth 23.0-24.0 mm). The
final depth was adjusted according to the strength
of the evoked potential
elicited by repeated
flashes of light and recorded through the stimulating electrode.
OF
STRIATE
NEURONS
1253
L
4
FIG. 1.
Groups of spikes elicited by electrical
stimulation at the site OR,, demonstrating
the variability
in
the latency
of firing in 20 stimulus
repetitions.
Cell
with nonoriented
receptive
field in lamina 4. Scale:
vertical bar, 0.75 mV, negative
upward;
horizontal
bar,
1.0 ms. Arrow
marks stimulus
onset.
The first pair of stimulating
electrodes in the
optic radiations (OR,) was positioned just above
the dLGN (H-C AP 5-6; ML 10; approximate
depth, 10 mm). The second pair of optic radiation
electrodes (OR,), with a longer protruding length
of tungsten wire (4 mm), was positioned close to
the cortex, a few millimeters from the recording
electrode. The positions of the stimulating electrodes in the optic chiasm and at OR, were
checked by gross dissection and the position of
the OR, electrode was verified in the histological
slides prepared for electrode reconstruction.
The criteria for distinguishing
antidromic from
orthodromic
activation
were conventional
(2,
1 I), and particular emphasis was placed on the
collision test to determine antidromic activation.
In a few cases, collision was observed in cells
that showed some jitter in their latency to electrical stimulation
(of the order of 0.3-0.4 ms).
This amount of jitter is unusual in antidromic
activation with short latencies (less than 2 ms)
(31) and these cases were, therefore, eliminated
from the data because of the uncertainty
of the
type of activation. Geniculate axons isolated in
the cortex were easily distinguished
from cortical cells by the constancy of their latency and
the aptitude to follow high frequencies of repetitive stimulation (up to 100 Hz).
The activation by electrical stimulation
was
presumed to be orthodromic
when the unit failed
to satisfy the conditions for antidromic or axontype activation.
Typically,
the jitter in orthodromic latency was of the order of 0.1-0.5 ms
for the first spike but was much larger for later
spikes (Fig. 1). The latency measures used
J. BULLIER
AND
G. H. HENRY
throughout
were the minimum latency between
the stimulus artifact and the foot of the A potential (2). In early experiments
the latency was
measured by examining the response pattern on
a conventional
oscilloscope,
but in later experiments a Polaroid camera or a storage oscilloscope was used in making the measurement.
No
difference was observed among the measurements obtained
with camera, storage oscilloscope, or with conventional
oscilloscope,
and we
pooled the data obtained using all three methods.
The minimal latency was obs&d
to vary by as
much as 0.2 ms, depending on the strength of
stimulation or whether the elicited spike was preceded by a spontaneously
occurring
spike. In
consequence,the precision of the latency measurement cannot be expected to be better than
0.2 ms and this value is certainly higher in the
case of polysynaptic
pathways.
first-, second-, and third-order neurons.
From the positions of these peaks one could
deduce characteristic values of latencies for
different ordinal positions and use these
values to assign an ordinal position to each
cell in the striate cortex. However, as shown
in Fig. 2A, the histogram of latencies to
electrical stimulation from site OR, early in
the optic radiations does not show narrow,
but rather broad peaks with a large number
of values between each peak. The variability
in conduction times for different dLGN
axons can be estimated from the latencies
to OR, stimulation of geniculate axon terminals isolated by the recording electrode
in the cortex. The distribution of OR, latenties for these axons is shown in Fig. 2B
where each axon was classified as BS (brisk
sustained) or BT (brisk transient) according
to its receptive-field properties (6). The scatter of latencies is similar to the interval between each peak in the histogram of Fig.
2A and the variability of conduction time
for different axons in the optic radiations
appears to be an important contributing factor in broadening the peaks in Fig. 2A.
We have attempted to reduce the effects
of variability in the conduction time of the
optic radiations by using a second stimulating electrode (OR,) positioned high in the
optic radiations, just below the striate cortex. Because of the shorter distance, the
RESULTS
In order to determine the ordinal position
of a cortical cell with respect to its thalamic
input, it would first seem appropriate to generate a frequency histogram of response latencies to electrical stimulation at a site
close to the thalamus. If the conduction time
in the optic radiations and the transmission
time from one cell to the next showed a small
degree of scatter, the histogram would show
sharp peaks representing populations of
A
3Or
q
q
BT
axon
BS
axon
I
4
OR,
2.
stimulation
FIG.
Histogram
of latencies for all spikes elicited
at site OR ,I. BT: brisk transi ent; BS: brisk
latency
( ms )
in cortical
su stained.
cells (A) and in dLGN
axons
(B) by electrical
ORDINAL
POSITION
OF STRIATE
variability of the conduction time from OR,
to the cortex is smaller than that from OR,.
This reduced variability is apparent in Fig.
3, which presents the scatter plot of OR,
latencies versus OR, latencies for nine geniculate axons. The scatter is significantly
smaller along the OR, (0.25 ms) than the OR,
(0.8 ms) axis. The regression line fitted to the
data points in Fig. 3 has a slope of 0.30
(95% confidence interval, 0.09-0.52) and intercepts the OR, axis at 0.21 ms (95% confidence interval, o-0.42). The value of the
slope is consistent with the ratio of the distances of the OR, and OR, electrodes from
the cortex (i.e., approximately
4 mm from
OR, and 14 mm for OR,). By contrast, the
deviation of the intercept of the regression
line away from the point of origin does not
carry statistical significance and, therefore,
does not require interpretation.
The reduced variability observed in the latencies of dLGN axons to OR, stimulation
has led us to adopt this measure for assessing the ordinal position of cortical cells.
However,
in doing so, there are two possible sources of error. The first is that the OR,
stimulating electrode may activate excitatory afferents to the cortex other than those
arising from the thalamus; and second, for
cells with a convergence of mono- and polysynaptic inputs, there is the possibility
that
the two radiation electrodes (OR, and OR,)
activate the cell through different synaptic
pathways.
The likelihood
of these errors
was reduced by selecting only the cells for
which similar spike patterns were elicited by
both OR, and OR, stimulation.
Also, cells
were rejected if the latency of homologous
spikes (from OR, and OR,) differed by more
than 1 ms. This led to a slight reduction in
1.07
x BT axon
c
0 BS axon
-
I
0.5
OR,
I
1.0
latency
I
1.5
J
2.0
( ms )
FLG. 3.
Scattergram
of OR, latency versus OR, latency for dLGN
axons isolated
in striate cortex.
Regression line has slope of 0.3 * 0.22 ms and intercept
at 0.21 5 0.21 ms (95% conf intervals;
Y = 0.78).
NEURONS
1255
-7 n
r
30
=m
82
%
b
$1
2
r0
r,
0
1
2
OR2 latency
3
4
5
( ms )
FIG. 4.
Histogram
of latencies for all spikes elicited
in cortical
cells by electrical
stimulation
at OR, site.
Arrows
mark criterion
levels for assigning ordinal position (see Table 1).
the sample sizes. Of the 166 cells that could
be activated from OR1, only 27 failed to be
activated from OR2, and among the 139 cells
that were activated from both OR, and ORz,
in only 4 units did the OR, and OR, latenties differ by more than 1 ms.
Figure 4 presents the histogram
of the
spikes elicited from OR, stimulation for cells
that satisfied the conditions
stated above.
The reduced variability in conduction time
results in sharper peaks in Fig. 4 than the
equivalent histogram of OR, latencies (Fig.
2A). The modes of these peaks occur at
1.2-1.4,2.2-2.6,
and 3.8-4.0 ms. The presence of low troughs between the peaks suggests that each peak is composed of a homogeneous group representing
a specific ordinal position but, in order to strengthen
this proposition,
we have sought to obtain
an independent estimate of the transmission
time from one cell to the next.
Trunsmission
time
To obtain an estimate of the cell-to-cell
transmission
time, we have examined the interspike intervals of the spikes generated by
electrical stimulation
when a single shock
elicited several spikes in the cell’s response
(Fig. 1). If each of the multiple spikes arises
from a signal passing through a different sequence of cortical cells, then the interspike
interval should be an integer multiple of the
cell-to-cell
transmission
time. To demonstrate first that the multiple spikes arose
from different inputs to the cell, we varied
the strength of electrical stimulation. Apart
from some minor variation in latencies for
each spike as the strength of suprathreshold
1256
J. BULLIER
AND
G. H. HENRY
dence of convergence of direct inputs we
concluded that, in general, such convergence is not manifested in the occurrence
of multiple spikes. Hence, in preparing the
distribution of interspike intervals in Fig. 6,
we included all cells that responded with
multiple spikes to stimulation at OR1, irrespective of whether a similar pattern of firing
could be elicited from stimulation
at ORz.
Peaking in Fig. 6 occurs at l-O/l .2 ms and at
2.0 ms, suggesting that 1.0 ms is an appropriate estimate for cell-to-cell transmission
time. In arriving at these values, however,
some caution should be exercised in interpreting the maximum point of the first peak
in the distributions
in Fig. 6 since the refractory state of the cell immediately followInterspike
interval
from OR, ( ms )
ing the first spike is likely to lead to truncation of the peak at intervals less than 0.8 ms.
FIG. 5.
Scattergram
of interspike
intervals
for mulThe value of 1 ms for the cell-to-cell transtiple spikes elicited
in cortical
cells by stimulation
at
mission time drawn from Fig. 6 agrees well
sites OR, and OR,. Line of unit slope represents
constant interspike
intervals
from both stimulation
sites.
with the value of interpeak separation (1.0
ms between peaks 1 and 2) in the histogram
of OR* latencies shown in Fig. 4. This
stimulation
was changed, the general pat- strengthens our claim that each peak in the
tern of response was independent
of the histogram corresponds to a different ordinal
strength of stimulation.
Then we also dem- position. It is interesting to note that the
onstrated that the multiple spikes did not re- cell-to-cell transmission time appears to be
sult from the convergence of fast and slowly
longer than the transmission time between
conducting afferents onto the cell. This was the dLGN axons and the cells represented
done by showing that the interspike interval
in the first peak of the histogram in Fig. 4.
remained the same for stimulation either at This axon-to-cell transmission time can be
OR, or OR,. A comparison of interspike in- estimated by taking the difference between
tervals for stimulation
at OR, and OR, is the mode of the first peak (an OR, value of
presented in Fig. 5. For the interspike inter1.2- 1.4 ms) and the mean OR, latency for
val to be invariant with the site of stimuladLGN axons (0.54 ms; SD. 0.1 ms; y2 = 9).
tion, the points should fall on the line of unit This gives a value of approximately 0.75 ms
slope passing through the point of origin.
(0.66-0.86 ms), which is clearly smaller
Most of the data points in Fig. 5 are within
than the 1 ms estimated for the cell-to-cell
0.2 ms (the estimated error of latency meas- transmission time.
urement) of the reference line and the greatest deviation is 0.4 ms. In the case of convergence of direct inputs on these cells, the
interspike interval would be diminished by
much more than 0.4 ms in going from stimulation at OR, to stimulation at ORz.
With the exclusion of the convergence of
direct inputs, the most likely explanation for
multiplicity
is that each spike comes through
a different synaptic pathway. If so, the cellInterspike
interval
from OR, ( ms )
to-cell transmission time may be derived
from the separation of peaks in the distribuFIG. 6.
Histogram
of interspike
intervals
for spikes
tion of interspike intervals. Since the sample
elicited in cortical cells by electrical
stimulation
at site
of points plotted in Fig. 5 produced no evi- OR,.
ORDINALPOSITION
OFSTRIATENEURONS
1257
Derivation oj* ordinal position from
TABLE
1.
Summary oj’ criteriu for
OR, latency
rating cells
The next step in the analysis is to arrive
No. of Spikes,
Criteria,
at the cutoff points on the OR, latency axis
to OR,
for Cells With0 < OR,
(Fig. 4) marking the boundaries for the Group
Stimulation
- OR, G 1.0 ms
groups of mono-, di- and polysynaptically
driven cells. A monosynaptic input from the
M
1 only
OR, s 1.6 ms
or OR, < 1.8 ms
thalamus is likely if the OR, latency is less
than 1.6 ms. This figure is based on the arguC I+n
Several
1 spike from M group,
ment that the minimum time for disynaptic
other spikes from D
or P groups
activation would equal the minimum OR,
latency for cells (0.8 ms from Fig. 4) plus D
1 or several
For first spike:
the minimum cell-to-cell transmission time
2.0 s OR, s 2.8 ms
(0.8 ms from Fig. 6). Similarly, monosynapP
1 or several
For first spike:
tic activation is unlikely for OR, latencies
0 R.,u 2 3.2 ms
longer than 2 ms. The reasoning here is that
the longest OR2 latency for an axon, showing an OR,-OR, latency difference of less We classed the se cells as bei ng no ndriven
(ND), with the implicat ion th at the y could
than 1 ms, will be smaller than 0.8 ms (Fig.
3). For a spike to be monosynaptically ac- not be driven electrically from the thalamic
tivated at 2.0 ms, the axon-to-cell transmis- region. Among the 166 cells that could be
sion time would need to be 1.2 ms, a dura- driven from OR1, 35 failed to meet the retion almost double the estimated value. This quirements laid down in Table 1 and could
appears unlikely in view of the sharpness of not be ascribed an ordinal rating. Thus, 131
cells were given an ordinal rating and Fig. 7
the first peak in the OR2 distribution.
By carrying through the reasoning applied shows how the various types of cells are disabove, it may be concluded with a high de- tributed among the four ordinal groups. The
gree of probability that spikes with OR, la- histograms in Fig. 7 record the percentage
tencies I) of less than 1.6 ms correspond to of each ordinal type (including the ND
a monosynaptic activation, 2) between 2 and group) for the various classes of cell. The
different classes of cell (S, Sh, C, B, cells
2.8 ms correspond to disynaptic activation,
with nonoriented and concentric receptive
3) greater than 3.2 ms are polysnaptically
(more than 2 synapses) driven. In Table 1 we fields) were grouped from their receptivefield properties as described in METHODS.
have drawn up a summary of the criteria
selected for giving each cell one of four or- Of the population of 223 cells (driven or not
dinal ratings: M (monosynaptic input), Cl+,) driven from OR,), 13 cells of uncertain clas(convergence of monosynaptic and other in- sification were not included in Fig. 7. The
puts), D (disynaptic input for the first spike), responses of 69 cells were too unreliable to
and P (polysynaptic input for the first spike). allow classification, and these are referred
In the case of seven cells, the conditions for to as NF (no field) cells and are included in
inclusion in the M group were relaxed be- Fig. 7. The high percentage of NF cells (69/
cause, although activation did not occur 258) possibly results from a decision to
from ORz, the OR, latency was clearly con- place cells in this category whenever a lack
sistent with monosynaptic activation (OR, of responsiveness created doubt about their
correct designation. The functional implicalatency < 1.8 ms).
tion of a group of celis with a low level of
Ordincrl position oj’ dijferen t types oj’
responsiveness is considered in the followstriate neuron
ing paper (5).
In all, 258 cells in 16 electrode penetraFigure 7 shows clearly that no single class
tions in 12 cats were studied for the re- of cell receives exclusively a mono-, di- or
sponses to visual and electrical stimulation.
POlY s ynaptic input from the thalamus . InOf these, 92 could not be driven with OR, stead , a varie tyofe xampl es were found with
stimulation although, in some instances, different ordinal ratings-in each class of cell.
cells of this group could be driven from OR,. With the exception that instances of poly-
1258
J. BULLIER
AND
G. H. HENRY
Sh
NF
N=2A
N=69
c
N=25
FIG. 7.
Percentage
distribution
among different
ordinal groups (Table 1) for cortical
cell types. N-O + cone,
cells with nonoriented
or concentric
receptive
fields. S, Sh, C, B as described
in METHODS.
NF, cells not
driven
bv visual stimulation.
ND, Cells not driven
by electrical
stimulation
at OR, site. N indtcates
the sample size for each cell type.
synaptic
input were not registered
for
S-cells and for cells with nonoriented or concentric receptive
fields, there was a full
spectrum of ordinal ratings to be found in
the various cell classes. However,
in some
groups-S,
C, and the nonoriented or concentric-there
is a clear preponderance
of
monosynaptically
driven cells, whereas in
the B, S,,, and NF groups, cells with a multisynaptic drive predominate. Almost all cells
with concentric
or nonoriented
receptive
fields may be regarded as first-order
cortical
neurons in that they belong either to the M
or C l+n grow.
Possibly associated
with
their first-order character, all cells with concentric and nonoriented
receptive
fields
could be driven electrically (see Fig. 7). In
sharp contrast,
only 14% of the cells that
could not be visually driven (NF cells) were
driven by electrical stimulation.
In general
terms, classes with a high incidence
of
monosynaptic
drive also have a high proportion of cells that can be driven electrically (e.g., S, C, cells with nonoriented or
concentric receptive fields), whereas classes
with a low incidence of monosynaptic
drive
(B and S, cells) have a higher proportion of
cells that were nonresponsive
to electrical
stimulation. Even the visually unresponsive
cells (NF class) may conform to this pattern
since there is some evidence that they are
higher order cortical neurons belonging to
laminae 2 and 3, which do not receive a
direct thalamic input (5).
Injlurnce oj’ ordinal position on
response properties
within each cluss
For three classes of cell-S,
Sh, and Cthere was a large enough sample to search
for within-class
variations that might be dependent on the ordinal position of the cell.
For this comparison of response properties,
cells of each class were subdivided into M,
C I+,,, and D subgroupings.
OCULAR
DOMINANCE.
In none of the three
classes of cell was the pattern of ocular dominance (21) found to vary with ordinal position. In particular, a growth in binocularity
(ocular dominance groups 3-5) was not observed to be associated with an increase in
the ordinal position of the cell.
DIRECTION
SELECTIVITY.
Within the S and
S,, classes there was a similar proportion of
direction-selective
units in all ordinal groups.
However, in the C category a decrease in the
direction-selective
character of the response
was observed in going from cells receiving
a purely monosynaptic
input (M) to those
ORDINAL
POSITION
OF STRIATE
receiving a convergence of mono- and polysynaptic inputs (C,,,). This variation is apparent in the distribution
histograms in Fig.
8 where direction selectivity is plotted on a
scale ranging from 50 (not direction selective) to 100 (totally direction selective).
Within
each
ORIENTATION
SPECIFICITY.
cell class no significant difference was observed in orientation
specificity,
recorded
by assessing the range of effective orientations, for cells of different ordinal position.
RECEPTIVE-FIELD
SIZE.
Similarly no significant size difference was apparent among
the ordinal groupings within each cell class
when the excitatory region of the receptive
field was plotted as a minimum response
field.
SPATIAL
DISPOSITION
OF LIGHTEDGE
DISCHARGE
REGIONS.
AND
i
M
Group
Cl+”
Group
D
FIG. 9. Distribution
of S-cells of different
ordinal
types of
groups CM. C,+n, D: Table 1) among different
recepttve-field
organization.
*, direction-selective
discharge; s, not direction-selective
discharge;
L, lightedge discharge;
D, dark-edge
discharge:
L,D,
hghtedge discharge
encountered
ahead of dark-edge
dtscharge:
0, miscellaneous
combination
of light- and
dark-edge
discharge.
M
“r
<Group
Grow
DARK-
For S cells in
the striate cortex the mapping of the receptive field is complicated by the fact that the
spatial disposition
of discharge
regions
varies when the moving stimulus is changed
from a light to a dark edge. The hypothesis
has been advanced, for the monkey at least,
that classes of cells based on the disposition
of light- and dark-edge discharge regions represent different ordinal positions in the striate cortex (27). To test this possibility
in
the cat, S-cells have been grouped into variGroup
1259
NEURONS
Cl+,
“1
ous combinations
of light- (L) and dark- (D)
edge discharge region disposition
for the
preparation of the frequency histograms for
M, G+n> and D groups in Fig. 9. The most
striking feature of the results in Fig. 9 is the
occurrence of examples of all subgroups of
S-cell in the monosynaptically
driven (M)
group. In consequence,
grouping based on
the spatial disposition of receptive-field
discharge regions does not appear to reflect
differences in the ordinal position of the cell.
Weight is added to this conclusion
in that
S-cells belonging to the C,,, and D groups
are also distributed
among several subclasses in Fig. 9.
DISCUSSION
’
8. Distribution
of C-cells of different
ordinal
groups (M and C,,,,) among the direction
selectivity
groups: 50, not direction
selective;
100, totally direction
selective.
FIG.
Ordinul position and intrrcellular
trunsmission
time
In interpreting the three peaks in the frequency distribution
histogram of OR, laten-
1260
J. BULLIER
AND
ties in Fig. 4, we have concluded that each
peak represents
a group of cells occurring
at a different ordinal position in the intracortical pathway. This interpretation
gained
support from multiple-spike
recordings with
the finding of a cell-to-cell transmission
time
of 1.0 to 1.2 ms -a duration similar to that
of the interpeak interval in Fig. 4. However,
this time for cell-to-cell transmission
is appreciably longer than the average synaptic
delay (0.4 ms) in the mammalian nervous
system (9) and one may question our estimate on the grounds that it is, in reality,
composed of transmission
through several
synapses. Any doubt of this nature seems
to be dispelled, however,
by the findings
of Toyama et al. (33). Recording intracellularly from cat striate cortex, Toyama and his
colleagues distinguished
two types of EPSP
-an
early and a late one-evoked
in cortical cells by electrical stimulation in the optic radiations.
After
stimulation
in the
dLGN the difference between the mean latencies of the two types of EPSP (early and
late) was approximately
1.1 ms-a
value
very similar to our estimate of the cell-tocell transmission
time. Moreover,
if a delay
of 0.2-0.3 ms is allowed for the generation
of the spike after the onset of an EPSP,
then the mean value of the OR, latency for
monosynaptically
(M and C,,,
groups:
mean 1.6 ms; SD 0.24 ms; n = 92) and disynaptically
(D group: mean 2.81 ms; SD
0.31 ms; n = 31) induced spikes in our results agrees almost exactly with the mean
latency for the early and late EPSPs in
Toyama et al’s study. A value of 1.0 ms,
therefore,
appears to be an accurate estimate of the time for cell-to-cell transmission
within the striate cortex.
The discrepancy
between the cell-to-cell
transmission
time and the time for synaptic
delay presumably
reflects the fact that the
transmission
time consists of other elements
in addition to synaptic delay. Components
such as conduction time of the impulse along
the presynaptic
axon and also the interval
in the second neuron between the onset of
the EPSP and the generation of the spike
must contribute
to the total transmission
time. There are reasons to believe that the
conduction time in the axon makes a major
contribution to the transmission
time. Cortical cells have axons that are generally unmyelinated and are of a fine caliber, equal
G. H. HENRY
to or less than 1 ,urn in diameter in Golgi
preparations
(unpublished
observations).
Assuming a conduction velocity of 1 m/s in
unmyelinated
fibers of 1 pm diameter (12,
18, 26), an impulse would take 0.5 ms to
travel 500 pm, a realistic distance for intralaminar conduction in the axons and axon
collaterals of striate neurons (15, 16). Such
a value together with a synaptic delay of 0.4
ms would account for most of the cell-to-cell
transmission
time.
Our observation
that the mean axon-tocell (from optic radiation to first-order
neuron) transmission
time is shorter than the
mean cell-to-cell transmission
time may be
related to the comparatively
small distance
traveled by the unmyelinated terminal twigs
of dLGN axons. For most of the dLGN
axons infiltrated with horseradish
peroxidase in a companion study (5), myelination
ceased when axon diameters reached approximately
1 pm within the lamina of termination and in close proximity (within 250
pm) to the receiving cell.
Interpretdon
of response patterns
The accuracy of our results depends on
the exclusion of errors that may arise from
the misinterpretation
of response patterns.
In particular,
to avoid distortion of the histogram of OR, latencies in Fig. 4 it is essential to know if every spike elicited by OR,
or OR, stimulation corresponds
to an excitatory input from the thalamus. This would
not be the case if the spike resulted from
antidromic
activation
traveling
either directly along the cell’s axon or indirectly via
a collateral from the efferent axon of a neighboring cell. Direct antidromic
activation
was systematically
excluded by applying the
collision test (11) when latency measurements showed a variability of less than 0.2
ms. To rule out the possibility of activation
of collaterals of efferent axons of other cortical cells, the cell was tested for orthodromic activation from the optic chiasm.
Signals induced from the optic chiasm only
activated dLGN cells and left unstimulated
the efferent axons arising from the cortex.
This test was applied to 86 of the 95 cells
apparently receiving a monosynaptic
input
from the thalamus (i.e., groups M and C,,,).
In every case among these 86 units, the OX
latency was consistent with disynaptic activation from the optic chiasm (one synapse
ORDINAL
POSITION
OF STRIATE
in dLGN and one in the cortex). We therefore concluded that the signal from OR, or
OR, seldom, if ever, passed to the cell via
the excitatory
collaterals of efferent axons.
The apparent rarity of cell activation
via
axon collaterals seems to indicate that the
excitatory
action of collaterals is too weak
to generate a spike in the postsynaptic
cell
and, therefore, the axon collaterals of pyramidal cells in laminae 5 and 6 may not
provide a path for sequential processing.
Another possible source of distortion of
the peaks in the histogram presented in Fig.
4 comes from the wide range of conduction
velocities
of dLGN
axons. Although
we
have demonstrated
that the use of OR, electrode reduces the scatter due to conduction
time in the optic radiations, it is worthwhile
to consider how much variability in the distribution of OR2 latencies can still be attributed to conduction in the optic radiations.
A scattergram
of OR, latency versus OR,OR, latency difference for cells receiving a
monosynaptic
(groups M and C,,,,) input
prepared to test this point, showed little sign
of correlation
between OR, and ORI-ORZ,
suggesting that the OR, latency is largely
independent
of the conduction
velocity
within the optic radiations.
Ordinul position und cdl types
Examination of the distribution of different cell types at various ordinal stages in
the processing sequence (Fig. 7) reveals that
no particular cell type is exclusively associated with a single ordinal position. On the
contrary, for almost all cell types, there are
examples of cells with almost identical response properties occurring at all ordinal
positions. This was a surprising result since
we had anticipated modifications in the receptive-field properties with progression
along the processing sequence. We were
prompted, therefore, to look for possible
sources of error in allocating the ordinal
position. First, had the cell failed to respond
with a monosynaptic spike because the electrode was inappropriately positioned? In
this case a monosynaptic input would be
overlooked and an incorrect classification
given to the cell if it otherwise met the
criteria for disynaptic activation. However,
it is unlikely that both OR, and OR2 electrodes would be out of position and that a
spike absent from stimulation at one site
NEURONS
1261
would also be absent from the other. Of the
139 cells that could be activated from both
OR, and ORB, only one cell showed an indication, in the form of an additional spike in
going from stimulation at OR, to ORz, that it
should be regarded as having a lower ordinal
position. Indeed, the majority of cells- 119
of the 139-showed exactly the same pattern of spikes when stimulated at OR, and
OR,. Then we also considered the possibility that a monosynaptic excitatory input was
masked by inhibition traveling in a more
rapidly conducting afferent pathway than
excitation. However, intracellular recordings of cortical cells suggest that they receive inhibitory activation disynaptically
from the thalamus (24, 33, 34). As a result,
inhibition is unlikely to precede monosynaptic excitation unless it travels much faster
over a long afferent pathway. By diminishing the distance of the afferent pathway with
the positioning of the OR, electrode, we
have excluded this possibility.
While the reasoning above eliminates the
chance of overlooking a monosynaptic input, there is some likelihood that a polysynaptic input may have been missed for
cells placed in the M or monosynaptic category. The disynaptically activated inhibition, referred to above, may arrive soon
enough to mask excitation taking a di- or
polysynaptic path to the cell. In consequence, in cells with evidence of inhibitory
activity in their visual responses, such as Sand &-cells, there may be more convergence of mono- with di- and polysynaptic
inputs (Cl+n) than indicated in our results.
Finally, in reviewing the effectiveness of
our methods, consideration should be given
to a shortcoming associated with electrical
stimulation. There is no indication, particularly in cells receiving a convergence of inputs, of the functional significance of one input relative to another. For example, for
cells receiving a mono- and disynaptic input
in the C,,,, group, a principal excitatory input cannot be recognized from a modulating
one and it should not be presumed that the
monosynaptic input is the key determinant
of the cell’s excitatory response.
Bearing in mind the points made above
there still can be little question that, for a
number of cell types (S, Sh, and C), there
are some examples of cells with a monosynaptic drive while there were others with a di-
J. BULLIER
AND
synaptic input. In those instances
where
groups of cells receiving a mono- or disynaptic input existed within a single cell type, we
could not find any evidence of systematic
differences in the receptive-field
properties
associated with an increase in ordinal position. This is contrary to the central theme
of the hierarchical
model, which proposes
that receptive-field
properties
increase in
the specificity
of their stimulus
requirements with progression along the processing
sequence. If the similarity in receptive-field
properties for cells in different ordinal position does not result simply from an inadequate assessment, then it is necessary to review the role of striate neurons as relaying
elements in a chain of excitatory events. It
is possible, for example, that a disynaptitally activated cell replicates the response
of a monosynaptically
driven cell because it
is acting as an interneuron
about to invert
the excitatory
response into an inhibitory
one. It is also possible that replication occurs when, for reference purposes, the response pattern of a monosynaptically
driven
cell is reproduced in cells, perhaps in other
laminae, lying off the main path taken by the
processing sequence. Under these circumstances the role of the disynaptically
driven
cell remains uncertain since we cannot tell
if it lies on the main processing
pathway
or along a secondary side band.
From our results, another difficulty for
the concept of serial processing arises in the
large number of cell types that appear to
receive a monosynaptic
input from the thalamus. We believe this diversity
of monosynaptically
driven neurons cannot be ascribed to an inappropriate
classification
based on differences in receptive-field
properties. There is now sufficient evidence to
suggest that B- and C-cells, at least, are performing different functional
roles from Sand &,-cells and from cells with nonoriented
and concentric receptive fields. It is possible, as mentioned above, that the mono-
G. H. HENRY
synaptic drive does not come from the principal input to the cell but from one which
exerts only a modulating influence. However, this seems unlikely in the majority of
S-, &-cells,
and cells with nonoriented or
concentric receptive field in laminae 4 or 6
where the principal drive would come as a
direct input from the dLGN. It is more difficult to exclude the possibility that a monosynaptic input to C-cells has only a modulating influence since the majority of C-cells
lie in lamina 5 (15), which does not receive
a direct input from the dLGN and commonly
receive a convergence
of inputs. Even allowing for such a contingency,
we are still
faced with a diversity of first-order
neurons
in the cat striate cortex. This finding may
not be so surprising
when consideration
is
given to the multiplicity
of efferents leaving the striate cortex. When a structure like
the dLGN appears to send at least three
efferent streams (X, Y, and W) to a single
destination, the striate cortex, we can begin
to envisage the multiplicity of streaming that
must occur within the striate cortex itself
to generate projections to a variety of cortical and subcortical centers. In light of these
considerations
and because of the wide variety of monosynaptically
driven striate neurons, we have undertaken
two additional
studies in this series. The first examines the
relationship that cortical processing streams
may have with the different divisions
of
dLGN input (4) and the second, the laminar
distribution
of cells occurring
at different
levels of processing
in different
cortical
streams (5).
ACKNOWLEDGMENTS
We express our thanks to Ken Collins for technical
assistance,
Bryson
Easton
for histological
preparations, and Jan Livingstone
and Josephine
Anderson
for secretarial
assistance.
Received
26 December
2 April 1979.
1978; accepted
in final
form
REFERENCES
1. BARLOW,
H. B., BLAKEMORE,
C., AND PETTIJ. D. The neural mechanism
of binocular
depth discrimination.
J. Yhysiol. Lo~~don 193: 327342, 1967.
2. BISHOP,
P. O., BURKE,
W., AND DAVIS,
R. Single
unit recording
from antidromically
activated
optic
GREW,
radiation
neurones.
J. Yhysiol.
London
162: 432450, 1962.
3. BISHOP, P. 0. AND HENRY,
G. H. Striate neurons:
receptive
field concepts.
Itwust.
Ophthalmol.
11:
346-354,
1972.
4. BULLIER,
J. AND HENRY,
G. H. Neural path taken
ORDINAL
POSITION
OF STRIATE
by afferent
streams
in striate cortex
of the cat.
J. Newophysiol.
42: 1264- 1270, 1979.
5. BULLIER,
J. AND HENRY,
G. H. Laminar
distribution of first-order
neurons
and afferent
terminals
in cat striate cortex. J. Newophysiol.
42: 1271- 128 1,
1979.
6. CLELAND,
B. G., DUBIN,
M. W., AND LEVICK,
W. R. Sustained and transient
neurones in the cat’s
retina and lateral geniculate
nucleus.
J. Physiol.
London
217: 473-496,
1971.
7. DENNEY,
D., BAUMGARTNER,
G., AND ADORJANI,
C. Responses
of cortical
neurones
to stimulation
of the visual afferent radiations.
Exp. Bruin Rccs. 6:
265-272,
21.
B. Hypercomplex
cells in the cat’s striate
Itwest.
Ophthulmol.
11: 355-356,
1972.
9. ECCLES,
J. C. The Physiology
oj’Synup.ws.
Berlin:
Springer,
1964, p. 37-43.
10. FINK,
B. R. AND SCHOOLMAN,
A. Arterial
blood
acid-base
balance
in unrestrained
waking
cats.
Proc. SW. Exp. Biol. Med. 112: 328330,
1963.
11. FULLER,
J. H. AND SCHLAG,
J. D. Determination
of antidromic
excitation
by the collision test: problems of interpretation.
Bruin Rus. 112: 283-298,
22.
23.
24.
25.
26.
27.
1976.
12. GASSER,
H. S. Unmedullated
dorsal root ganglia. J. Gen.
fibers
Physiol.
originating
33: 651-690,
in
1950.
C. Laminar
differences
in receptive
field
13. GILBER-I‘,
properties
of cells in cat primary
visual cortex. J.
Physiol.
London
268: 391-421,
1977.
14. HENRY,
G. H. Receptive
field classes of cells in
the striate cortex of the cat. Bruin Rcs. 133: l-28,
1977.
15. HENRY,
G. H., HARVEY,
A. R., AND LUND,
J. S.
The afferent
connections
and laminar distribution
of cells in the cat striate cortex. J. Comp. Ncwrol.
In press.
16. HENRY,
G. H., LUND,
J. S., AND HARVEY,
A. R.
Cells of the striate cortex projecting
to the ClareBishop area of the cat. Bruin Rus. 151: 154-158,
1978.
D. A. AND MITCHELI-,
R. A. Blood gas
tensions
and acid-base
balance in awake cats. J.
Appl. Physiol.
30: 434-436,
1971.
18. HODGKIN,
A. L. A note on conduction
velocity.
J. Phy.siol. London
125: 221-224,
1954.
19. HOFFMANN,
K.-P. AND SI‘ONE,
J. Conduction
velocity of afferents
to cat visual cortex:
a correlation with cortical
receptive
field properties.
Bruin
Rcs. 32: 460-466,
1971.
20. HUBEL,
D. H. AND WIESEL,
T. N. Receptive
fields
of single neurones
in the cat’s striate cortex.
J.
Phvsiol.
London
148: 574-591.
1959.
17.
HERBERT,
HUBEL,
binocular
the cat’s
1968.
8. DREHER,
cortex.
NEURONS
28.
D. H. AND WIESEL,
T. N. Receptive
fields,
interaction
and functional
architecture
in
visual cortex.
J. Physiol.
London
160:
106-154,
HUBEL,
1962.
229-289,
LEVICK,
1965.
D. H. AND WIESEL,
T. N. Receptive
and functional
architecture
in two nonstriate
areas ( 18 and 19) of the cat. J. Ncurophysiol.
fields
visual
28:
W. R. Another
tungsten
microelectrode.
Med. Biol. Eng. 10: 510-515,
1972.
LI, C. L., ORTIZ-GOLVIN,
A., CHOU,
S. N., AND
HOWARD,
S. Y. Cortical
intracellular
potentials
in
response to stimulation
of lateral geniculate
body.
J. Ncwophysiol.
23: 592-601,
1960.
RODIECK,
R. W., PETTIGREW,
J. D., BISHOP,
P.
O., AND NIKARA,
T. Residual
eye movements
in
receptive-field
studies
of paralyzed
cats. Vision
Res. 7: 107-I 10, 1967.
RUSHTON,
W. A. H. A theory of the effects of fibre
size in medullated
nerve. J. Physiol.
London
115:
101-122,
1951.
SCHILLER,
P. H., FINLAY,
B. L., ANDVOLMAN,
S.
F. Quantitative
studies of single-cell
properties
in
monkey
striate cortex.
I. Spatiotemporal
organization
of receptive
fields. J. Neurophysiol.
39:
1288-1319,
SILLITO,
1976.
A. M. Inhibitory
processes underlying
the
directional
specificity
of simple,
complex
and
hypercomplex
cells in the cat’s visual cortex.
J.
Physiol.
London
271: 699-720,
1977.
W., TRETTER,
F., AND CYNADER,
M. Or29. SINGER,
ganization
of cat striate cortex:
a correlation
of
receptive-field
properties
with afferent and efferent
connections.
J. Neurophysiol.
38: lOSO- 1098, 1975.
30. STONE, J. AND DREHER,
B. Projection
of X- and Ycells of the cat’s lateral geniculate
nucleus to areas
17 and 18 of visual cortex.
J. New-ophysiol.
36:
551-567,
1973.
31. SWADLOW,
H. A., WAXMAN,
S. G., AND ROSENE,
D. L. Latency
variability
and the identification
of
antidromically
activated
neurons
in mammalian
brain. Exp. Bruin Rcs. 32: 439-443,
1978.
32 TOYAMA,
K., MAEKAWA,
K., ANDTAKEDA,
T. An
analysis
of neuronal
circuitry
for two types of
visual cortical
neurones
classified
on the basis of
their responses
to photic stimuli.
Bruin Res. 61:
33
395-399,
TOYAMA,
TOKASHIKI,
organization
21: 45-66,
WATANABE,
1973.
K.,
MATSUNAMI,
K., OHNO,
T., AND
S. An intracellular
study of neuronal
in the visual cortex.
Exp. Bruin Rcs.
1974.
S., KONISHI,
M., AND CREU-I‘ZFELDT,
0. Postsynaptic
potentials
in the cat’s visual cortex
following
electrical
stimulation
of afferent
pathways. Exp. Bruin. Res. 1: 272-282,
1966.