Download Synchrony between Neurons with Similar Muscle Fields in Monkey

Neuron, Vol. 38, 115–125, April 10, 2003, Copyright 2003 by Cell Press
Synchrony between Neurons with Similar
Muscle Fields in Monkey Motor Cortex
Andrew Jackson,1 Veronica J. Gee,
Stuart N. Baker,2 and Roger N. Lemon*
Sobell Department of Motor Neuroscience
and Movement Disorders
Institute of Neurology
UCL, London WC1N 3BG
United Kingdom
Summary
Synchronous firing of motor cortex cells exhibiting
postspike facilitation (PSF) or suppression (PSS) of
hand muscle EMG was examined to investigate the
relationship between synchrony and output connectivity. Recordings were made in macaque monkeys performing a precision grip task. Synchronization was
assessed with cross-correlation histograms of the activity from 144 pairs of simultaneously recorded neurons, while spike-triggered averages of EMG defined
the muscle field for each cell. Cell pairs with similar
muscle fields showed greater synchronization than
pairs with nonoverlapping fields. Furthermore, cells
with opposing effects in the same muscles exhibited
negative synchronization. We conclude that synchrony
in motor cortex engages networks of neurons directly
controlling the same muscle set, while inhibitory connections exist between neuronal populations with opposing output effects.
Introduction
Synchronous neural activity in the cerebral cortex has
been the focus of much recent interest. For instance,
synchronous spike discharge in the visual cortex has
been implicated in the mechanism of perceptual binding
(von der Malsburg, 1981; Eckhorn et al., 1988; Singer
and Gray, 1995). The “binding hypothesis” proposes that
V1 neurons become synchronized when they encode
information pertaining to the same object in a visual
scene. This synchrony increases the salience of related
features at a higher processing level. It has also been
suggested that assemblies of synchronously active neurons play a more widespread role in the processing of
information by the brain. If synchronous EPSPs exert a
greater influence over a postsynaptic neuron than asynchronously arriving inputs, then synchrony could provide a mechanism by which information can be reliably
transmitted through a network of weak synaptic connections (Abeles, 1982, 1991; Vaadia et al., 1995; Alonso et
al., 1996; Singer et al., 1997). These issues can now
be addressed due to developments in multielectrode
*Correspondence: [email protected]
1
Present address: Department of Physiology and Biophysics, University of Washington, Seattle, Washington 98195.
2
Present address: Department of Anatomy, University of Cambridge,
Cambridge CB2 3DY, United Kingdom.
recording techniques (Eckhorn and Thomas, 1993; Nicolelis et al., 1997; Baker et al., 1999a).
Synchrony has also been observed in the discharge
of neurons in primary motor cortex (M1) (Murphy et al.,
1985; Smith and Fetz, 1989; Hatsopoulos et al., 1998;
Baker et al., 2001), and it is possible that synchronous
cortical activity may be more effective at driving
motoneurons (Baker et al., 1999b; Kilner et al., 2002).
The observation of cortico-muscular coherence at the
same frequency as cortical oscillations (Conway et al.,
1995; Baker et al., 1997; Kilner et al., 1999) suggests
that the temporal structure of synchronous cortical output can be transmitted to the spinal level. If synchrony
in M1 arises from the organization of cells into functional
assemblies, then one possibility is that the activity of
cells influencing common muscles could be synchronized together. Indeed, an early study by Smith (1989)
(reviewed by Fetz et al., 1990) provided evidence to
support such an organization. We have conducted a
systematic test of this hypothesis for pairs of M1 neurons exhibiting postspike facilitation (PSF) or suppression (PSS) of hand muscles in monkeys performing a
precision grip task.
The postspike effects of a single cortical neuron on
muscle activity can be revealed by spike-triggered averaging of rectified EMG (Fetz and Cheney, 1980; Lemon
et al., 1986). PSF typically occurs at a latency consistent
with a monosynaptic excitatory pathway from the cortex
to motoneurons (Fetz and Cheney, 1980; Lemon et al.,
1986; Baker and Lemon, 1998). Corticospinal projections originate from layer V of M1 and other motor areas,
and descend via the pyramidal tract to contralateral
spinal segments (Porter and Lemon, 1993). Corticomotoneuronal (CM) cells produce monosynaptic excitation of motoneurons, while corticospinal projections to
inhibitory interneurons cause postspike suppression of
muscles. Many CM cells produce postspike effects in
more than one muscle; the extent of these effects defines the target “muscle field” (Fetz and Cheney, 1980;
Buys et al., 1986).
Using multiple electrodes we were able to record simultaneously motor cortex neurons with both overlapping and nonoverlapping muscle fields. Synchrony between cells was assessed by cross-correlation analysis
(Perkel et al., 1967; Kirkwood, 1979; Baker et al., 2001).
Particular care was taken to ensure that the underlying
firing pattern of each neuron did not bias the measure
of synchronization.
We found that cell pairs with similar muscle fields
exhibited significantly greater synchronization than pairs
with nonoverlapping muscle fields. Furthermore, cells
with opposing effects in the same muscles exhibited
fewer synchronous spikes than would be expected by
chance, referred to as “negative synchronization.” These
results could not be explained by a disproportionate
contribution of synchronous spikes to STA effects. We
conclude that synchrony arises within networks of CM
cells with common target muscles and that inhibitory
mechanisms act between neurons with opposing effects
in the same muscles.
Neuron
116
Figure 1. Cross-Correlation Histograms of
Cell Activity
(A) Cross-correlation histogram (CCH) for a
synchronous pair of PTNs (cell L236p5:
35,000 spikes; cell L236p6: 38,400 spikes).
Dashed line shows predicted correlation calculated from instantaneous firing rate estimates.
(B) Percentage excess of observed counts
above predicted. Dashed lines indicate 95%
significance; dark shading indicates ⫾5 ms
of zero lag.
(C and D) Equivalent analysis for a PTN pair
exhibiting negative synchronization (cell
L205p2: 67,100 spikes; cell L205p5: 45,000
spikes).
(E) Distribution of synchronization values for
all cell pairs. Dark shading indicates crosscorrelations exhibiting a peak or trough
greater than 95% significance.
(F) Distribution of peak width at half-maximum (PWHM) of significant peaks (upward
bars) and troughs (downward bars).
(G) Distribution of time lags of significant
peaks/troughs.
Results
Two monkeys (macaca mulatta) performed a precision
grip task (Baker et al., 2001). These results are based
on 31 recording sessions (monkey [M] 33: 6; M36: 25)
in which at least two neurons with significant postspike
effects were simultaneously recorded from primary motor cortex. A total of 101 neurons (M33: 21; M36: 80, 84%
positively identified as pyramidal tract neurons [PTNs])
yielded 144 cell pairs. Analysis was performed on continuous sections of data typically lasting 30 min and incorporating around 300 trials. Between 9,700 and 136,500
spikes in total were recorded from each neuron (mean
44,200).
Cross-Correlation of Spike Trains
Figure 1A shows a cross-correlation histogram (CCH)
for a pair of simultaneously recorded PTNs. Because
the activity of both cells may covary in relation with the
task, we also calculated a cross-correlation predictor
using estimates of the instantaneous firing rate of each
cell (IFR) (Pauluis and Baker, 2000; Baker et al., 2001).
This is shown by the dashed line in Figure 1A and takes
into account correlated firing rate modulations but not
the precise timing of spikes. Figure 1B shows that the
percentage excess of observed CCH counts above this
predictor exhibits a significant central peak. An excess
of 18% in the zero-lag bin fell to below half this value
within one bin width (2 ms). Synchronization was defined
as the mean excess within ⫾5 ms of zero lag; the value
in this case of 6% was typical for the cell pairs we
analyzed. In addition to positive synchrony, we also observed CCH troughs indicative of negative synchrony.
Figures 1C and 1D shows one such cell pair. In this case
the synchronization value was ⫺7%, reflecting fewer
synchronous spikes between ⫾5 ms than would be expected by chance.
Figure 1E shows the distribution of synchronization
values obtained from all cell pairs used in this study
(mean, 2.4%; SD, 5.6%; n ⫽ 144). 80% of CCHs exhibited
a significant central peak or trough. The shading in Figure 1E indicates the synchronization values for these
cell pairs. Note that the CCH peaks for some cell pairs
with low synchronization values nevertheless reached
significance. This is partly because significance testing
was applied to individual bins, while the synchronization
value represents the average of the five central bins. In
addition, the significance level will depend upon the
number of spikes used to compile the CCH, and if this
is large, then small effects can reach significance.
The peak width at half-maximum (PWHM) and time
lag of significant CCH peaks and troughs are shown in
Figures 1F and 1G, respectively. Note that many of these
effects were quite narrow, although the width at the base
of the peak was typically two to four times the PWHM
(c.f. Figure 1B). Most maxima occurred in the central
bin (⫾1 ms), whereas the distribution of minima was
offset by 2–4 ms (Figure 1G). We also observed symmetrical CCH troughs with minima either side of zero lag
(c.f. Figure 1C). These differences in timing might suggest that positive synchronization arises mainly from
common inputs to each cell, whereas negative synchronization involves serial inhibitory connections (Perkel
et al., 1967). In this study, no attempt was made to
differentiate between types of cross-correlation effect.
In the subsequent analysis, only the synchronization
value as defined above was used, irrespective of the
significance, width, or timing of peaks and troughs. The
range of ⫾5 ms of zero lag was chosen since it covers
the large majority of effects. The advantages of this
method are that no assumptions are made about the
Synchrony and Muscle Field in Motor Cortex
117
Figure 2. Spike-Triggered Averages of EMG
(A) Muscle fields for two cells exhibiting positive synchrony (same pair as Figure 1A).
Dashed line shows time of trigger spike. Vertical scale represents percentage excess
above a prediction based on IFR estimates.
Asterisks indicate significant effects at 95%
level calculated from the prespike period. “S”
indicates effects which reached significance
but were discarded according to the PWHM
criterion.
(B) Muscle fields for two cells exhibiting negative synchronization (Figure 1C).
(C) Distribution of the latencies of peak PSF
(upward bars) and PSS (downward bars).
Shading indicates effects in intrinsic and extrinsic hand muscles.
(D) Distribution of the PWHM of PSF and PSS
effects.
shape of effects, and that the measure is unbiased by
total spike count. In addition, 5 ms is close to the membrane time constant of motoneurons (Zengel et al.,
1985), so this measure should have some relevance to
the physiological consequences of synchrony at the spinal level.
Spike-Triggered Averaging of EMG
Postspike effects were identified from spike-triggered
averages (STAs) of rectified EMG. As with the cell-cell
cross-correlations, these cell-EMG averages were normalized and expressed as percentage excess above a
predictor calculated from the IFR estimate. The size
of the muscle field, as determined by the number of
significant effects at the 95% level, was between one
and six muscles per CM cell (average M33: 1.4; M36:
3.2), although since a restricted set of muscles were
sampled, this is likely to underestimate the true extent.
Figure 2A shows all STAs for two cells exhibiting positive synchronization (same pair as Figure 1A). Both cells
showed similar facilitation of EDC and ECR. Figure 2B
shows muscle fields for two cells exhibiting negative
synchronization (Figure 1C). Note that these cells produce
opposing postspike effects in AbPL, EDC, and ECR. The
second cell also facilitated FDP, FDS, and AbPB.
Two quantities were used to characterize each effect:
the latency of peak facilitation (or suppression) and the
PWHM. Figure 2C shows the latencies of all PSF and
PSS effects. Effects in intrinsic hand muscles had a
slightly longer latency (mean, 11.0 ms) compared with
extrinsic hand muscles (mean, 10.0 ms; two-sample t
test, p ⫽ 0.0004), consistent with a longer peripheral
conduction distance. In addition, PSF had a shorter latency (mean, 10.1 ms) than PSS (mean, 10.9 ms; two-
sample t test, p ⫽ 0.03) probably reflecting transmission
via an additional inhibitory synapse (Jankowska et al.,
1976; Kasser and Cheney, 1985; Maier et al., 1998).
The PWHM of postspike effects is relevant since it
provides one means of separating genuine facilitation
from effects which can arise if a non-CM cell fires synchronously with CM cells facilitating the same muscle
(Hamm et al., 1985; Smith and Fetz, 1989; Kirkwood,
1994). Baker and Lemon (1998) used a computer simulation to conclude that the duration of PWHM provided a
good criterion for accepting PSF effects as genuine.
Accordingly, 12/208 PSF effects (6%) with a PWHM ⬎
7 ms were discarded as potentially arising from synchrony. Examples of excluded PSFs can be seen for
muscle AbPL in Figure 2A. Three cells were removed
from the analysis as a result of all their postspike effects
being judged as caused by synchrony.
Baker and Lemon (1998) discussed two potential
PWHM criteria (7 or 9 ms). In choosing the more stringent
of these, it is possible that some genuine effects may
have been excluded. However, for the purposes of the
present study, this was considered preferable to the
inclusion of synchrony effects which could potentially
bias the results. Figure 2D shows the PWHM distribution
for accepted PSF and PSS effects. Note that no criterion
was applied to the PWHM of PSS effects; these would
be expected to be wider due to the additional inhibitory
synapse. Therefore the 10% of PSS effects with a PWHM
greater than 7 ms were included in the analysis.
Relationship between Synchronization and
Postspike Effects
To determine whether synchrony between a pair of CM
cells was related to the similarity between the two mus-
Neuron
118
Figure 3. Example of a Synchronized Cell Pair with Overlapping Muscle Field
(A) Spike-triggered averages of rectified EMG from 1DI and FDP for two PTNs (cell L03p7: 47,000 spikes; cell L03p8: 52,000 spikes). No
significant effects were observed with other EMGs.
(B) Vector representation of the significant postspike effects of each cell. Muscle field divergence is defined as the angle between vectors.
(C) CCH for the same cell pair indicates synchronous discharge.
(D) Typical finger and thumb lever position traces for a single trial and average firing rate profile for these cells (average of 430 trials aligned
to the start of the hold period).
cle fields, the postspike effects of each cell were represented as a vector (see Experimental Procedures). This
allowed us to quantify the similarity between two muscle
fields as the angle between two vectors. This muscle
field divergence can have values between 0⬚ and 180⬚,
where 0⬚ indicates identical effects for the two cells, 90⬚
indicates mainly nonoverlapping (orthogonal) effects,
and 180⬚ indicates opposing effects (PSF versus PSS)
in the same muscles. Figure 3 describes the calculation
for one cell pair. Both neurons exhibited PSF effects
in 1DI; cell L03p7 also facilitated FDP (Figure 3A). No
significant effects were observed in the remaining five
muscles for either cell. Therefore, the muscle field divergence for these cells was 45⬚ (Figure 3B). A positive
CCH peak (Figure 3C) indicated that the discharges of
this cell pair were synchronized. Both cells’ discharges
showed marked modulation during the precision grip
task (Figure 3D).
Figures 4A and 4B plots the relationship between the
muscle field divergence and the synchronization (as
quantified from the CCHs) for all cell pairs. A significant
negative correlation was observed with both animals
(M33: n ⫽ 29, Pearson’s r ⫽ ⫺0.57, p ⫽ 0.001; M36: n ⫽
115, r ⫽ ⫺0.55, p ⬍ 0.0001). Cell pairs with positive
synchronization tended to have similar postspike effects
(divergence ⬍ 90⬚), while pairs exhibiting negative synchronization values had opposing effects (divergence ⬎
90⬚). This result was confirmed by dividing the cell pairs
into three groups according to the muscle field divergence: similar, orthogonal, and opposing. Figure 4C
shows the mean synchronization for each group. Similar
cell pairs had a mean synchronization of 5.3%, whereas
for opposing cell pairs it was ⫺2.0%. Both groups were
significantly different from the orthogonal group (mean
synchronization, 1.7%; two-sample t test: similar-orthogonal, p ⫽ 0.0003; opposing-orthogonal, p ⫽ 0.0001).
The total number of postspike effects produced by
both cells was calculated for all pairs. This combined
muscle field size was not significantly correlated with
either synchronization or the absolute magnitude of synchronization (n ⫽ 144, Pearson’s r ⫽ ⫺0.11 and ⫺0.03,
respectively). Therefore, synchrony between CM cells did
not influence the size of the muscle field for each cell.
Task Relationship of Synchronous Neurons
CM neurons tend to exhibit either phasic activity during
movement, tonic activity during the hold period, or a
combination of both (Cheney and Fetz, 1980; Maier et al.,
1993). Figure 3D shows the average firing rate profiles for
two cells with overlapping muscle fields. The profiles
are aligned to the start of the hold period as defined
by both finger and thumb traces entering the target
displacement window. Despite clear synchrony between these cells (Figure 3C), inspection of Figure 3D
reveals different task modulation. Cell L03p7 exhibited
a phasic burst during movement with declining activity
during the hold period; cell L03p8 showed no phasic
component but was active tonically throughout the hold
period. This suggests that synchrony can be observed
between neurons with similar postspike effects but different task dependence.
The similarity between the task modulation of each
cell pair was quantified by calculating the correlation
coefficient between firing rate profiles (compiled from
1 s before to 2 s after the hold period start with a bin
width of 0.1 s). Figure 4D shows this firing profile correlation plotted against synchronization. As has been reported by previous studies (Smith, 1989; Fetz et al.,
1990; Pinches, 1999), there is a wide scatter of points,
with a small but significant positive correlation between
firing rate similarity and synchronization (data from both
monkeys pooled: n ⫽ 144, r ⫽ 0.20, p ⫽ 0.02). Since
Synchrony and Muscle Field in Motor Cortex
119
Figure 4. Relationship between Synchronization and Muscle Field Divergence
(A) Scatter plot of muscle field divergence
(defined as the angle between two muscle
field vectors, see Experimental Procedures)
against synchronization for cell pairs recorded in M33.
(B) Equivalent plot for cell pairs recorded in
M35.
(C) Mean synchronization for all cell pairs
grouped by muscle field divergence: similar
(⬍90⬚, n ⫽ 67), orthogonal (⫽90⬚, n ⫽ 39),
and opposite (⬎90⬚, n ⫽ 38). Data from both
animals pooled.
(D) Similarity of task modulation (defined as
the correlation coefficient between firing rate
profiles) plotted against synchronization for
all cell pairs.
other studies have reported an association between the
firing pattern during movement and the muscle field of
some CM cells (Cheney et al., 1991; Bennett and Lemon,
1996), a pair of cells with similar task modulation might
be expected also to have similar muscle fields. However,
this association cannot explain the relationship between
synchronization and muscle field divergence. Having
accounted for the similarity of task modulation using
partial correlation, the relationship between synchronization and muscle field divergence remains statistically
significant (n ⫽ 144, rpartial ⫽ ⫺0.54, p ⬍ 0.0001). However,
the partial correlation coefficient between similarity of
task modulation and synchronization, accounting for the
effect of muscle field divergence, is not significant (n ⫽
144, rpartial ⫽ 0.13, p ⫽ 0.12). Therefore, this weak correlation between synchronization and similarity of task modulation may be explained by a stronger interaction with
muscle field similarity, combined with an association
between firing patterns and muscle fields.
Effect of Electrode Separation
One possible explanation for the correlation between
synchronization and muscle field similarity could be covariation of both factors with the distance between neurons. Neighboring neurons might be expected to exhibit
greater synchrony (Fetz et al., 1990; Abeles, 1991; Matsumura et al., 1996; Hatsopoulos et al., 1998) and project
to similar muscles (Cheney and Fetz, 1985; Lemon et
al., 1987). Estimating the neuronal separation as the
distance between the tips of the recording electrodes,
we found that synchronization was significantly correlated with electrode separation (r ⫽ ⫺0.35, p ⫽ 0.001).
Figure 5A shows the mean and standard deviation of
synchronization for cell pairs grouped according to electrode separation. The range of observed synchronization values is greatest for the closest pairs, and the
mean value declines with increasing separation. However, there was no correlation between muscle field divergence and electrode separation (r ⫽ ⫺0.02, p ⬎ 0.2)
as shown in Figure 5B. Therefore, over the range of
interneuronal separations sampled here, distance does
not seem to influence the probability that the two neurons will have overlapping muscle fields.
Double Spike-Triggered Averaging of EMG
Another possible explanation for the correlations in Figures 4A and 4B is that the synchronous spikes from a
Figure 5. Relationship between Synchronization and Electrode
Separation
(A) Mean and standard deviation of synchronization for cell pairs
grouped according to electrode separation (300–500 ␮m, 500–700
␮m, etc.). Both the mean and the range of observed synchronization
values decreases with increasing separation.
(B) Mean and standard deviation of muscle field divergence of cell
pairs, again grouped according to separation. There is no effect of
separation on the probability of muscle field overlap.
Neuron
120
Figure 6. Double Spike-Triggered Averages of EMG
(A) Method of double spike-triggered averaging. Time relative to each spike train runs along separate axes (t1, t2). Each spike pair contributes
to the average along a diagonal, offset from the main diagonal according to the interspike interval.
(B) Typical double spike-triggered average for two cells which exhibited PSF effects on muscle EDC. The CCH for this pair is shown in Figure
1A. The facilitation caused by each cell is separated and averaged along the different axes.
(C) Linear sum of average postspike effects closely matched the observed interaction.
(D) Double spike-triggered average for a cell pair with opposing effects on muscle ECR. The CCH for this pair is shown in Figure 1C.
(E) Linear sum of average postspike effects.
pair of CM cells could contribute disproportionately to
the PSF observed in STAs. This might arise either from
nonlinearities in the response of motoneurons to synchronous EPSPs, or as a result of higher order synchrony between M1 neurons (Martignon et al., 1995,
2000). The latter would be characterized by correlated
firing among a group of cells such that some of the
synchronous spikes from any cell pair would coincide
with spikes from the wider population of CM cells. In
either case, we would expect that if a pair of cells exhibits a CCH synchrony peak, then the STAs for each cell
would be similar since the same synchronous events
would be dominating both averages. We assessed this
possibility by compiling EMG averages triggered by a
pair of CM cell spikes.
Figure 1A shows the CCH for a synchronized cell pair.
Both these cells exhibited PSF effects in EDC (Figure
2B). Figure 6B shows the double spike-triggered average of this muscle for the same cell pair. In this average,
time relative to each spike train runs along different axes
(t1, t2; see Figure 6A). Thus the PSF caused by each
cell independently can be separated from the effect of
synchronous spike pairs. If it is only synchronous spikes
which contribute to PSF effects, then facilitation should
mainly be observed along the diagonal t1 ⫽ t2. Figure
6B indicates that this is not the case. The vertical band
of facilitation around time t1 ⫽ 10 ms shows the PSF
caused by cell 1 (L236p5). The horizontal band at t2 ⫽
10 ms is the PSF by cell 2 (L236p6). Plotted below and
to the left of Figure 6B are these responses, averaged
over the vertical (t2) and horizontal (t1) dimensions, respectively. Cell 1 individually produces a maximum facilitation of 8%, for cell 2 the peak is 7%. Where the two
bands intersect shows the response to a synchronous
spike pair (t1 ⫽ t2 ⫽ 10 ms). At this point there is a
15% modulation above baseline, suggesting that the
facilitation due to the synchronous discharge of these
cells is similar to the linear sum of the individual effects.
Figure 6C shows the prediction of the linear model for
the double STA of this cell pair, calculated by recombining the one-dimensional averages in Figure 6B (see Experimental Procedures).
Splitting the STA into two dimensions in this way is
computationally time consuming and severely reduces
the signal-to-noise ratio. Therefore, double STAs were
compiled only for a subset of 39 cell-cell-muscle combinations with large spike numbers and clear facilitation
of the muscle by both cells individually. For this sample,
the value of the double STA was compared with the
prediction of linear summation at the time lags appropriate for maximum facilitation by each cell. The result
is shown in Figure 7. Most points fall around the dotted
line representing an observed facilitation by synchronous spikes equal to that predicted by a linear summation of individual effects. This was true for cell pairs
evoking both small and quite large effects, with the mean
(⫾ standard error) ratio of observed to predicted facilitation equal to 1.04 ⫾ 0.05. Therefore, on the basis of
Synchrony and Muscle Field in Motor Cortex
121
ferred directions (Georgopoulos et al., 1993; Lee et al.,
1998).
Figure 7. Facilitation by Synchronous Spikes versus the Linear Sum
of Individual Effects
Plot of double-spike triggered average versus the prediction of the
linear summation model at the time lags appropriate for maximum
facilitation by each cell individually. Dotted line represents facilitation due to a synchronous pair of spikes that equals the linear sum
of the facilitation caused be each cell separately. Points above
this line represent supralinear effects. Data from 39 cell-cell-EMG
combinations.
these data, there was no statistically significant difference between the facilitation by synchronous pairs of
spikes from cells which share a target muscle and an
equivalent number of unsynchronized spikes. However,
the 95% confidence range for this ratio encompasses
values between 0.94 and 1.14; thus, small nonlinearities
would not be resolved by this method.
Figure 6D shows a double spike-triggered average for
two cells with opposing effects in muscle ECR (Figure
2B). The cross-correlation for this pair is shown in Figure
1C. Once again, the double STA reveals the effect of
each cell independently and that the combined effect
is well described by a linear sum (Figure 6E).
Discussion
Synchronous Networks within the Motor Cortex
This study has investigated synchrony among motor
cortex neurons causing postspike facilitation or suppression of muscle activity. CM cells active during precision grip have restricted muscle fields (Buys et al., 1986;
Maier et al., 1993; Bennett and Lemon, 1996). We found
that the greatest synchronization was observed between pairs of cells with overlapping muscle fields. Cells
with nonoverlapping muscle fields exhibited less synchrony. Furthermore, cells with opposing effects in the
same muscles had negative synchronization, implying
fewer synchronous spikes than would be expected by
chance. Thus, synaptic connections mediating synchrony exist predominantly between CM cells sharing
the same output connectivity. This system of organization could underlie the synchronization of cells with similar task modulation (Fetz et al., 1990; Pinches, 1999),
including during reaching movements for which correlations are strongest for pairs of neurons with similar pre-
Influence of Synchrony on the Magnitude
of Postspike Effects
One issue already raised is the effect of synchrony on
the magnitude of postspike facilitation and suppression
as revealed by STAs (Fetz and Cheney, 1980; Smith and
Fetz, 1989; Kirkwood, 1994). In the most extreme case,
a cell with no physiological connection to motoneurons
could exhibit PSF if it is active in synchrony with genuine
CM cells. More significantly, synchrony effects could
have contributed to the main finding of this study, the
correlation between synchrony and muscle field overlap
(Figure 4). Several approaches to resolving this problem
can be considered. First, the size of muscle field was
not correlated with synchronization, suggesting that
synchrony between the recorded neurons is not responsible for a significant number of spurious postspike effects. Second, by using the PWHM criterion set out in
Baker and Lemon (1998), we discarded some postspike
effects that may have arisen from synchrony alone. This
criterion is based on the logic that synchrony PSF will
result from genuine effects convolved with the shape of
the cross-correlation peak between cells, resulting in a
wider profile. It should be pointed out however that the
PWHM criterion is based on a computer simulation depending critically on several factors, particularly the
width of the CCH peaks between cells. Classification of
genuine and synchrony effects was found to be adequate for cross-correlation widths (at half-maximum)
down to at least 2.8 ms (Baker and Lemon, 1998), but
for some of the narrower CCH peaks summarized in
Figure 1F, separation of postspike effects based on
PWHM may be unreliable.
A third approach was developed by Smith and Fetz
(1989) who selectively removed spikes to eliminate
cross-correlation peaks before compiling spike-triggered averages. For a cell pair exhibiting 21% synchronization (around the upper limit of the values found in this
study), the authors concluded that this accounted for
only around 10% of the PSF magnitude. Furthermore,
the effect of the CCH peak was assessed by convolution
with the PSF produced by the synchronous cell. Such
approaches assume linear summation of the facilitation
by each cell individually, and the absence of higher order
synchrony. The validity of these assumptions can now
be qualified using the double STA results described
here. First, in all cases the double STA revealed that it
was not only the synchronous spikes producing
postspike facilitation (Figure 6). Despite this, the contribution of synchrony to PSF magnitudes could affect the
proportion of effects reaching significance, producing
a spurious correlation between synchronization and
muscle field overlap. This situation will be exacerbated
if synchronous spikes contribute disproportionately to
the PSF. However the double STA revealed that at worst
(within the 95% confidence range of the data) the effect
of synchronous spikes was 1.14 times the effect of asynchronous spikes. Incorporating this nonlinearity into the
convolution method will influence the result by only
14%. Simulations using typical spike trains recorded
from this study showed that for a synchronization of
Neuron
122
6%, only 1.5% of the PSF from one cell appears in
the STA for the second cell, and such effects cannot
account for the strength of correlation observed in Figure 4 (see Supplementary Material at http://www.neuron.
org/cgi/content/full/38/1/115/DC1). Nevertheless, pairwise synchrony with other cells in the population may
exaggerate the observed facilitation by each cell independently. The extent of this cannot be assessed from
this study, since it depends on the total number of CM
cells within each assembly, which cannot be determined
from the present data.
Functional Significance of Synchrony
In sensory systems it has been proposed that synchrony
may provide a solution to the problem of “binding” lowlevel features encoded by individual neurons into more
complex high-level representations of objects. A similar
combinatorial problem can be envisaged in the motor
system. For example, if features of a movement are
represented by the distributed activity of a population
of neurons, a synchrony code could help to distinguish
and organize the processing of information pertaining
to individual muscles. Although corticospinal projections to target motoneurons are anatomically constrained, it is possible that the synchrony observed between CM cells reflects “motor binding” at a level higher
than the CM cell output, or has functions at the other
targets influenced by CM cells through their collaterals
at spinal and supraspinal levels (Ugolini and Kuypers,
1986).
Some theories propose that synchronous presynaptic
activity should exert a greater influence over postsynaptic neuronal discharge than asynchronous activity
(Abeles 1982; Singer et al., 1997). We attempted to test
this for CM projections by using double STAs to compare
the facilitation caused by synchronous and asynchronous pairs of spikes. Because of extensive branching
of CM cells within the motor nucleus of the target muscle
(Porter and Lemon, 1993), we would expect considerable convergence in the projections to motoneurons
from CM cell pairs which share postspike effects. We
observed no significant difference between the facilitation caused by synchronous spikes and the linear sum
of the effect of each cell individually, suggesting that
synchronous activity may be no more effective at driving
motoneurons than the same number of spikes occurring
asynchronously. Unitary EPSPs from CM cells are probably of small amplitude and derived from a small number
of synaptic contacts (Lawrence et al., 1985). It has been
shown that pairs of small EPSPs arriving at separate
dendrites on the same motoneuron sum approximately
linearly (Prather et al., 2001), although this may not be
true for larger or more numerous potentials arriving synchronously (Porter and Hore, 1969; Poliakov et al., 1997;
Matthews, 1999). Our result is consistent with the finding
of no interaction between the responses to ICMS pulses
delivered simultaneously to separate M1 sites (Baker
et al., 1998). Nonlinearities have been observed in the
muscle response evoked by two spikes from the same
CM cell separated by a short interval (Fetz and Cheney,
1980; Lemon and Mantel, 1989), but this involves sequential activation of the same synapses on target
motoneurons and probably invokes different biophysi-
cal mechanisms than pairs of inputs spatially separated
on the motoneuron’s dendrites.
Even without nonlinear summation of motoneuron
EPSPs, widespread higher order synchrony among large
populations of CM cells would be expected to produce
an apparent facilitation which is greater than the sum
of individual effects for any cell pair. However, the large
background of synchronous spikes occurring by chance
would dilute any effect of genuine higher order synchrony events in the average. To illustrate this for a CM
cell pair with a typical synchronization value, we could
suppose that any increased facilitation revealed by the
double STA arises from just 6% of the total number
of synchronous spike pairs. A ratio of synchronous to
asynchronous facilitation of 1.14 would be consistent
with these synchronous spike pairs being accompanied
by on average an extra 5.7 spikes from the entire population of CM cells facilitating that muscle. Although it
should be emphasized that this represents the upper
bound consistent with these results, and that the observed mean ratio of 1.04 leads to a more modest 1.3
additional spikes, it therefore remains a possibility that
small assemblies of CM cells could exhibit higher order
synchrony. In relation to this, it is interesting to note
that Martignon et al. (2000) concluded that higher order
interactions in frontal areas did not seem to be frequent.
In addition, Baker and Lemon (2000) found that repeating temporal patterns of spikes from three or more
neurons (of which higher order synchrony is a special
case) occurred at chance levels within the motor cortex.
Synchronization between PTNs within the cortex
might be due to shared inputs, or arise from recurrent
networks of intracortical axon collaterals (Landry et al.,
1980; Ghosh and Porter, 1988; Jackson et al., 2002). In
either case, it is clear that synchronization can occur
between cells whose activity relates to the same muscles but different features of a movement (Figure 3).
Synchrony between such cells could result from the
necessity to coordinate activity during different phases
of a movement. Furthermore, modulation of synchrony
in the motor cortex during motor tasks (Vaadia et al.,
1995; Riehle et al., 1997; Baker et al., 2001) suggests
systematic variations in the degree of synchronization
within or between network assemblies required for
specific motor activities. This may encode movementrelated information (Hatsopoulos et al., 1998) and possibly account for changes in the size and form of postspike
effects which can occur across and within tasks (Buys
et al., 1986; Bennett and Lemon, 1996).
Finally, the negative synchronization observed between cells with opposing effects in the same muscles
could be explained if some of the cells we recorded
were interneurons, inhibiting CM cells. This would lead
to a cross-correlation trough and, potentially, postspike
suppression of muscles facilitated by the inhibited cells.
However, since the majority of cells were identified as
PTNs, the most likely explanation is that suppression
of activity in networks with opposing postspike effects
occurred via axon collaterals of the recorded cells projecting to inhibitory interneurons (Renaud and Kelly,
1974; Thomson et al., 1995).
The connectivity underlying positive and negative
synchronization may serve to form discrete dynamic
assemblies of motor cortex output neurons with com-
Synchrony and Muscle Field in Motor Cortex
123
mon target muscles. Different assemblies might reflect
the enormous repertoire of skilled hand movements in
primates as a task-related pattern of muscle combinations (Hughlings Jackson, 1874; Turton et al., 1993).
Experimental Procedures
Behavioral Task
Two purpose-bred adult female monkeys (m. mulatta) weighing 6.1
kg (M33) and 5.0 kg (M36) were used for this study. The precision
grip task (see Baker et al., 2001) involved squeezing two springloaded levers between thumb and index finger. A successful trial
required a movement into target followed by a hold period of 1 s
before release. Auditory cues were given when the levers were within
target and at the end of the hold period.
Recording
Details of surgical procedures and the Eckhorn multiple-electrode
recording system (Thomas Recording Ltd., Marburg, Germany) have
been described by Baker et al. (1999a, 2001). EMG was recorded
from muscles flexor digitorum profundus (FDP), flexor digitorum
sublimis (FDS), extensor digitorum communis (EDC), extensor carpi
radialis (ECR), abductor pollicis longus (AbPL), abductor pollicis
brevis (AbPB), and first dorsal interosseous (1DI) in both animals.
In M33, we additionally recorded from adductor pollicis (AdP) and
abductor digiti minimi (AdDM). In M36, all EMGs were recorded
using implanted patch electrodes (Miller et al., 1993); in M33, a
combination of implanted, surface, and needle electrodes was used.
Cross-talk between EMG recordings was not more than 20%; for
most pairs it was negligible. Cortical recordings were made in the
hand area of M1, contralateral to the performing hand; both hemispheres were investigated in both animals. Chamber center coordinates were A13 L18 (M33) and A10 L17 (M36). Up to 14 glassinsulated platinum electrodes (impedance 1–3 M⍀, 4 ⫻ 4 grid with
interelectrode spacing of 300 ␮m) were independently lowered into
the cortex to search for cells. PTNs were identified by their antidromic response to PT stimulation (latencies, 0.9–4 ms; thresholds,
20–200 ␮A) and collision testing (Lemon, 1984). Off-line, single units
were discriminated using principal component analysis on the spike
waveform and cluster cutting (Eggermont, 1990).
All procedures were performed in accordance with appropriate
UK Home Office regulations.
Analysis
Synchronization
Synchrony between pairs of simultaneously recorded neurons was
assessed by compiling cross-correlation histograms with a bin width
of 2 ms, which were compared with a predictor calculated from
instantaneous firing rate (IFR) estimates for each cell (Pauluis and
Baker, 2000; Baker et al., 2001). Synchronization was quantified as
the percentage excess of observed counts above expected counts
averaged across the five central bins (⫾5 ms of zero lag). In addition,
we tested the significance of peaks and troughs, assuming Poisson
distributed bin counts. Peaks and troughs which exceeded 95%
significance were characterized according to timing and peak width
at half-maximum (PWHM). As a precaution, cross-correlations for
which the PWHM equalled 1 bin width (2 ms) were recompiled with
a bin width of 0.5 ms to ensure that no peaks were precise at the
shorter time scale; this would have suggested an artifact contaminating both channels.
Muscle Field Divergence
Postspike facilitation (PSF) or suppression (PSS) of muscles was
identified from spike-triggered averages of rectified EMG sampled
at 5 kHz. These were compiled between 50 ms either side of the
trigger spike and normalized by the cross-correlation between IFR
and rectified EMG. Effects were considered significant if they exceeded twice the standard deviation of the prespike period and
satisfied the PWHM criterion discussed elsewhere. The muscle field
for each cell with at least one significant effect was defined as a
vector m where mi ⫽ ⫹1, ⫺1, or 0 for respectively a PSF, PSS, or
no effect in muscle i. For a pair of cells, the similarity between
muscle fields was assessed by calculating the divergence angle, ␪,
between the two muscle field vectors:
␪ ⫽ cos⫺1
m ·m
.
冢|m
||m |冣
1
1
2
(1)
2
Double Spike-Triggered Averaging
The contribution of synchrony to postspike effects was investigated
by averaging rectified EMG triggered by a pair of spikes from two
cells. A double spike-triggered average is a function of two dimensions X(t1,t2), with time relative to each spike train running along
separate axes t1, t2. Let the times of spikes from the two cells be
represented T i1, T j2. E(t), the level of rectified EMG at time t is then
averaged separately at every point for which a spike pair i, j satisfies:
T i1 ⫹ t1 ⫽ T j2 ⫹ t2 ⫽ t.
(2)
As a result, each spike pair contributes one sweep of EMG to the
average along the diagonal defined by Equation (2). The effect of
synchronous spikes is represented along the main diagonal (t1 ⫽
t2), while asynchronous pairs are offset from this (Figure 6A). Averages were expressed as the percentage excess above a predicted
average XIFR(t1,t2) calculated from IFR estimates for each cell F1(t ),
F2(t ):
冮E(t)F1(t ⫺ t1)F2(t ⫺ t2)dt
,
冮F1(t ⫺ t1)F2(t ⫺ t2)dt
(3)
Xobserved (t1,t2) ⫺ XIFR(t1,t2)
⫻ 100.
XIFR(t1,t2)
(4)
XIFR (t1,t2) ⫽
Xnorm(t1,t2) ⫽
If there is no interaction between the postspike effects of the individual cells, then this excess Xnorm should approximate to a linear sum
of the contribution from each cell independently:
Xlinear (t1,t2) ⫽ X1(t1) ⫹ X2 (t2) ⫹ c,
(5)
where X1(t1), X2 (t2) are obtained by averaging Xnorm(t1,t2) over the
independent dimension and c is a constant required to conserve
the overall mean.
To reduce computational time and improve signal-to-noise, double spike-triggered averaging was performed on rectified EMG
which had been smoothed and downsampled to 500 Hz.
Acknowledgments
The authors thank Daniel Wolpert, Peter Kirkwood, Rachel Spinks,
Thomas Brochier, Helen Lewis, and Lucy Anderson for their assistance. This work was funded by the Medical Research Council, The
Wellcome Trust, and the Human Frontiers Science Program.
Received: July 23, 2002
Revised: February 20, 2003
Accepted: March 6, 2003
Published: April 9, 2003
References
Abeles, M. (1982). Role of cortical neuron: integrator or coincidence
detector? Isr. J. Med. Sci. 18, 83–92.
Abeles, M. (1991). Corticonics: Neural Circuits of Cerebral Cortex
(Cambridge: Cambridge University Press).
Alonso, J.M., Usrey, W.M., and Reid, R.C. (1996). Precisely correlated firing in cells of the lateral geniculate nucleus. Nature 383,
815–819.
Baker, S.N., and Lemon, R.N. (1998). Computer simulation of postspike facilitation in spike-triggered averages of rectified EMG. J.
Neurophysiol. 80, 1391–1406.
Baker, S.N., and Lemon, R.N. (2000). Precise spatiotemporal repeating patterns in monkey primary and supplementary motor areas
occur at chance levels. J. Neurophysiol. 84, 1770–1780.
Baker, S.N., Olivier, E., and Lemon, R.N. (1997). Task-dependent
coherent oscillations recorded in monkey motor cortex and hand
muscle EMG. J. Physiol. 501, 225–241.
Neuron
124
Baker, S.N., Olivier, E., and Lemon, R.N. (1998). An investigation of
the intrinsic circuitry of the motor cortex of monkey using intracortical microstimulation. Exp. Brain Res. 123, 397–411.
Jankowska, E., Padel, Y., and Tanaka, R. (1976). Disynaptic inhibition
of spinal motoneurones from the motor cortex in the monkey. J.
Physiol. 258, 467–487.
Baker, S.N., Philbin, N., Spinks, R., Pinches, E.M., Wolpert, D.M.,
MacManus, D.G., Pauluis, Q., and Lemon, R.N. (1999a). Multiple
single unit recording in the cortex of monkeys using independently
moveable microelectrodes. J. Neurosci. Methods 94, 5–17.
Kasser, R.J., and Cheney, P.D. (1985). Characteristics of corticomotoneuronal post-spike facilitation and reciprocal suppression
of EMG activity in the monkey. J. Neurophysiol. 53, 959–978.
Baker, S.N., Kilner, J.M., Pinches, E.M., and Lemon, R.N. (1999b).
The role of synchrony and oscillations in the motor output. Exp.
Brain Res. 128, 109–117.
Baker, S.N., Spinks, R., Jackson, A., and Lemon, R.N. (2001). Synchronization in monkey motor cortex during a precision grip task.
I. Task-dependent modulation in single-unit synchrony. J. Neurophysiol. 85, 869–885.
Bennett, K.M., and Lemon, R.N. (1996). Corticomotoneuronal contribution to the fractionation of muscle activity during precision grip
in the monkey. J. Neurophysiol. 75, 1826–1842.
Kilner, J.M., Baker, S.N., Salenius, S., Hari, R., and Lemon, R.N.
(1999). Task-dependent modulation of 20–30 Hz coherence between
rectified EMGs from human hand and forearm muscles. J. Physiol.
516, 559–570.
Kilner, J.M., Alonso-Alonso, M., Fisher, R., and Lemon, R.N. (2002).
Modulation of synchrony between single motor neurons during precision grip in humans. J. Physiol. 541, 937–948.
Kirkwood, P.A. (1979). On the use of cross-correlation measurements in the mammalian central nervous system. J. Neurosci. Methods 1, 107–132.
Buys, E.J., Lemon, R.N., Mantel, G.W., and Muir, R.B. (1986). Selective facilitation of different hand muscles by single corticospinal
neurones in the conscious monkey. J. Physiol. 381, 529–549.
Kirkwood, P.A. (1994). The identification of corticomotoneuronal
connections. In Movement Control, P. Cordo and S. Harnad, eds.
(Cambridge, United Kingdom: Cambridge University Press), pp.
164–165.
Cheney, P.D., and Fetz, E.E. (1980). Functional classes of primate
corticomotoneuronal cells and their relation to active force. J. Neurophysiol. 44, 773–791.
Landry, P., Labelle, A., and Deschenes, M. (1980). Intracortical distribution of axonal collaterals of pyramidal tract cells in the cat motor
cortex. Brain Res. 191, 327–336.
Cheney, P.D., and Fetz, E.E. (1985). Comparable patterns of muscle
facilitation evoked by individual corticomotoneuronal (CM) cells and
by single intracortical microstimuli in primates: evidence for functional groups of CM cells. J. Neurophysiol. 53, 786–804.
Lawrence, D.G., Porter, R., and Redman, S.J. (1985).
Corticomotoneuronal synapses in the monkey: light microscopic
localization upon motoneurons of intrinsic muscles of the hand. J.
Comp. Neurol. 232, 449–510.
Cheney, P.D., Fetz, E.E., and Mewes, K. (1991). Neural mechanisms
underlying corticospinal and rubrospinal control of limb movements.
Prog. Brain Res. 87, 213–252.
Lee, D., Port, N.L., Kruse, W., and Georgopoulos, A.P. (1998). Variability and correlated noise in the discharge of neurons in motor
and parietal areas of the primate cortex. J. Neurosci. 18, 1161–1170.
Conway, B.A., Halliday, D.M., Shahani, U., Maas, P., Weir, A.L., and
Rosenberg, J.R. (1995). Synchronization between motor cortex and
spinal motoneuronal pool during the performance of a maintained
motor task in man. J. Physiol. 489, 917–924.
Lemon, R.N. (1984). Methods for neuronal recording in conscious
animals. In IBRO Handbook Series: Methods in Neurosciences, Volume 4, A.D. Smith, ed. (London: Wiley), pp. 1–162.
Eckhorn, R., and Thomas, U. (1993). A new method for the insertion
of multiple microprobes into neural and muscular tissue, including
fiber electrodes, fine wires, needles and microsensors. J. Neurosci.
Methods 49, 175–179.
Eckhorn, R., Bauer, R., Jordan, W., Brosch, M., Kruse, W., Munk,
M., and Reitboeck, H.J. (1988). Coherent oscillations: a mechanism
for feature linking in the visual cortex? Biol. Cybern. 60, 121–130.
Lemon, R.N., and Mantel, G.W. (1989). The influence of changes in
discharge frequency of corticospinal neurones on hand muscles in
the monkey. J. Physiol. 413, 351–378.
Lemon, R.N., Mantel, G.W., and Muir, R.B. (1986). Corticospinal
facilitation of hand muscles during voluntary movement in the concious monkey. J. Physiol. 381, 497–527.
Eggermont, J.J. (1990). The Correlative Brain. Theory and Experiment in Neural Interaction (Berlin: Springer).
Lemon, R.N., Muir, R.B., and Mantel, G.W. (1987). The effects upon
the activity of hand and forearm muscles of intracortical stimulation
in the vicinity of corticomotor neurones in the concious monkey.
Exp. Brain Res. 66, 621–637.
Fetz, E.E., and Cheney, P.D. (1980). Post-spike facilitation of forelimb
muscle activity by primate corticomotoneuronal cells. J. Neurophysiol. 44, 751–772.
Maier, M.A., Bennett, K.M., Hepp-Reymond, M.C., and Lemon, R.N.
(1993). Contribution of the monkey corticomotoneuronal system to
the control of force in precision grip. J. Neurophysiol. 69, 772–785.
Fetz, E.E., Toyama, K., and Smith, W.S. (1990). Synaptic interactions
between cortical neurons. In Cerebral Cortex, Volume 9, A. Peters,
ed. (New York: Plenum), pp. 1–47.
Maier, M.A., Illert, M., Kirkwood, P.A., Nielsen, J., and Lemon, R.N.
(1998). Does a C3-C4 propriospinal system transmit corticospinal
excitation in the primate? An investigation in the macaque monkey.
J. Physiol. 511, 191–212.
Georgopoulos, A.P., Taira, M., and Lukashin, A. (1993). Cognitive
neurophysiology of the motor cortex. Science 260, 47–52.
Ghosh, S., and Porter, R. (1988). Morphology of pyramidal neurones
in monkey motor cortex and the synaptic actions of their intracortical
axon collaterals. J. Physiol. 400, 593–615.
Hamm, T.H., Reinking, R.M., Roscoe, D.D., and Stuart, D.G. (1985).
Synchronous afferent discharge from a passive muscle of the cat:
significance for interpreting spike-triggered averages. J. Physiol.
365, 77–102.
Hatsopoulos, N.G., Ojakangas, C.L., Paninski, L., and Donoghue,
J.P. (1998). Information about movement direction obtained from
synchronous activity of motor cortical neurons. Proc. Natl. Acad.
Sci. USA 95, 15706–15711.
Hughlings Jackson, J. (1874). On the anatomical and physiological
localisation of movements in the brain. In Selected Writings of John
Hughlings Jackson, Volume 1, J. Taylor, ed. (London: Staples), pp.
37–76.
Jackson, A., Spinks, R.L., Freeman, T.C.B., Wolpert, D.M., and
Lemon, R.N. (2002). Rhythm generation in monkey motor cortex
explored using pyramidal tract stimulation. J. Physiol. 541, 685–699.
Martignon, L., von Hasseln, H., Grün, S., Aertsen, A., and Palm, G.
(1995). Detecting higher order interactions among the spiking events
of a group of neurons. Biol. Cybern. 73, 69–81.
Martignon, L., Deco, G., Laskey, K., Diamond, M., Freiwald, W., and
Vaadia, E. (2000). Neural coding: higher-order temporal patterns in
the neurostatistics of cell assemblies. Neural Comput. 12, 2621–
2653.
Matsumura, M., Chen, D., Sawaguchi, T., Kubota, K., and Fetz, E.E.
(1996). Synaptic interactions between primate precentral cortex
neurons revealed by spike-triggered averaging of intracellular membrane potentials in vivo. J. Neurosci. 16, 7757–7767.
Matthews, P.B.C. (1999). The effect of firing on the excitability of a
model motoneurone and its implications for cortical stimulation. J.
Physiol. 518, 867–882.
Miller, L.E., van Kan, P.L.E., Sinkjaer, T., Andersen, T., Harris, G.D.,
and Houk, J.C. (1993). Correlation of primate red nucleus discharge
with muscle activity during free-form arm movements. J. Physiol.
469, 213–243.
Murphy, J.T., Kwan, H.C., and Wong, Y.C. (1985). Cross correlation
Synchrony and Muscle Field in Motor Cortex
125
studies in primate motor cortex: synaptic interaction and shared
input. Can. J. Neurol. Sci. 12, 11–23.
Nicolelis, M.A.L., Fanselow, E.E., and Ghazanfar, A.A. (1997). Hebb’s
dream: the resurgence of cell assemblies. Neuron 19, 219–221.
Pauluis, Q., and Baker, S.N. (2000). An accurate measure of the
instantaneous discharge probability, with application to unitary
joint-event analysis. Neural Comput. 12, 647–669.
Perkel, D.H., Gerstein, G.L., and Moore, G.P. (1967). Neuronal spike
trains and stochastic point processes. II. Simultaneous spike trains.
Biophys. J. 7, 419–440.
Pinches, E.M. (1999). The contribution of population activity in motor
cortex to the control of skilled hand movements in the primate.
Ph.D. thesis, Institute of Neurology, University College London, London, UK.
Poliakov, A.V., Powers, R.K., and Binder, M.D. (1997). Functional
identification of the input-output transforms of motoneurones in the
rat and cat. J. Physiol. 504, 401–424.
Porter, R., and Hore, J. (1969). Time course of minimal
corticomotoneuronal excitatory postsynaptic potentials in lumbar
motoneurons of the monkey. J. Neurophysiol. 32, 443–451.
Porter, R., and Lemon, R.N. (1993). Corticospinal Function and Voluntary Movement (Oxford: Clarendon Press).
Prather, J.F., Powers, R.K., and Cope, T.C. (2001). Amplification and
linear summation of synaptic effects on motoneuron firing rate. J.
Neurophysiol. 85, 43–53.
Renaud, L.P., and Kelly, J.S. (1974). Simultaneous recordings from
pericruciate pyramidal tract and non-pyramidal tract neurons; response to stimulation of inhibitory pathways. Brain Res. 79, 29–44.
Riehle, A., Grün, S., Diesmann, M., and Aertsen, A. (1997). Spike
synchronization and rate modulation differentially involved in motor
cortical function. Science 278, 1901–1902.
Singer, W., and Gray, C.M. (1995). Visual feature integration and the
temporal correlation hypothesis. Annu. Rev. Neurosci. 18, 555–586.
Singer, W., Engel, A.K., Kreiter, A.K., Munk, M.H.J., Neuenschwander, S., and Roelfsema, P.R. (1997). Neuronal assemblies:
necessity, signature and detectability. Trends Cogn. Sci. 1, 252–261.
Smith, W.S. (1989). Synaptic interactions between identified motor
cortex neurons in the active primate. PhD thesis, University of Washington, Seattle, Washington.
Smith, W.S., and Fetz, E.E. (1989). Effects of synchrony between
corticomotoneuronal cells on post-spike facilitation of primate muscles and motor units. Neurosci. Lett. 96, 76–81.
Thomson, A.M., West, D.C., and Deuchars, J. (1995). Properties
of single axon excitatory postsynaptic potentials elicited in spiny
interneurons by action potentials in pyramidal neurons in slices of
rat neocortex. Neuroscience 69, 727–738.
Turton, A., Fraser, C., Flament, D., Werner, W., and Bennett, K.M.
and Lemon, R N. (1993). Origanization of cortico-motoneuronal projections from the primary motor cortex: evidence for task-related
function in monkey and man. In Spasticity—Mechanisms and Management, A. Thilman, D.J. Burke, and W.Z. Rymer, eds. (Berlin:
Springer), pp. 8–24.
Ugolini, G., and Kuypers, H.G.J.M. (1986). Collaterals of corticospinal
and pyramid fibres to the pontine grey demonstrated by a new
application of the fluorescent fibre labelling technique. Brain Res.
365, 211–227.
Vaadia, E., Haalman, I., Abeles, M., Bergman, H., Prut, Y., Slovin,
H., and Aertsen, A. (1995). Dynamics of neuronal interactions in
monkey cortex in relation to behavioural events. Nature 373,
515–518.
von der Malsburg, C. (1981). The correlation theory of brain function.
MPI Biophysical Chemistry, Internal Report 81–82. Reprinted in
Models of Neural Networks II (1994), E. Domany, J.L. van Hemmen,
and K. Schulten, eds. (Berlin: Springer).
Zengel, J.E., Reid, S.A., Sypert, G.W., and Munson, J.B. (1985).
Membrane electrical properties and prediction of motor-unit type
of medial gastrocnemius motoneurons in the cat. J. Neurophysiol.
53, 1323–1344.