䉬

Human Brain Mapping 13:226 –238(2001) 䉬

Early Discrimination of Coherent versus
Incoherent Motion by Multiunit and Synaptic
Activity in Human Putative MTⴙ
Istvan Ulbert,3,4 George Karmos,3 Gary Heit,4 and Eric Halgren1,2*
1

Massachusetts General Hospital NMR Center, Harvard Medical School,
Charlestown, Massachusetts
2
INSERM E9926, Marseilles, France
3
Institute for Psychology, Hungarian Academy of Sciences, Budapest, Hungary
4
Department of Neurosurgery, Stanford University, Stanford, California

䉬

䉬

Abstract: A laminar probe was chronically implanted in human putative MT⫹. The area was specifically
responsive to globally coherent visual motion, a crucial aspect of the perception of movement through space.
The probe contained 23 microcontacts spaced every 175␮ in a linear array roughly perpendicular to the cortical
surface. Current-source density (CSD) and multiunit activity (MUA) were recorded while viewing initially
stationary random dot patterns that either moved incoherently or dilated from the central fixation. Onset of
visual motion evoked large MUA/CSD activity, with coherent motion evoking earlier and faster-rising
MUA/CSD activity than incoherent, in both superficial and deep pyramidal layers. The selective response,
peaking at ⬇115 ms, was especially large in deep pyramids, providing evidence that information necessary for
visual flow calculations is projected from MT⫹ at an early latency to distant structures. The early onset of
differential MUA/CSD implies that the selectivity of this area does not depend on recurrent inhibition or other
intrinsic circuitry to detect coherent motion. The initially greater increase of MUA to coherent stimuli was
followed by a greater decrease beginning at ⬇133 ms, apparently because of recurrent inhibition. This resulted
in the total MUA being greater to incoherent than coherent stimuli, whereas total rectified CSD was overall
greater to coherent than to incoherent stimuli. However, MUA distinguished stationary from moving stimuli
more strongly than did CSD. Thus, while estimates of total cell firing (MUA), and of total synaptic activity
(CSD) generally correspond to previously reported BOLD results, they may differ in important details. Hum.
Brain Mapping 13:226 –238, 2001. © 2001 Wiley-Liss, Inc.
Key words: cerebral cortex; neocortex; occipital lobe; visual cortex; electroencephalography; vision;
evoked potentials; visual perception; motion perception
䉬

䉬

INTRODUCTION
Contract grant sponsors: Human Frontiers Science Program Organization (RG0025/1996), National Institutes of Health (NS 18741),
National Foundation for Functional Brain Imaging (DOE grant DEFG03-99ER62764), and Office of Naval Research.
Received for publication 4 February 2000; Accepted 30 March 2001.
*Correspondence to: Eric Halgren, Ph.D., MGH-NMR Center, Rm.
2301, Bldg. 149, 13th St., Charlestown, MA 02129.
E-mail: halgren@nmr.mgh.harvard.edu

© 2001 Wiley-Liss, Inc.

In the human, moving visual stimuli have been
found to evoke specific hemodynamic activation in a
small area at the dorsolateral occipitotemporal junction, using positron emission tomography (PET)
[Watson et al., 1993] and functional magnetic resonance imaging (fMRI) [Tootell et al., 1995; Beauchamp
et al., 1997]. This area is termed MT⫹ because it ap-

䉬 Human Synaptic/Unit Responses to Visual Flow 䉬

pears to be homologous to area MT (or V5) in the
monkey, plus the adjacent smaller area MST. Neurons
in monkey MT fire specifically to moving visual stimuli [Maunsell and VanEssen, 1983]. Anatomically,
MSTd follows MT in the dorsal visual processing
stream [Ungerleider and Desimone, 1986; Boussaoud
et al., 1990], and its cells also respond to moving visual
stimuli. However, compared to MT, MSTd neurons
are especially sensitive to the coherence of visual motion [Tanaka et al., 1986; Duffy and Wurtz, 1991; Graziano et al., 1994; Geesaman et al., 1997]. For example,
a set of dots that suddenly start expanding radially
outward evokes much greater activity than would a
set of dots in the same locations that begin to move at
the same velocity and average direction but at random
orientations to the radial [Albright, 1989; Tanaka et al.,
1989; Lagae et al., 1994]. Radial motion of visual stimuli commonly happens when we move through space,
looking ahead, and is termed “optic flow.” The responsiveness of MSTd to such patterns suggests that it
may assist in visual navigation by calculating the direction of heading with respect to the direction of gaze
[Lappe and Duffy, 1999; Andersen et al., 2000].
This possibility has been elaborated in quantitative
models in which MSTd calculates optic flow by combining inputs from MT, which is posited to detect local
motion [Lappe et al., 1996; Perrone and Stone, 1998].
For example, several MT cells, each sensitive to local
retinal movement away from a given point, are hypothesized to converge on a particular MSTd cell that
would then fire selectively when the animal moves
toward that location. Under this model, the net synaptic activity to coherent vs. incoherent motion could
be the same, but increased convergence to coherent
stimuli would result in an increased initial multiunit
response. An alternative hypothesis notes that, although MT shows much less selectivity for coherent
stimuli than MSTd, some MT cells do exhibit that
property [Lagae et al., 1994]. This raises the possibility
that such cells selectively project to MSTd, and thus
that selectivity for coherent motion may be present in
synaptic as well as multiunit activity from the earliest
onset of excitation.
In a third model, the sensitivity to global motion
would be produced by using reciprocal inhibition to
integrate local motion with location. In MT, cells excited by motion in a particular direction and location
tend to be inhibited by movements in other directions
in surrounding locations [Bradley and Andersen,
1998]. In an expanding dot pattern, dots that share
receptive fields would tend to move in the same directions, and thus would inhibit each other less than if
they were moving randomly with respect to each

䉬

other, thus explaining the larger MSTd activation to
coherently moving stimuli. Note that in this model,
coherent and incoherent stimuli would initially activate MSTd cells equally, and the responses to incoherent stimuli would then be attenuated by inhibition via
an interpolated interneuron after a delay.
In the current study, we attempted to distinguish
between these alternative models by measuring neuronal activity from putative human MT⫹ during coherent vs. incoherent motion. PET and fMRI are not
able to resolve this issue because they integrate neural
activity over 1– 60s, and thus cannot resolve the onset
of activation. Furthermore, PET and fMRI integrate
activation over 3– 8 mm, and thus do not resolve different cortical layers. Finally, because PET and fMRI
appear to measure a consequence of energy consumption, they do not distinguish between synaptic currents and neuronal action potentials. Laminar recordings from multiple closely spaced microcontacts have
been used to characterize the synaptic currents and
action potentials underlying cortical processing in animals [Mitzdorf, 1985; Schroeder et al., 1998; Sukov
and Barth, 1998]. We believe that this report is the first
use of laminar recordings during behavioral testing in
a human. It was found that the selective activation by
coherent stimuli in putative human MT⫹ is present
from its earliest activation. The laminar technique may
also be applied profitably to studying the synaptic
bases of other types of cognitive processing in humans.
METHODS
Subject
The subject was a 41-year-old right-handed woman
of normal intelligence and personality with a 25-year
history of medically intractable complex partial seizures. She participated after fully informed consent,
monitored by the Stanford IRB and according to NIH
guidelines conforming to the Declaration of Helsinki.
Spontaneous seizures began with an aura of spatial
distortion of the room and/or distortions of observed
faces. MRI was normal. An incidental finding of the
intracarotid sodium amytal (Wada) test was a left
peri-clinoid internal carotid artery aneurysm, which
was embolized with an endovascular coil. Ictal SPECT
showed increased uptake in the right temporal region,
and scalp EEG at ictal onset was localized to the right
occipito-temporal region. On the basis of these data,
surgical therapy was planned to involve the right
occipito-temporal junction. As an aid to definition of
the margins of the resection, a subdural grid and two

227 䉬

䉬 Ulbert et al. 䉬

subdural strips were implanted in the region of the
focus for 4 days (Fig. 1A). Intracranial EEG of spontaneous seizure onset, supplemented by intraoperative
ECOG, permitted a tailored resection of the right oc-

cipito-temporal area. The patient remains seizure-free
at 12-month follow-up. No abnormalities were observed in the pathology specimen.
Electrodes

Figure 1.
Location of laminar probes. A. Intraoperative photograph showing
placement of the laminar probe (arrow) underneath the subdural
grid of macrocontacts, at the lateral occipito-temporal junction.
The laminar probe is in the middle occipital gyrus (mOg), just
posterior to the middle temporal gyrus (mTg). mTg is identified in
relation to the superior temporal gyrus (sTg). B. Sagittal preoperative MRI showing the laminar probe’s location (arrow) in the lip
of the lateral occipital sulcus, near its intersection with an ascending branch of the inferior temporal sulcus (corresponding to the
anterior occipital sulcus). C. Photomicrograph of the electrode
track, with lines separating layers I/II, III/IV, and IV/V. Luxol fast
blue.

䉬

A laminar electrode was placed under visual observation in a location within the planned resection. The
electrode consisted of a 400␮ diameter polyimide tube
with twenty-three 40␮ diameter 90%Pt–10%Ir contacts
spaced at 175␮ center-to-center, arrayed in a line perpendicular to the cortical surface. In gross anatomical
terms, the laminar probe was located in the lip of the
lateral occipital sulcus, near its intersection with the
inferior temporal sulcus (Fig. 1B). In Talairach [1988]
coordinates, the laminar probe was located at ⫹50
(right), ⫺60 (posterior) ⫹3 (up). This is 9 –16 mm
anterior to the center of motion-selective activation
previously found for human MT⫹ using PET
(42,⫺69,0) [Watson et al., 1993] and fMRI (45,⫺76,3)
[Tootell et al., 1995]. In these studies, the center of
activation had a standard deviation of about 8 mm in
the anterior-posterior dimension, and the entire MT⫹
activation was about 10 –15 mm in each dimension.
Thus, the laminar probe appeared to be located at the
anterior edge of MT⫹, i.e., in the proposed location of
MSTd. Because MT neurons may also display a preference for globally coherent motion, it is possible that
the electrode was located in MT. Thus, we will refer to
the probe location as MT⫹, and the conclusions
should be construed as referring to how sensitivity to
coherent motion arises within this general area.
The cortical tissue surrounding the electrode was
removed en bloc at the definitive surgery. Alternate
sections were stained with luxol fast blue,
Bielschowsky, and haematoxilin and eosin, in order to
identify cortical laminae using standard criteria (Fig.
1C). A silastic disk was attached to the top of the
laminar probe. This disk was held against the pial
surface by the subdural grid, thus fixing mechanically
the distance from the pial surface to the first contact,
and allowing the microcontacts to be approximately
located with respect to cortical lamina.
Recordings
Differential recordings were made from 22 pairs of
successive contacts. After wideband (DC-10,000 Hz)
preamplification (gain 10⫻, CMRR 90db, input impedance 1012 ohms), the signal was split into field potentials (filtered at 0.2–500 Hz, gain 1,000⫻, digitized at
1,000 Hz, 16 bit) and multiunit activity (MUA: filtered
at 200 –5,000 Hz, gain 1,000⫻, digitized at 10,000 Hz,

228 䉬

䉬 Human Synaptic/Unit Responses to Visual Flow 䉬

Figure 2.
Visual stimuli. Left. Randomly placed dots in the coherent stimuli moved radially away from the
center of the display. Right. Dots in incoherent stimuli moved in random directions but with the
same average starting points and directions.

12 bit), and stored continuously with stimulus markers.
Estimated current source densities (CSD) at electrode contacts 3–20 (j) were calculated as:
CSDj ⫽ 共.23共ij ⫺ ij⫺1兲 ⫹ .54共ij⫹1 ⫺ ij兲
⫹ .23共ij⫹2 ⫺ ij⫹1兲兲/h

(1)

where ij is the estimated current between electrode
contacts j and j-1, and h is the intercontact distance,
0.175 mm. The ij are calculated from differences in the
field potentials v (i.e., the output of a differential amplifier channel) divided by the interelectrode tissue
resistance rh (equal to the resistance r, assumed to be
homogeneous and equal to 20 ohms/mm [Nicholson
and Freeman, 1975; Mitzdorf, 1985], multiplied by the
intercontact distance):
ij ⫽ 共vj ⫺ vj⫺1兲/rh

(2)

Combining formula (1) and (2), one arrives at the
standard five-point Hamming filter estimate for CSD
[Rappelsberger et al., 1981]:
CSDj ⫽ 共.23vj-2 ⫹ .08vj⫺1 ⫺ .62vj
⫹ .08vj⫹1 ⫹ .23vj⫹2兲/rh2

(3)

CSD at electrode contacts 2 and 21 were calculated by
applying the same formula after estimating the potentials at (pseudo-) contacts 0 and 23 using linear extrapolation [Vaknin et al., 1988]. MUA was obtained by

䉬

further filtering (300 –3,000 48 db/oct, zero phase
shift) the signal in a range found to detect actionpotentials in previous studies [Grover and Buchwald,
1970; Halgren et al., 1977; Legatt et al., 1980]. A continuous estimate of MUA was derived by rectifying,
and then low-pass filtering (50 hz digital, 24 db/oct
zero phase shift) the signal. The field potential gradient was also filtered (1– 40 Hz, 24 db/oct zero phase
shift), prior to calculating the CSD. Appropriate artifact rejection was performed using amplitude threshold criteria on field potential gradient and MUA. All
steps of the analysis were performed using Neuroscan
Edit 4.1.1 software (Neuroscan, Inc.). Details of electrode construction, recordings and analysis can be
found in Ulbert et al. [2001].
Rationale of CSD/MUA analysis
CSD identifies possible synaptic generators using
high-resolution maps of field potentials within the
generating structure [Schroeder et al., 1998; Sukov and
Barth, 1998]. The difference in potential between adjacent recording contacts is equal to the current flowing between them times the resistance. In cortical
structures, current flows in the tangential direction
can be assumed to be symmetrical (at least to the first
approximation), and thus to cancel each other out.
Consequently, changes in current in the radial dimension imply local passage of current across the neuronal
membrane [Nicholson and Freeman, 1975]. EPSPs result from the passage of positive charge from the
extracellular to the intracellular space, termed a sink in

229 䉬

䉬 Ulbert et al. 䉬

CSD. Surrounding the sink are passive sources of current. In contrast, typical hyperpolarizing IPSPs result
from active current sources, and are surrounded by
passive sinks. Active and passive transmembrane current flows can be distinguished using simultaneous
MUA. For example, an increase in MUA suggests that
simultaneous sinks locate active excitatory synapses,

and simultaneous sources locate passive return currents. The origin of these synapses may then be inferred from the known anatomical distribution of different pathways across cortical laminae. Although it is
theoretically possible that axonal action potentials (especially in terminal arborizations), dendritic spikes,
nonsynaptic active currents, and even glial currents
could contribute to CSD measures; these sources are
considered unlikely to contribute significantly in the
current experimental protocol [for discussion, see
Mitzdorf, 1985; Schroeder et al., 1995].
Coherent vs. incoherent motion

Figure 3.
Population synaptic and unit activity to moving and stationary
random dot patterns A. Population synaptic activity is measured
as current-source density (CSD) at 21 locations, separated by
175␮ intervals, spanning the cortical laminae. Recordings are displayed with the most superficial location at top. Activation is
earlier and stronger to coherently moving stimuli (thick lines) than
to incoherently moving (thin lines). B. Population unit activity is
measured as multiunit activity (MUA) from the same contacts.
Again, coherent motion evokes a larger and earlier response,
especially in the cortical layers containing superficial pyramids
(level 9). The MUA response in the deep pyramidal layer tends to
be biphasic to coherent stimuli and monophasic to incoherent.
The vertical bar indicates a t value of 4.0, compared to baseline,
corresponding to a P ⬍ 0.0001.

䉬

At the beginning of each trial, 20 white dots (each
subtending 0.15° of visual angle) were randomly distributed on a black background in a region subtending
3.5° vertical by 7.0° horizontal (Fig. 2). Between trials,
fixation was maintained on a white dot in the center of
that region. After remaining stationary for 533 ms, the
dots moved for 373 ms at 6.7°/sec and then remained
stationary for another 533 ms. On the 160 trials with
incoherent motion, the dots moved in random directions; on the 160 trials with coherent motion, the dots
moved radially away from the center (i.e., the pattern
dilated). The onsets of successive trials were separated
by 3,500 ms. The subject was instructed to observe the
stimuli attentively, and avoid moving her eyes or
blinking during the trials.
Separate averages were made of single sweep CSD
and single sweep MUA from trials with coherent vs.
incoherent motion, after rejecting artifacts using an
amplitude criterion. Statistical comparisons were
made using data from the ⬇120 individual trials remaining after artifact rejection. At each latency point,
t tests (Neuroscan Edit 4.1.1) were used to compare
trials with coherent motion to those with incoherent
motion, as well as each condition to the baseline period. To provide protection against type 1 errors resulting from multiple comparisons across channels
and latencies, an uncorrected statistical significance
threshold of P ⬍ 10-5 was used.
RESULTS
Response to stationary visual stimuli
The responses are presented as waveforms (Fig. 3)
and as spatiotemporal color-coded maps (Fig. 4). Onset of the stationary visual pattern evoked a small but
significant increase in MUA in the deep pyramidal
layer, with an onset latency of ⬇100 ms and peak
latency of ⬇140 ms (see red contour lines in channels

230 䉬

䉬 Human Synaptic/Unit Responses to Visual Flow 䉬

17–19 of Fig. 4D, beginning ⬇400 ms prior to stimulus
movement). Stationary patterns also evoked a CSD
sink, in layer IV and a source-sink pair in the deep
pyramidal layers V and VI (black-white contour lines,
Fig. 4D). The deep MUA/CSD response beginning at
⬇100 ms is followed by a strong superficial cortical
CSD source that begins at ⬇140 ms and peaks at ⬇190
ms (black contours, channels 2– 4, at –300 ms, Fig. 4D).
Like the deep source, the superficial source is followed
by a sink beginning at ⬇370 ms peaking at ⬇450 ms
after stationary stimulus onset (white contours, channels 2– 4, at –100 ms, Fig. 4D). The superficial CSD
sink/source, as well as the deep sink/source, are not
associated with significant changes in MUA (Fig. 4D).
However, the total MUA across all layers shows a
tendency to decrease during this period (Fig. 5A).
Response to coherent visual motion
The onset of visual motion by formerly stationary
stimuli evoked a large MUA and CSD response, especially when the motion was coherent (Figs. 3–5).
Again, the most prominent MUA response was located in the deep pyramidal layers, but a response was
now also present in the superficial pyramidal layers.
Coherent visual motion evoked a strong MUA increase in the deep pyramidal layer, beginning at ⬇110
ms, peaking at ⬇120 –150 ms, and persisting weakly
throughout the period of stimulation (thick lines,
channels 17–21, Fig. 3B, and red contours, Fig. 4A).
Coherent motion also evoked a significant CSD sink in
layer IV (white contours, Fig. 4A), surrounded by
sources in layers III and V-VI (black contours, Fig. 4A).
In addition, a later but prominent CSD sink is present
in layer VI. This source-sink-source-sink pattern began
at ⬇105 ms, peaked at ⬇140 ms, with portions of it
persisting throughout the period of visual motion. The
temporal association of increased MUA with the CSD
sink indicates that the sink represents active excitatory
population synaptic currents. The initial sink was centered in layer IV, followed by the layer VI sink.
Coherent visual motion also evoked a weak but
significant MUA increase in the superficial pyramidal
layer, peaking at ⬇145 ms (red contours, channel 9,
Fig. 4A). This firing was temporally correlated with a
superficial layer III sink, surrounded by sources in
layers I–II and deep layer III. These MUA and CSD
responses in the superficial pyramidal layer were all
transient, and did not persist beyond 200 ms after
motion onset. Finally, coherent motion evoked a sustained source in layer I, beginning at ⬇135 ms and
continuing throughout the stimulation period (black
contours, channels 1–3, Fig. 4A).

䉬

Comparison of phasic responses to coherent vs.
incoherent motion
The overall pattern of the CSD/MUA response to
incoherent motion was similar to that evoked by coherent, except that the phasic response was weaker
and delayed, whereas the sustained response was
more robust. These differences can be examined by
comparing the thick (coherent) and thin (incoherent)
lines in Figures 3A and 3B, by comparing Figure 4A
(coherent) to 4B (incoherent), and by examining Figure
4C, which shows the pattern obtained by subtracting
the CSD/MUA response to incoherent motion from
that to coherent.
The decreased phasic response is seen most clearly
in the deep pyramidal layer. Specifically, the deep
pyramidal MUA increase to incoherent visual motion
did not begin until ⬇120 ms, and did not peak until
⬇140 –160 ms (red contours, channels 17–21, Fig. 4B).
Similarly, the deep pyramidal layer source evoked by
incoherent motion did not begin until ⬇110 ms, and
did not peak until ⬇180 ms (black contours, channels
17–21, Fig. 4B). The direct comparison of deep pyramidal responses indicates that coherent motion evokes
a larger CSD response beginning at ⬇100 ms, and a
larger MUA response beginning at ⬇110 ms, with
coherent/incoherent differences peaking at about 120
ms (channels 19 –21, Fig. 4C).
The CSD/MUA response in the superficial pyramidal layer also appeared later and weaker to incoherent
stimuli, at ⬇200ms (MUA- red contours in channel 8;
CSD source, black contours in channels 9 –10; CSD
sink, white contours in channels 6 – 8; Fig. 4B). Note
that the superficial pyramidal MUA and CSD responses are consistently delayed from those in the
deep pyramidal layers. However, the deep and superficial responses appear to be linked, in that when the
deep response is weaker (i.e., between incoherent vs.
coherent motion), the superficial response is also, and
when the deep response is delayed (again, between
incoherent vs. coherent motion), the superficial response is also.
Comparison of sustained responses to coherent
vs. incoherent motion
In striking contrast to the decreased phasic response
to incoherent as compared to coherent motion, the
sustained MUA response was greater to incoherent
motion. This is especially evident in channels 17–22,
where initially (from ⬇100 –150 ms) the MUA activity
is larger to coherent motion, and then (from ⬇160 –260
ms) the MUA activity is larger to coherent motion

231 䉬

䉬 Ulbert et al. 䉬

(thick vs. thin lines; Fig. 3B, 4C). At ⬇200 ms, for example, the firing of the deepest pyramids to coherent motion is seen to actually fall below baseline, whereas their
firing to incoherent motion remains high.

At the same time that the deep pyramidal MUA is
showing this increased sustained response to incoherent motion, the deep pyramidal CSD source/sink is
unchanged or indeed (after ⬇250 ms) is somewhat

Figure 4.

䉬

232 䉬

䉬 Human Synaptic/Unit Responses to Visual Flow 䉬

larger to coherent motion. Yet, as noted above, the deep
pyramidal CSD source/sink appears to represent the
population EPSP that is causing the deep pyramidal
MUA. This dissociation suggests that active inhibition
may be present during this period, especially to coherent stimuli.
CSD/MUA responses integrated across cortical
layers and time windows
To compare the current results with lower-resolution noninvasive hemodynamic measures, the CSD
and MUA measures were combined across the recording sites. The CSD measure was rectified before summing in order to arrive at an estimate of total transsynaptic current flow (cf. the “AVREC” measure of
Schroeder et al. [1995]). The resulting waveforms are
shown in Figure 5A. Examination of these waveforms
reveals the overall greater response of this area to
moving vs. stationary stimuli and to coherent vs. incoherent motion. However, clear distinctions also are
apparent. For example, the CSD response to stationary
stimuli appears to be relatively large (gray arrow, Fig.
5-1A), as does the sustained MUA response to incoherently moving stimuli (black arrow, Fig. 5-2A).
These points were examined quantitatively by integrating the responses across time windows corresponding to the sustained (Fig. 5B) and phasic (Fig.
5C) neuronal responses, followed by statistical analysis between conditions across individual trials. Both
phasic and sustained CSD and MUA responses, to
both stationary and moving stimuli, were significantly
different from baseline (t ⬎ 12.45, P ⬍ 10-5; all tests are

two-tailed with ⬇240 – 480 degrees of freedom), except
for the sustained MUA responses, which were not
significantly different (t ⫽ .26). All responses (phasic
and sustained, MUA and CSD) were greater to moving than to stationary stimuli (t ⬎ 9.22, P ⬍ 10-5),
except for the sustained CSD responses, which were
not significantly different (t ⫽ .91). Finally, all responses (phasic and sustained, MUA and CSD) were
greater to coherently than to incoherently moving
stimuli (t ⬎ 4.79, P ⬍ 10-5), except for the sustained
MUA responses, which were actually greater to the
incoherently moving dots (t ⫽ 21.15, P ⬍ 10-5).
DISCUSSION
Local cortical circuits processing
coherent visual motion
The laminar distribution of population synaptic currents and action potentials were recorded from putative MT⫹ while the subject viewed the onset of a
pattern of stationary dots that after a delay began to
move, either coherently or incoherently. Moving stimuli were more effective in activating this structure
than stationary stimuli, and coherent motion was
more effective than incoherent.
An early sink in layer IV was evoked by coherently
as well as incoherently moving visual stimuli. This is
consistent with primate neuroanatomical studies
showing that the projection from V1 and V2 to MT
terminates predominately in layer IV [Rockland, 1989,
1995], as is typical for feedforward connections in the
neocortex [Rockland and Pandya, 1979; Felleman and

Figure 4.
Responses in putative MT⫹ to coherent vs. incoherent motion. time base, a phasic deep MUA and CSD response to stationary
Current sources and sinks (indicated with black and white contour stimulus onset is seen. Stationary stimulus onset also evokes
lines, respectively) are superimposed over a background showing sustained rhythmic CSD in superficial layers without accompanythe level of MUA activity (decreases indicated as blue and purple, ing MUA. This superficial activity continues through the onset of
increases as orange and red). MUA increases and decreases are motion (coherent and incoherent are combined in this panel). This
also indicated by red and green contour lines, respectively. Cor- overlap prevents clear interpretation of superficial CSD activity to
tical location (with more superficial lamina at top) is plotted visual motion. In deeper cortical laminae, the onset and offset of
against time relative to when the stationary dots (presented during motion both evoke much larger CSD and MUA responses than
the black bars) begin to move (red bars). A. The onset of coherent does the onset of the stationary pattern. Responses are shown
motion produces a large MUA increase as well as source-sink relative to a baseline prior to stimulus motion onset (⫺67 to 33
configurations in both deep (layers V–VI), as well as superficial ms) in panels A–C or to a baseline prior to stationary stimulus
(layer III) pyramidal laminae. The earliest response is in the deep- onset (⫺1033 to ⫺633 ms) in panel D, with zero time indicating
est pyramidal layers, with MUA onset at about 90 ms and peak motion onset. In all panels, for both CSD and MUA, thick contour
CSD response of about 1 ␮A/mm3 at 150 ms. B. Incoherent lines indicate P ⬍ 10-5 (uncorrected for multiple comparisons).
motion onset produces a similar but much weaker and slightly For panels A, B and D, the first thin contour line indicates P ⬍ 10-3
delayed CSD and MUA response. C. Subtraction of the response with additional thin contour lines at t-score increments of 1 (for
to incoherent from that to coherent motion reveals the initially MUA) or 2 (for CSD). In panel C, thin contour lines are at P ⬍
greater CSD and MUA in both superficial and deep pyramidal 0.002, 0.001, 0.0001, and 0.000002.
layers, followed by a greater decrease. D. Using a compressed

䉬

233 䉬

䉬 Ulbert et al. 䉬

Figure 5.
Total estimated transcortical synaptic currents and action potentials. The left panels (1A,B,C) display the total CSD, as rectified
and summed across the different cortical layers. The right panels
(2A,B,C) display the total MUA, again summed across cortical
layers. The top panels (1A,2A) show the waveforms calculated
over the entire trial, including the prestimulus baseline, stationary
visual stimuli, moving stimuli, and again stationary stimuli. The
middle panels (1B,2B) show the total CSD and MUA integrated
over long time windows chosen to encompass the entire neuronal
responses, to stationary stimuli (90 –583 ms after stationary stimulus onset) and coherently or incoherently moving stimuli (90 –
423 ms after motion onset) in comparison to the baseline (500 –
100 ms prior to stationary stimulus onset). The responses are
divided by the duration of the integration window, and normalized
to the phasic response to coherently moving stimuli. The bottom
panels (1C,2C) show the total CSD and MUA integrated over time
windows chosen to reflect the phasic neuronal responses to
stationary and moving stimuli (90 –180 or 90 –160 ms after the
respective stimulus onsets) in comparison to the same prestimulus
baseline. The phasic reponses display the general effects predicted
by the BOLD response in previous studies: a progressive increase
from baseline, to stationary, to incoherently moving, to coherently
moving stimuli. The sustained responses deviate from these predictions, as discussed in the text.

VanEssen, 1991]. An early deep pyramidal layer
source was also noted, as has previously been found in
laminar CSD/MUA recordings from macaque visual
association cortex [Mehta et al., 2000]. This may correspond to the observation that the V1 projection to

䉬

MT is not only to layer IV but also includes layer VI
[Rockland, 1989].
The presence of a differentiated response in the
deepest pyramids is of particular significance because
the neocortex is generally organized such that the
deep pyramids are the usual origin of long-range subcortical and cortico-cortical projections [Rockland and
Pandya, 1979; Felleman and VanEssen, 1991]. From
MT⫹, these fibers project to the superior colliculus,
basal ganglia, thalamus, brainstem, and parieto-frontal occulomotor and visual attention areas, where they
are thought to play a role in integrating optic flow
with perceptual stabilization and oculomotor control
[Ungerleider et al., 1984; Ungerleider and Desimone,
1986; Boussaoud et al., 1990]. Thus, the current results
show that by ⬇110 ms after motion onset, there is a
strong output from MT⫹ indicating to distant target
structures whether that motion is coherent.
The CSD/MUA response in the deep pyramidal
layer is followed after ⬇20 ms with a CSD/MUA
response in the superficial pyramidal layer. These responses appeared to be linked, in that they covaried
across conditions in latency and strength. This delayed coupling between deep and superficial pyramidal activity is consistent with numerous studies in
animals that have identified a powerful excitatory local projection from deep to superficial pyramidal cells
(see for example the laminar studies of rat sensory
cortex by Sukov and Barth [1998]).
Following these phasic responses, moving stimuli
evoke a sustained deep pyramidal CSD response that
is larger to coherent motion, and a MUA response that
is larger to incoherent. This dissociation could be explained by increased recurrent inhibition following
the increased initial excitation evoked by coherent
stimuli. A prominent feature of local intracortical circuitry is that the recurrent collaterals of pyramidal
cells strongly activate inhibitory neurons, which then
recurrently inhibit the pyramids [Berman et al., 1992].
The relatively short latencies recorded here correspond to previous observations of responses in the
dorsal visual stream. Anatomical [Snowden et al.,
1992; Anderson et al., 1998] and physiological [Maunsell et al., 1990; Sclar et al., 1990; Gegenfurtner et al.,
1994] evidence in macaques suggests that motion-sensitive activation is projected from the magnocellular
division of the lateral geniculate body to cells in layer
IVB of area V1, which project directly to MT⫹ via
rapid highly myelinated fibers [Rockland, 1989]. The
lower layers of human MT⫹ are also myelinated
[Clarke and Miklossy, 1990], and MT⫹ hemodynamic
activation has high contrast–sensitivity and low chromatic-sensitivity, suggesting that the human MT⫹ re-

234 䉬

䉬 Human Synaptic/Unit Responses to Visual Flow 䉬

ceives motion-selective activation from V1 [Tootell et
al., 1995]. Although MST follows MT in the dorsal
visual processing stream [Ungerleider and Desimone,
1986; Boussaoud et al., 1990], the response latencies of
MST neurons are about the same as those of MT
neurons in macaques [Schmolesky et al., 1998]. In
general, these latencies are quite short in comparison
with the ventral stream pathway. In fact, laminar recordings in the macaque found that average response
latencies from the region of MT⫹ were shorter than
those from V1 [Schroeder et al., 1998]! In human extracranial EEG and MEG recordings, the earliest responses in V1 occur at a latency to peak of about 70 ms
[Regan, 1989], and motion-sensitive components over
MT⫹ occur with a peak latency of about 150 ms [Kuba
and Kubova, 1992; Dale et al., 1996; Ahlfors et al., 1999;
Nakamura and Ohtsuka, 1999], which correspond
well with those observed in the current study.
It is noteworthy that the differentiation in the MUA
responses to coherent vs. incoherent motion is visible
at the onset of the response, and is significant shortly
thereafter. This argues strongly against models
wherein the selectivity for coherent motion arises from
network interactions within MT⫹. In any case, these
data show clearly that not only was the initial MUA
increase stronger to coherent stimuli, the subsequent
decrease (peaking about 70 ms after the excitatory
peak) was also much stronger to coherent stimuli. This
finding is in direct contradiction to the opponent inhibition feedback model, which predicts that the inhibition following the initial excitation should be larger
to the incoherent stimuli.
Overall, the patterns of synaptic and unit activation
observed here correspond to that expected from previous studies in animals, and thus confirms that these
model studies are applicable to humans performing
behavioral tasks. The current results are clearly limited because they are multiunit rather than based on
an examination of individual neurons with previously
identified receptive fields. Nonetheless, the current
results describe with an unprecedented level of detail
the synaptic and unit responses of MT⫹ neurons in
the awake human to coherent and incoherent stimuli,
and thus should provide data relevant to future theorizing.
Neural correlates of the BOLD response
These data also have implications for the relation
between neuronal activity and the BOLD (blood oxygenation level dependent) response measured with
fMRI. This is a crucial issue because although the
BOLD response has been used extensively to nonin-

䉬

vasively localize brain activation during behavioral
tasks in humans, BOLD is a hemodynamic measure
that has a delayed and indirect relation to the synaptic
and neural activity that actually embodies ongoing
information processing [Buxton et al., 1998]. In particular, there is no well-accepted quantitative model that
calculates the BOLD response from detailed electrophysiological measurements and no empirical studies
of that relationship for high-level cognitive tasks in
people.
Because much of the metabolic effort in the brain
goes toward restoring ionic balances disturbed by
trans-synaptic current flows, it is possible that the
hemodynamic response closely mirrors those flows
[Frostig et al., 1990]. This hypothesis would predict a
close correspondence between the behavioral response profiles of CSD (as reported here) with those
found in previous fMRI/PET studies. Other studies
have suggested that a linear relationship exists between unit activity recorded in macaques and the
BOLD response measured in humans in the same or a
similar task [e.g., Heeger et al., 1999]. Because unit
activity arises from synaptic activity, these measures
are usually coupled. However, high levels of synaptic
activity without cell-firing may occur if there is simultaneous excitatory currents at distal dendrites and
inhibition at proximal locations [e.g., Larkum et al.,
1999].
Indeed, in the current study, both total synaptic
activity (as estimated by CSD) and total neuronal firing (as estimated by MUA), strongly increased over
baseline in response to moving visual stimuli and (at
least acutely) in response to coherent vs. incoherent
motion (Fig. 5), both cardinal properties of the BOLD
response in MT⫹ [Tootell et al., 1995]. Furthermore,
both MUA and CSD responded transiently to stationary stimulus onset and to the termination of visual
motion, as has been reported for the BOLD response
[Tootell et al., 1995].
However, important dissociations can also be found
between the CSD or MUA responses and the BOLD
responses reported in the literature to similar stimuli.
For example, the total CSD was clearly greater to
coherent than incoherent motion both acutely and
over the entire stimulation period. In contrast, the
initially greater increase of MUA to coherent stimuli
was followed by a greater decrease beginning at ⬇133
ms, possibly because of recurrent inhibition. This resulted in the total MUA actually being greater to incoherent stimuli when measured over the entire stimulus period (Fig. 5). This conclusion runs counter to
that of Rees et al. [2000] who concluded that increasing
the coherence of visual motion increased macaque MT

235 䉬

䉬 Ulbert et al. 䉬

unit activity as well as the human MT⫹ BOLD response. Rees et al. [2000] used the macaque data of
Britten et al. [1993] who quantified the responses of
MT⫹ neurons that had been preselected to be responsive to random dot patterns in a directionally sensitive
manner. They used a sharpened high-impedance microelectrode that was moved until a single cell was
isolated. In contrast, the current study used a relatively large recording surface and sampled MUA at
regular fixed intervals of 175␮ across the cortical
depth. In addition, we sampled only one cortical traverse, whereas Britten et al. (1993) sampled many. The
sampling bias induced by these multiple factors is
difficult to evaluate.
Although in this comparison of the sustained response to coherent vs. incoherent motion, CSD seems
to track the BOLD response better than does MUA; the
opposite pattern was observed for the overall sustained responses to stationary vs. moving stimuli,
which MUA distinguished much more strongly than
did CSD. Sustained MUA levels were over tenfold
higher for moving than for stationary stimuli or baseline, which were about equal. In contrast, sustained
CSD levels did not differ significantly between stationary and moving stimuli, although both were greater
than baseline. It should be noted that in their acute
responses, MUA and CSD behaved similarly, being
greater to stationary stimuli than baseline, and greater
to moving than stationary.
Thus, in both cases where the neuronal responses to
stimulation parameters differed from what BOLD
would predict, the divergence was observed only for
the sustained measures, not for the acute response.
This is unexpected because the BOLD response is
thought to act like a low-pass filter on the metabolic
demand [Buxton et al., 1998], and the BOLD responses
to successive neuronal demands are thought to add
linearly [Dale and Buckner, 1997]. The net effect of
these two properties should be that the BOLD response would track the integrated neuronal activity
level over the entire stimulation period rather than the
initial phasic response.
In the two critical cases where the tracking between
neuronal and BOLD responses are reversed from that
expected there are particular explanations that may
not generalize to other situations. Specifically, stationary stimuli evoked large sustained CSD oscillations in
the superficial cortical layers without inducing significant MUA and possibly without entailing increased
metabolic demand. Similarly, the net decrease in MUA
to coherent vs. incoherent motion at longer latencies
appeared to be caused by recurrent inhibition, i.e.,
sustained excitatory drive accompanied by inhibitory

䉬

currents at the soma that effectively decrease cell firing. An apparently similar phenomenon was reported
by Ackermann et al. [1984] who found that metabolism (measured with 2DG) increases in the deafferented hippocampus when the fornix is stimulated,
producing brief excitation followed by a long and
profound active inhibition.
Thus, in this study, special factors seem to render
the later CSD and MUA levels less reliable than the
initial for predicting the BOLD response. It is unclear
from this initial study whether these factors will also
be operative in other cortical areas or behavioral tasks.
Theoretical physiological analysis suggests that many
dissociations are possible between population synaptic activity, unit activity, and the BOLD response. If
the findings of this study generalize, then it may be
found that the initial neuronal response is more representative of the metabolic demand than is the sustained response. Experimental studies in the primary
sensory cortices of animals have emphasized the relationship of the initial layer IV neuronal response to the
overall hemodynamic response, and thus are consistent with this interpretation [e.g., Woolsey et al., 1996].
CONCLUSION
Population synaptic and unit activity in putative
human MT⫹ respond strongly and differentially to
moving vs. stationary visual stimuli, and especially to
coherent motion. The differential response is especially early and strong in the layers that project to
other visual motion and occulomotor areas. The differential response to coherent vs. incoherent motion is
present from the onset of the response, and thus does
not depend on secondary processing by interneurons.
However, the MUA/CSD patterns suggest that recurrent inhibition may play an important role in mediating tonic differences between such stimuli. In addition, the current results emphasize the complexity of
the relationship between neural activity and the hemodynamic response and the importance of further
detailed studies with simultaneous metabolic and
electrophysiological measures. The potential variability in this relationship requires that conclusions from
model studies be validated in humans performing
cognitive tasks.
ACKNOWLEDGMENTS
We thank L. Shuer and M. Reisinger for clinical
collaboration, A. Dale and J. Mendola for scientific
collaboration, and B. Harris for histology.

236 䉬

䉬 Human Synaptic/Unit Responses to Visual Flow 䉬

REFERENCES
Ackermann RF, Finch DM, Babb TL, Engel J Jr. (1984): Increased
glucose metabolism during long-duration recurrent inhibition of
hippocampal pyramidal cells. J Neurosci 4:251–264.
Ahlfors SP, Simpson GV, Dale AM, Belliveau JW, Liu AK, Korvenoja A, Virtanen J, Huotilainen M, Tootell RB, Aronen HJ,
Ilmoniemi RJ (1999): Spatiotemporal activity of a cortical network for processing visual motion revealed by MEG and fMRI.
J Neurophysiol 82:2545–2555.
Andersen RA, Shenoy KV, Crowell JA, Bradley DC (2000) Neural
mechanisms for self-motion perception in area MST. Int Rev
Neurobiol 44:219 –233.
Anderson JC, Binzegger T, Martin KA, Rockland KS (1998): The
connection from cortical area V1 to V5: a light and electron
microscopic study. J Neurosci 18:10525–10540.
Berman NJ, Douglas RJ, Martin KA (1992): GABA-mediated inhibition in the neural networks of visual cortex. Prog Brain Res
90:443– 476.
Boussaoud DB, Ungerleider LG, Desimone R (1990): Pathways for
motion analysis: cortical connections of the medial superior temporal and fundus of the superior temporal visual areas in the
macaque. J Comp Neurol 296:462– 495.
Bradley DC, Andersen RA (1998): Center-surround antagonism
based on disparity in primate area MT. J Neurosci 18:7552–7565.
Britten KH, Shadlen MN, Newsome WT, Movshon JA (1993): Responses of neurons in macaque MT to stochastic motion signals.
Vis Neurosci 10:1157–1169.
Buxton RB, Wong EC, Frank LR (1998): Dynamics of blood flow and
oxygenation changes during brain activation: the balloon model.
Magn Reson Med 39:855– 864.
Clarke S, Miklossy J (1990): Occipital cortex in man. Organisation of
callosal connections, related myelo- and cytoarchitecture, and
putative boundaries of functional visual areas. J Comp Neurol
298:188 –214.
Dale AM, Ahlfors SP, Aronen HJ, Belliveau JW, Huotilainen M,
Ilmoniemi RJ, Kennedy WA, Korvenoja A, Liu AK, Reppas JB,
Rosen BR, Sereno MI, Simpson GV, Standertskjold-Nordenstam
C-G, Tootell RBH, Virtanen J (1996): Spatiotemporal imaging of
motion processing in human visual cortex. Neuroimage 3:S359.
Dale AM, Buckner RL (1997): Selective averaging of rapidly presented individual trials using fMRI. Hum Brain Mapp 5:329 –340.
Duffy CJ, Wurtz RH (1991): Sensitivity of MST neurons to optic flow
stimuli. I. A continuum of response selectivity to large-field
stimuli. J Neurophysiol 65:1329 –1345.
Felleman DJ, VanEssen DC (1991): Distributed hierarchical processing in the primate cerebral cortex. Cereb Cortex 1:1– 47.
Frostig RD, Lieke EE, Ts’o DY, Grinvald A (1990): Cortical functional architecture and local coupling between neuronal activity
and the microcirculation revealed by in vivo high-resolution
optical imaging of intrinsic signals. Proc Natl Acad Sci U S A
87:6082– 6086.
Geesaman BJ, Born RT, Andersen RA, Tootell RB (1997): Maps of
complex motion selectivity in the superior temporal cortex of the
alert macaque monkey: a double-label 2-deoxyglucose study.
Cereb Cortex 7:749 –757.
Gegenfurtner KR, Kiper DC, Beusmans JM, Carandini M, Zaidi Q,
Movshon JA (1994): Chromatic properties of neurons in macaque MT. Vis Neurosci 11:455– 466.
Graziano MS, Andersen RA, Snowden RJ (1994): Tuning of MST
neurons to spiral motions. J Neurosci 14:54 – 67.

䉬

Grover FS, Buchwald JS. 1970. Correlation of cell size with amplitude of background fast activity in specific brain nuclei. J Neurophysiol 33:160 –171.
Halgren E, Babb TL, Crandall PH (1977): Responses of human
limbic neurons to induced changes in blood gases. Brain Res
132:43– 63.
Heeger DJ, Boynton GM, Demb JB, Seidemann E, Newsome WT
(1999): Motion opponency in visual cortex. J Neurosci 19:7162–
7174.
Kuba M, Kubova Z (1992): Visual evoked potentials specific for
motion onset. Doc Ophthalmol 80:83– 89.
Lagae L, Maes H, Raiguel S, Xiao DK, Orban GA (1994): Responses
of macaque STS neurons to optic flow components: a comparison of areas MT and MST. J Neurophysiol 71:1597–1626.
Lappe M, Bremmer F, Pekel M, Thiele A, Hoffmann KP (1996): Optic
flow processing in monkey STS: a theoretical and experimental
approach. J Neurosci 16:6265– 6285.
Lappe M, Duffy CJ (1999): Optic flow illusion and single neuron
behaviour reconciled by a population model. Eur J Neurosci
11:2323–2331.
Larkum ME, Zhu JJ, Sakmann B (1999): A new cellular mechanism
for coupling inputs arriving at different cortical layers. Nature
398:338 –341.
Legatt AD, Arezzo JC, Vaughan HGJ (1980): Averaged multiple unit
activity as an estimate of phasic changes in local neuronal activity: effects of volume-conducted potentials. J Neurosci Meth
2:203–217.
Maunsell JH, Nealey TA, DePriest DD (1990): Magnocellular and
parvocellular contributions to responses in the middle temporal
visual area (MT) of the macaque monkey. J Neurosci 10:3323–
3334.
Maunsell JH, VanEssen DC (1983): Functional properties of neurons
in middle temporal visual area of the macaque monkey. I. Selectivity for stimulus direction, speed, and orientation. J Neurophysiol 49:1127–1147.
Mehta AD, Ulbert I, Schroeder CE (2000): Intermodal selective attention in monkeys. II. Physiological mechanisms of modulation.
Cereb Cortex 10:359 –370.
Mitzdorf U (1985): Current source-density method and application
in cat cerebral cortex: investigation of evoked potentials and
EEG phenomena. Physiol Rev 65:37–100.
Nakamura Y, Ohtsuka K (1999): Topographical analysis of motiontriggered visual-evoked potentials in man. Jpn J Ophthalmol
43:36 – 43.
Nicholson C, Freeman JA. 1975. Theory of current source density
analysis and determination of the conductivity tensor for anuran
cerebellum. J Neurophysiol 38:356 –368.
Perrone JA, Stone LS (1998): Emulating the visual receptive-field
properties of MST neurons with a template model of heading
estimation. J Neurosci 18:5958 –5975.
Rappelsberger P, Pockberger H, Petsche H (1981): Current source
density analysis: methods and application to simultaneously
recorded field potentials of the rabbit’s visual cortex. Pflugers
Arch 389:159 –170.
Regan D (1989): Human brain electrophysiology. New York:
Elsevier.
Rockland KS (1989): Bistratified distribution of terminal arbors of
individual axons projecting from area V1 to middle temporal
area (MT) in the macaque monkey. Vis Neurosci 3:155–170.
Rockland KS (1995): Morphology of individual axons projecting
from area V2 to MT in the macaque. J Comp Neurol 355:15–26.

237 䉬

䉬 Ulbert et al. 䉬

Rockland KS, Pandya DN (1979): Laminar origins and terminations
of cortical connections of the occipital lobe in the rhesus monkey.
Brain Res 179:3–20.
Schmolesky MT, Wang Y, Hanes DP, Thompson KG, Leutgeb S,
Schall JD, Leventhal AG (1998): Signal timing across the macaque visual system. J Neurophysiol 79:3272–3278.
Schroeder CE, Mehta AD, Givre SJ (1998): A spatiotemporal profile
of visual system activation revealed by current source density
analysis in the awake macaque. Cereb Cortex 8:575–592.
Schroeder CE, Steinschneider M, Javitt DC, Tenke CE, Givre SJ,
Mehta AD, Simpson GV, Arezzo JC, Vaughan HG Jr. (1995):
Localization of ERP generators and identification of underlying
neural processes. Electroencephalogr Clin Neurophysiol Suppl
44:55–75.
Sclar G, Maunsell JHR, Lennie P (1990): Coding of image contrast in
central visual pathways of the macaque monkey. Vis Res 30:1–
10.
Snowden RJ, Treue S, Andersen RA (1992): The response of neurons
in areas V1 and MT of the alert rhesus monkey to moving
random dot patterns. Exp Brain Res 88:389 – 400.
Sukov W, Barth DS (1998): Three-dimensional analysis of spontaneous and thalamically evoked gamma oscillations in auditory
cortex. J Neurophysiol 79:2875–2884.
Talairach J, Tournoux P (1988): Co-planar stereotaxic atlas of the
human brain. New York: Thieme.
Tanaka K, Fukada Y, Saito HA (1989): Underlying mechanisms of
the response specificity of expansion/contraction and rotation
cells in the dorsal part of the medial superior temporal area of
the macaque monkey. J Neurophysiol 62:642– 656.

䉬

Tanaka K, Hikosaka K, Saito H, Yukie M, Fukada Y, Iwai E (1986):
Analysis of local and wide-field movements in the superior
temporal visual areas of the macaque monkey. J Neurosci 6:134 –
144.
Tootell RBH, Reppas JB, Kwong KK, Malach R, Born RT, Brady TJ,
Rosen BR, Belliveau JW (1995): Functional analysis of human MT
and related visual areas using magnetic resonance imaging.
J Neurosci 15:3215–3230.
Ulbert I, Halgren E, Heit G, Karmos G (2001): Multiple microelectrode-recording system for human intracortical applications.
J Neurosci Meth 106:69 –79.
Ungerleider L, Desimone R, Galkin TW, Mishkin M (1984): Subcortical projections of area MT in the macaque. J Comp Neurol
223:368 –386.
Ungerleider LG, Desimone R (1986): Cortical connections of visual
area MT in the macaque. J Comp Neurol 248:190 –222.
Vaknin G, DiScenna PG, Teyler TJ (1988): A method for calculating
current source density (CSD) analysis without restoring sites
outside the sampling volume. J Neurosci Meth 24:131–135.
Watson JD, Myers R, Frackowiak RSJ, Hajnal JV, Woods RP, Mazziotta JC, Shipp S, Zeki S (1993): Area V5 of the human brain:
evidence from a combined study using positron emission tomography and magnetic resonance imaging. Cereb Cortex 3:79 –
94.
Woolsey TA, Rovainen CM, Cox SB, Henegar MH, Liang GE, Liu D,
Moskalenko Y, Sui J, Wei L (1996): Neuronal units linked to
microvascular modules in cerebral cortex: response elements for
imaging the brain. Cereb Cortex 6:647– 660.

238 䉬