Download The response of cat visual cortex to flicker stimuli of variable frequency

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Emotional lateralization wikipedia , lookup

Optogenetics wikipedia , lookup

Allochiria wikipedia , lookup

Eyeblink conditioning wikipedia , lookup

Neuroesthetics wikipedia , lookup

Surface wave detection by animals wikipedia , lookup

Animal echolocation wikipedia , lookup

Neuroethology wikipedia , lookup

Neural coding wikipedia , lookup

Time perception wikipedia , lookup

Neural correlates of consciousness wikipedia , lookup

Neural oscillation wikipedia , lookup

Visual N1 wikipedia , lookup

Response priming wikipedia , lookup

Metastability in the brain wikipedia , lookup

Psychophysics wikipedia , lookup

P200 wikipedia , lookup

Stimulus (physiology) wikipedia , lookup

C1 and P1 (neuroscience) wikipedia , lookup

Perception of infrasound wikipedia , lookup

Evoked potential wikipedia , lookup

Feature detection (nervous system) wikipedia , lookup

Transcript
European Journal of Neuroscience, Vol. 10, pp. 1856–1877, 1998
© European Neuroscience Association
The response of cat visual cortex to flicker stimuli of
variable frequency
Günter Rager and Wolf Singer1
University Fribourg, Institute of Anatomy, 1, rue A. Gockel, CH-1700 Fribourg, Switzerland
1Max Planck Institute for Brain Research, Deutschordenstr. 46, D-60528 Frankfurt/Main, Germany
Keywords: flicker, frequency response, oscillations, visual cortex
Abstract
We examined the possibility that neurons or groups of neurons along the retino-cortical transmission chain have
properties of tuned oscillators. To this end, we studied the resonance properties of the retino-thalamo-cortical
system of anaesthetized cats by entraining responses with flicker stimuli of variable frequency (2–50 Hz).
Responses were assessed from multi-unit activity (MUA) and local field potentials (LFPs) with up to four spatially
segregated electrodes placed in areas 17 and 18. MUA and LFP responses were closely related, units
discharging with high preference during LFP negativity. About 300 ms after flicker onset, responses stabilized
and exhibited a highly regular oscillatory patterning that was surprisingly similar at different recording sites due
to precise stimulus locking. Fourier transforms of these steady state oscillations showed maximal power at the
inducing frequency and consistently revealed additional peaks at harmonic frequencies. The frequencydependent amplitude changes of the fundamental and harmonic response components suggest that the retinocortical system is entrainable into steady state oscillations over a broad frequency range and exhibits
preferences for distinct frequencies in the θ- or slow α-range, and in the β- and γ-band. Concomitant activation
of the mesencephalic reticular formation increased the ability of cortical cells to follow high frequency stimulation,
and enhanced dramatically the amplitude of first- and second-order harmonics in the γ-frequency range between
30 and 50 Hz. Cross-correlations computed between responses recorded simultaneously from different sites
revealed pronounced synchronicity due to precise stimulus locking. These results suggest that the retino-cortical
system contains broadly tuned, strongly damped oscillators which altogether exhibit at least three ranges of
preferred frequencies, the relative expression of the preferences depending on the central state. These
properties agree with the characteristics of oscillatory responses evoked by non-temporally modulated stimuli,
and they indicate that neuronal responses along the retino-cortical transmission chain can become synchronized
with precision in the millisecond range not only by intrinsic interactions, but also by temporally structured stimuli.
Introduction
Neuronal networks have a tendency to engage in synchronous,
oscillatory activity. The extent to which neurons synchronize dominant
oscillation frequencies differs for different structures of the brain and
exhibits a marked state dependency (for review see Steriade et al.,
1990; Basar & Bullock, 1992; Singer, 1993; McCormick & Bal,
1994; Contreras & Steriade, 1997). Frequencies below 5 Hz (delta)
are observed during pathological states, e.g. coma, but also during
anaesthesia and slow wave sleep. Activity in the theta-band prevails
in limbic structures during exploratory behaviour. Oscillatory activity
in the 10 Hz range, also known as alpha activity, occurs during
drowsiness or states of relaxation, and is particularly pronounced in
thalamic nuclei and cortical areas involved in the processing of
sensory information. Up to this frequency range, oscillatory activity
can readily be recorded with macro-electrodes, which indicates that
a large number of neurons must have synchronized their activities at
the respective oscillation frequencies. During states characterized by
high levels of arousal and attention, conventional electroencephalo-
Correspondence: W. Singer, as above.
Received 29 July 1997, revised 14 January 1998, accepted 16 January 1998
gram (EEG) recordings exhibit low amplitude fluctuations whose
Fourier spectrum covers a broad range of frequencies extending
up to 60 Hz and more. This pattern is commonly referred to as
desynchronized EEG, but analysis with refined methods has revealed
that regular oscillatory activity also occurs during these desynchronized states, whereby the dominant oscillatory activity is in the β- and
γ-range (Munk et al., 1996; Steriade et al., 1996; Steriade & Amzica,
1996). Furthermore, there is evidence that under these conditions,
groups of cortical neurons involved in ongoing sensory-motor processing can synchronize their discharges with a precision in the
millisecond range and over large distances (Bressler & Nakamura,
1993; Murthy & Fetz, 1996a,b; Bressler, 1996; Munk et al., 1996;
Roelfsema et al., 1997). These results suggest the action of two,
perhaps related mechanisms: (i) a mechanism that causes an oscillatory
patterning of neuronal responses; and (ii) a process that synchronizes
distributed neuronal discharges.
To further characterize the dynamic properties and state dependency
The response of cat visual cortex to flicker stimuli 1857
of the mechanism responsible for this oscillatory modulation, we
investigated the frequency response of the retino-cortical pathway to
temporally modulated input activity, using the cat visual system as a
model. We were particularly interested to see whether the retinothalamo-cortical system exhibits preferences for certain oscillation
frequencies and whether these preferences depend on the state of
central core modulatory systems. Evidence for resonance in distinct
frequency bands has been obtained previously by studying the steady
state responses to flicker in EEG recordings from human subjects
(Regan & Spekreijse, 1986). In addition, we examined how the
precision with which retinal stimuli synchronize cortical responses
compares to that achieved by internal synchronizing mechanisms.
This comparison is relevant for the following reasons. It had been
proposed that internal synchronization of neuronal discharges may
serve to select distributed neuronal responses and to ‘bind’ them
together for further joint processing (von der Malsburg, 1985; Gray
et al., 1989; for review see Singer & Gray, 1995). This suggests the
possibility that the synchronous responses evoked by simultaneously
appearing stimuli may also be used for binding. Psychophysical
studies actually indicate that visual stimuli get bound perceptually if
they are coincident in time (Ramachandran & Rogers-Ramachandran,
1991; Leonards et al., 1996; Leonards & Singer, 1997; but see Kiper
et al., 1996). The synchronicity achieved by internal synchronization
has a precision in the millisecond range and we wondered whether a
similar degree of precision is attained by stimulus-induced synchronization.
In order to study these questions, we recorded local field potentials
(LFPs) and multi-unit activity (MUA) with multiple electrodes from
areas 17 and 18 of anaesthetized cats while presenting flicker stimuli
at frequencies ranging from 2 to 50 Hz. For the investigation of the
dependence of frequency response on central states, we activated the
mesencephalic reticular formation (MRF) in conjunction with the
light stimuli. Responses obtained from the various electrodes were
then subject to auto- and cross-correlation analysis for the assessment
of oscillatory patterning and synchronization.
Materials and methods
Animal preparation
Data were obtained from four adult (. 1 year) cats. Surgical techniques and recording procedures were similar to those described in
detail previously (Engel et al., 1990). Anaesthesia was induced with
an intramuscular injection of ketamine hydrochloride (Ketanest®,
Parke-Davis, Berlin, Germany; 10 mg/kg) and xylazine hydrochloride
(Rompun®, Bayer, Leverkusen, Germany; 2.5 mg/kg) and continued
after tracheotomy throughout the experiment by ventilation with a
mixture of 70% O2 and 30% N2O, supplemented with 0.2–1%
halothane (Hoechst, Frankfurt/Main, Germany). Trepanations were
made to provide access to areas 17 and 18, and for the placement of
epidural EEG electrodes and MRF stimulating electrodes (MRF;
Horsley Clark coordinates H: 8 mm; A: 2 mm and L: 3 mm). Once
all surgical interventions were terminated, the head was fixed with a
bolt to the stereotaxic apparatus, eye and ear bars were removed,
and muscle relaxation was induced by continuous i.v. infusion
of hexacarbocholinbromide (Imbretil®, Hormon-Chemie, Munich,
Germany; 0.7 mg/kg/h). End-tidal CO2 was kept within the range of
3.5–4.0% and body temperature at 38 6 0.2 °C. The EEG, electrocardiogram (ECG) and pulmonary pressure were continuously monitored.
Recording and stimulation
MUA and LFP responses were recorded using glass-coated Tungsten
electrodes. The location of the recording electrodes was assessed
from electrolytic lesions in Nissl-stained frontal sections of blocks
containing the visual cortex.
The flicker stimuli were generated with a strobe that evenly
illuminated a frosted tangent screen positioned 1.14 m in front of the
cat’s eye plane. We applied either whole-field stimuli or stimuli
confined to the receptive fields (RFs) of the respective recording
sites. In the latter case, the screen was covered with black paper
containing rectangular apertures whose location, size and orientation
matched the RFs of the MUA responses recorded by the different
electrodes. These RFs had previously been plotted with hand-held
stimuli and were all located within 10 ° of the centre of gaze. The
intensity (luminance) and duration of the stimuli were 2.2 mW/
1337 cm2 and 20 µs, respectively. In the first two experiments, the
frequency of the flicker stimuli was adjusted by varying the intervals
between flashes in steps of 2 ms between 60 and 80 ms, and in larger
steps beyond 80 ms. In the last two experiments, flicker frequencies
were increased in steps of 2 Hz from 2, and respectively, 6 to 50 Hz.
In cases where stimuli were presented together with MRF stimuli,
each flash was paired with a single electrical stimulus that was applied
either to the right or left MRF with coaxial electrodes (10 V, 50 µs).
Each series of flicker stimuli lasted between 3 and 4 s, and was
repeated 10 times at intervals of 10 s. Responses were recorded and
digitized over epochs of 5000 ms comprising stimulus-free epochs
before and after the flicker, so that initial entrainment, steady stateand off-responses could be studied.
Data analysis
MUA responses were obtained by amplification and filtering the
electrode signals in the frequency range of 1–3 kHz. This signal was
fed through a Schmitt trigger whose threshold was set to at least
three times the noise level. The trigger pulses were digitized at 1 kHz
and then stored on disk. LFP responses were obtained by filtering
the electrode signal from 1 to 100 Hz. They were also digitized at
1 kHz and stored together with the MUA. For the evaluation of MUA
responses, peri-stimulus time histograms (PSTH) were compiled for
10 successive stimulus presentations with a bin width of 1 ms. Then,
a window was placed over the steady state phase of the response
(from 500 to 3500 ms), and auto-correlation functions (ACF) and
cross-correlation functions (CCF) were computed within this window.
In order to assess the frequency characteristics of the responses,
CCFs were calculated between the neuronal responses and the strobetrigger. Then, a window of 1000 ms was placed over the crosscorrelogram starting at T0 and the Fast Fourier Transform (FFT) was
computed for this window. The obtained amplitude spectra served as
a basis for surface plots in which the frequency spectrum of the
responses (0–100 Hz) is represented on the X-axis, the flicker frequency of the respective stimulus (2–50 Hz) on the Y-axis, and the
amplitude of the spectra on the Z-axis. To allow for a direct comparison
between response components at fundamental and harmonic frequencies, amplitude distributions of the fundamental, and first and second
harmonic were plotted as a function of stimulus frequency. In addition,
the ratios of the first harmonic over the fundamental and the second
harmonic over the fundamental were computed. In order to determine
the temporal precision with which the stimuli synchronized neuronal
responses, we selected 1-s-long intervals from the steady state
responses and computed CCFs between MUA responses recorded
from pairs of electrodes located either within the same area (17 or
18) or in different areas (17 and 18). These CCFs were then
averaged over 10 successive sweeps. In order to assess whether the
synchronization of responses was solely due to stimulus locking or
whether internal synchronizing mechanisms contributed, in addition
we also computed CCFs between two 1-s-long response epochs that
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
1858 G. Rager and W. Singer
were offset by 1 s, thus correlating responses evoked by flashes
belonging to different, temporally shifted stimulus trains (shift predictor, Gerstein & Perkel, 1972). Correlations persisting in these shift
predictors are exclusively due to stimulus locking. Thus, by comparing
correlograms computed from the original data with shift predictor
correlograms, it is possible to assess to which extent correlations are
due to internal synchronizing mechanisms and stimulus locking,
respectively.
For the evaluation of LFP responses, these were first averaged
over 10 successive trials. These averages were computed either over
the full length of the response or for windows placed over the initial
response phase (first 200 ms after onset of the flicker stimulus),
steady state phase, or the phase around stimulus offset (from 100 ms
before to 200 ms after the end of the flicker stimulus). For the steady
state phase we computed, in addition, LFP averages within windows
of 6 200 ms, using either the spike discharges (spike-triggered
averages) or the flashes (stimulus-triggered averages) as trigger.
For the spectral analysis, FFTs were computed from single sweep
LFP responses with a resolution of 1 ms for a frequency band of 0–
150 Hz and for windows from 500 to 1499 ms after stimulus onset.
These FFTs were subsequently averaged over 10 successive presentations of the same stimulus and average amplitude spectra were plotted
in the same manner as was described for MUA analysis above.
Results
MUA responses to whole field flicker
At low flicker frequency (2 Hz), MUA responses to whole field
stimuli resemble responses to single flashes. They are biphasic
(Fig. 1A,B). The first component consists of one or two brief bursts
which peak at µ 10–120 ms post-stimulus and are followed by an
interval of reduced firing. The second component is more sustained,
starts at µ 200 ms, reaches a maximum between 300 and 500 ms,
and then decays slowly over the next 200–300 ms.
No differences were noted between responses recorded from
different electrodes located within the same area, but in responses
from area 18, the second, sustained component was more pronounced
than in responses from area 17. At this low stimulus frequency, there
were no significant further after-discharges at the end of the flicker
sequence (Fig. 1C). As flicker frequency increased, the tonic response
component became truncated more and more by the inhibition
following the phasic response to the respective next stimulus. Moreover, there was an indication for an entrainment at the beginning of
the response, the later bursts being larger in amplitude than the first.
Moreover, a tonic off-response appeared at the end of stimuli above
6 Hz. Examples of responses to 6 Hz flicker are shown in Fig. 1D,F.
As flicker increased above 10 Hz, tonic response components
became more and more suppressed, but the phasic discharge followed
reliably up to 50 Hz, the highest of the tested frequencies. Usually it
took µ 200–300 ms after stimulus onset until the cells engaged in a
regular, stimulus-locked bursting pattern. During the initial phase, the
responses tended to be less well modulated, bursts occurred with
variable amplitudes and were still intermingled with more sustained
response components. This is exemplified in Fig. 1E,G for responses
to 20 Hz flicker.
Auto- and cross-correlation analysis of MUA
As expected from the stimulus-locked discharge patterns, ACFs
computed from the steady state phase of the flicker response exhibited
a periodic modulation, whereby the intervals between the maxima
corresponded to the respective flicker frequency. However, in most
ACFs, especially at flicker frequencies below 40 Hz, there were
additional peaks of smaller amplitude which occurred at regular
intervals between the principal peaks. The amplitude of these additional peaks varied with stimulus frequency and could even differ
between recording sites for a given flicker frequency. Typically, the
number of additional peaks ranged from one to three, indicating that
grouped discharges had not only occurred at intervals corresponding
to the flicker frequency, but also at shorter intervals, corresponding
to the first, second and sometimes even third harmonic of the actual
flicker frequency. At flicker frequencies between 8 and 20 Hz, two
satellite peaks were usually distinguished, while in general only one
peak remained at higher frequencies. No difference was observed
between areas 17 and 18 with respect to the occurrence of these
satellite peaks: examples of ACFs are shown in Fig. 2 for two groups
of simultaneously recorded neurons, both from area 17, for flicker
frequencies of 10, 20, 30 and 40 Hz. Some ACFs (see, e.g. Fig. 2A)
show, in addition to the phasic modulation, a single hump which is
centred around zero and decays slowly over shift intervals of 200–
300 ms. This reflects periodic fluctuations of response amplitudes
which were also apparent in the corresponding PSTHs. These fluctuations occurred in a frequency range of µ 2 Hz and appeared unrelated
to the frequency of the flicker.
The CCFs computed between MUA responses and the stimulus
triggers revealed that the responses were precisely locked to the
stimuli: all CCFs had a prominent peak centred around zero and were
periodically modulated at intervals corresponding to the flicker
frequency. They also exhibited additional, regularly spaced satellite
peaks whose number and position corresponded to the occurrence of
the various harmonic response components that were present in the
ACFs of the respective responses.
As expected from the precise stimulus locking of the responses,
CCFs computed between MUA recorded from different electrodes
again showed a prominent oscillatory modulation with a large centre
peak and side peaks whose amplitude was related to the stimulation
frequency and its harmonics in the same way as in the ACFs and
CCFs computed between MUA and trigger pulses. In the CCFs
between MUA responses from different electrodes, the amplitude and
width of the centre peak provide a measure of the temporal precision
with which the flashes synchronized the responses at the different
recording sites. We have investigated such CCFs for three electrode
pairs located within area 17, for one pair in area 18, and for three
pairs distributed across the two areas. In all cases, responses were
evoked by whole-field stimuli. Apart from the fact that responses
recorded across the two areas were slightly less synchronized (smaller
and broader centre peak) than responses recorded within the same
area, we noted no significant differences among the CCFs computed
between the various electrode pairs. Figure 3 shows two representative
examples, one for an electrode pair located within the same area
(A 17), the other for electrodes located in different areas. These plots
reveal that the amplitude and width of the centre peak and satellite
peak of the CCFs depend strongly on stimulation frequency. At
frequencies below 8 Hz, CCFs exhibit broad smoothly modulated
peaks whose spacing reflects stimulation frequency. Beyond 8 Hz,
these peaks become larger and sharper, and additional peaks appear
whose spacing corresponds to harmonics of the flicker frequency.
However, this increase in the precision of stimulus locking is not
monotonously related to stimulation frequency. Synchronicity is high
in the frequency band between 8 and 10 Hz, decreases for 12–14 Hz,
becomes particularly pronounced in the range from 16 to 28 Hz, and
then decreases again for higher frequencies with occasional recovery
of the intra-areal correlations at frequencies beyond 40 Hz. Comparison with CCFs computed across successive response segments (shift
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
The response of cat visual cortex to flicker stimuli 1859
predictors, see Materials and methods) revealed no difference between
the actual and shifted CCFs, indicating that the synchronization was
exclusively due to stimulus locking and not enhanced further by
internal synchronizing mechanisms.
Spectral analysis of multi-unit activity
In order to assess more quantitatively the frequency response of
cortical cell groups, the CCFs computed between steady state MUA
responses to whole-field stimuli and the flash triggers were subject
to Fourier analysis. A 1 s window was placed over the CCFs and a
FFT of the correlograms was computed. As expected, the Fourier
spectra had a sharp and prominent peak at the frequency of the
stimulus. Further well-delineated peaks were present at multiples of
the stimulus frequency (Figs 4 and 5). The absolute and relative
amplitudes of these peaks varied substantially with stimulus frequency,
but were similar for responses obtained from different recording sites.
To allow for a comprehensive overview of the frequency response of
MUAs at all tested flicker frequencies, the amplitude spectra of
responses from one recording site in area 17 (Fig. 4A,B) and one in
area 18 (Fig. 5A,B) are represented together both as surface and
contour plots. These plots show that responses to all stimulus
frequencies contain, in addition to the fundamental, at least a first
harmonic. At stimulus frequencies below 30 Hz second harmonics
are also clearly distinguishable, and at frequencies below 20 Hz
higher order harmonics can also be identified. There are virtually no
components at response frequencies lower than the stimulus frequency.
The spectral analysis of MUA responses does not reveal any activity
in this zone. Profiles along the fundamental, first and second harmonic
(Figs 4C and 5C) show the variation in amplitude of the spectra.
There is a first peak between 4 and 8 Hz, which is followed by a
sharp decline around 12 Hz. Then a much higher peak is present
between 16 and 30 Hz. From 30 to 50 Hz, the amplitude fluctuates
on a high level. Comparison of the amplitudes of the first and second
harmonic with that of the fundamental shows that harmonics are
particularly pronounced at a stimulus frequency of µ 12 Hz reaching
amplitudes 10 times larger than that of the fundamental. At lower
and higher stimulus frequencies, the harmonics are usually smaller
or of the same size as the fundamental (Figs 4D and 5D).
Field potential responses to whole field flicker
Field potentials recorded intracortically with microelectrodes reflect
with great fidelity the bulk activity of neurons in the vicinity of
the electrode tip (for review see Mitzdorf, 1985). Field negativity
corresponds mainly to ligand and voltage-gated inward currents,
while positivity is primarily due to passive loop closing outward
currents related to remote excitatory events, and to a minor extent,
to IPSP-related outward currents. Thus, there is usually a tight
correlation between LFP negativity and excitatory responses (see,
e.g. Gray & Singer, 1989). This is also true in the present case:
comparison of averaged MUA and LFP responses revealed that
neuronal discharges were closely time-locked to LFP negativities at
all flicker frequencies. Accordingly, spike-triggered averages of LFP
responses show that maximum firing probability coincides with the
steepest slope of the rising phase of the LFP negativity (Fig. 6).
These relations between MUA and LFP responses were virtually the
same at all recording sites, and LFPs recorded simultaneously from
different sites closely resembled each other, even for recording sites
in different areas. Because of this consistency and because of the
continuous nature of LFPs, which makes them more suitable for
spectral analysis than MUA activity, we based the quantification of
frequency characteristics on LFP responses.
At low flicker frequencies (, 4 Hz) the LFP responses closely
resembled the well known flash-evoked potential (Fig. 6). It consisted
of an initial, sharply rising negativity that peaked at µ 130 ms poststimulus and was followed by one or two additional negative
deflections riding on a large, positive potential. This positive potential
peaked at µ 200 ms and was followed by another, slowly rising and
decaying negativity of large amplitude. As shown in Fig. 6, the early
and late negativities correspond to the phasic and tonic MUA
responses, respectively, and the positivity coincides with the period
of reduced firing after the initial phasic discharges. At higher
frequencies (. 4 Hz) the responses to the first and subsequent stimuli
of a series began to differ from one another, and it took µ 200–
300 ms until the steady state responses emerged. Up to frequencies
of 6 Hz (Fig. 6), the positive components of the LFP were still
pronounced, but as frequencies increased these positivities became
more and more suppressed, and at the same time the negative response
components increased in amplitude up to frequencies of µ 30 Hz.
Up to this frequency, each flash gave rise to more than one negativity,
whose number decreased as frequency increased. Beyond 30 Hz, the
responses became more uniform consisting essentially of a single
negative deflection. At frequencies above 10 Hz, an off-response
appeared in addition after the end of the stimulus train, but there was
no evidence that oscillatory activity outlasts the stimulus.
Spectral analysis of field potentials
To quantify the frequency response of cortical cell groups, segments
of LFP responses taken from the steady state phase of the response
were subject to Fourier analysis (see Materials and methods). As in
the MUA data, the Fourier spectra had a sharp and prominent peak
at the frequency of the stimulus (fundamental), and at the first and
second harmonics (Figs 7 and 8). The absolute and relative amplitudes
of these peaks were similar for responses obtained from different
recording sites both in areas 17 and 18. In contrast to the MUA
responses, however, there are also always components at response
frequencies lower than the stimulus frequency. These are concentrated
in a frequency range of 1–5 Hz, but there is no clear evidence that
they might represent subharmonics, as their spectral composition does
not depend in a systematic way on stimulus frequency. Typically, and
this holds for all experiments, the amplitude of all components tended
to decrease with increasing stimulus frequency for stimuli above
10 Hz, but this decay was not monotonous, as can be seen particularly
well in Figs 7C and 8C where the frequency-dependent changes in
amplitude are plotted for the fundamental, and first and second
harmonic. The fundamental component was maximal at µ 6–8 Hz,
then decayed rapidly as frequencies approached 12–14 Hz, was again
enhanced between 16 and 30 Hz, and then fell off slowly towards
50 Hz, the highest frequency tested. The amplitude of the first
harmonic was maximal at stimulus frequencies of µ 2 Hz, dropped
to a minimum at frequencies of µ 4–8 Hz, increased again for
stimulus frequencies of µ 12–30 Hz and then decayed slowly for
higher frequencies. The modal distribution of the power of the second
harmonic was even more complex. Its amplitude was maximal at a
stimulus frequency of 2 Hz, which corresponds to a response frequency
of 6 Hz, decreased for stimulus frequencies of µ 4–6 Hz, rose again
for frequencies of µ 10–16 Hz, corresponding to a response frequency
of 30–48 Hz, and after a further reduction at 14–18 Hz, showed a
third plateau at stimulus frequencies of 22–30 Hz. The ratio of the
first harmonic over the fundamental and second harmonic over the
fundamental was highest at µ 12 Hz, and very low between 6 and
8 Hz (Figs 7D and 8D).
To show the relation between the amplitude of fundamental and
harmonic oscillations on the one hand, and the actual frequency of
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
1860 G. Rager and W. Singer
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
The response of cat visual cortex to flicker stimuli 1861
FIG. 2. Autocorrelograms of steady state MUA responses to stimulus frequencies of 10 Hz (A), 20 Hz (B), 30 Hz (C) and 40 Hz recorded from two different
electrodes in area 17 (position 1, upper boxes; position 3, lower boxes). In each of these examples the autocorrelation is computed for shift-intervals ranging
from –200 to 1 200 ms for a window ranging from 500 to 3500 ms.
FIG. 1. PSTHs of MUA responses (n 5 10) to flicker of 2 Hz (A–C), 6 Hz (D,F) and 20 Hz (E,G) from two recording positions in areas 17 and 18 located
within the representation of the area centralis. (B,C) and (F,G) show response segments at an expanded time scale, taken from the respective response sequences
at times indicated on the abscissa. Squares below the histograms indicate the timing of flashes. The dimensions for the coordinates in (B–G) are the same as in (A).
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
1862 G. Rager and W. Singer
the response components on the other, the amplitude distributions
were replotted as a function of response frequency rather than
stimulus frequency in Figs 7E and 8E. This revealed that all response
components have their maximal power at a response frequency of
µ 6–8 Hz: a further, albeit much smaller, increase in amplitude occurs
for the fundamental at a response frequency of µ 16–30 Hz, for the
first harmonic at µ 25–60 Hz and for the second harmonic at µ 30–
40 Hz. Again, no major differences were noted between the amplitude
distributions in areas 17 and 18. Taken together, these results indicate
that under anaesthesia the visual system exhibits a preference to
engage in oscillatory activity in the frequency ranges of 6–8 Hz and
16–40 Hz.
The effect of reticular stimulation
In order to determine the dependence of the frequency response of
the thalamo-cortical system on global changes of neuronal excitability,
we investigated the effects of MRF stimulation on MUA and LFP
responses to whole-field flicker. As exemplified in Figs 9 and 10,
LFP responses retained their basic characteristics when MRF stimuli
were added, but there were major differences in the frequency
response. At all stimulation frequencies the steady state response
phase was reached faster and the amplitudes of the responses showed
less fluctuations than without MRF stimulation (data not shown). In
the spectra, the trough between the first (6–8 Hz) and second (µ 16 Hz)
maximum in response amplitude was reduced because of a relative
increase in power of responses in the α-range. The ratio of the first
harmonic over the fundamental was no longer maximal at 12 Hz, but
now peaked between 34 and 42 Hz. Thus, the relative amplitudes of the
harmonic response components increased dramatically for response
frequencies in the γ-range, suggesting that the system now engages
preferentially in oscillatory activity in this frequency range. At the
same time, there was a reduction of frequency components other than
the fundamental and various harmonics, in particular of frequencies
below the respective fundamental. These effects were similar in areas
17 and 18.
Comparison between whole-field and RF stimulation
Switching from whole-field to RF stimuli reduced the attenuation of
response amplitudes that occurred with higher stimulation frequencies
(. 10 Hz) [Fig. 11]. This effect was equally pronounced in MUA
and LFP responses recorded from both areas 17 and 18, and occurred
for both the fundamental and harmonic response components. In the
example shown in Fig. 11, which represents MUA responses from A
17, there is even an enhancement of responses with increasing
stimulation frequency. With whole-field stimuli (Fig. 12) the fundamental response component is maximal around a stimulation frequency
of 10 Hz and declines at higher frequencies. With whole-field stimuli
the peaks reflecting the various resonance frequencies decrease with
stimulation frequencies above 15 Hz, while they stay essentially
unattenuated up to stimulation frequencies of 25 Hz with RF stimuli.
Frequency-dependent phase locking
Another variable that we considered as being of interest for the
description of the frequency response of the system is the extent to
which the responses are phase locked to the individual stimuli. The
extent of phase locking is reflected by the latency jitter of responses
to individual flashes. On average, large jitter is expected to lead to a
smearing of the respective response components and a decrease of
the amplitudes when responses are averaged over different trials. In
order to quantify the degree of phase locking we compared the
amplitude spectra of averaged responses with those obtained from
single sweep analysis. The ratios of the former over the latter give
the percentage of phase locking. Phase locking of the fundamental
response component was found to be nearly perfect over the whole
frequency range tested (ù 85% without and ù 95% with MRF
stimulation). For the harmonics, phase locking was less successful in
the range of response frequencies from 12 to 30 Hz (minimum 55%)
and µ 60 Hz (80%), but nearly perfect (µ 90%) for the remaining
frequency range (Fig. 13A). When the MRF was stimulated simultaneously with the global visual stimulus, phase locking was improved
(Fig. 13B). Only in the second harmonic was a trough seen at 20 Hz
(60%). For the remaining frequency range, phase locking was better
than 80%. Phase locking during local stimulation was comparable
and in some cases even better than during global stimulation (data
not shown).
Discussion
We have deliberately based our analysis of the frequency response
of the visual cortex on signals which reflect group activity in order
to make the results comparable with data on internally generated
oscillatory activity. Endogenous oscillatory activity is also a group
phenomenon that emerges from cooperative interactions among reciprocally coupled neurons and is associated with the synchronization
of local clusters of neurons (see Introduction). In this respect, the
oscillatory activity that is generated by intrinsic mechanisms either
spontaneously (Steriade et al., 1996) or in response to non-temporally
structured light stimuli (Gray & Singer, 1987, 1989; Eckhorn et al.,
1988) closely resembles the flicker-induced oscillations. In both cases,
clusters of units discharge synchronously, and these episodes of
synchronous firing appear as bursts in MUA recordings and transient
negativities in the LFP.
For quantification of frequency responses we have relied more on
LFP than MUA responses because continuous signals are more
appropriate for spectral analysis than spike trains. We consider this
legitimate because the two complementary signals gave very similar
results. LFP signals differed from MUA activity only because
they contained more prominent low frequency components. This is
probably due to the fact that positive deflections of the LFP contribute
to the spectra but are not associated with spike discharges. These
positivities most likely reflect inhibitory events and passive loop
closing currents, and are of large amplitudes at stimulus frequencies
up to 10 Hz. We believe that this is the reason why the spectra of
LFP responses had relatively more power at low frequencies than
those of the MUA responses.
Periodic stimulation of the retina resulted in regular oscillatory
responses over the whole range of tested frequencies (from 2 to
50 Hz). After a transitory phase of µ 300 ms, which is approximately
the duration of the response to a single flash, the responses to flicker
stabilized and exhibited a very regular oscillatory pattern that was
maintained throughout the whole stimulation period and ended
abruptly with the last flash.
No marked differences were observed between responses recorded
from different sites within area 17, nor were there any major
differences between the response patterns obtained from areas 17 and
18, except that in the latter the attenuation of responses with increasing
flicker frequency was less pronounced than in the former. This agrees
with the evidence that in cat LGN afferents to area 18 are exclusively
of the transient or y-type, while those to area 17 are mixed, the
sustained or x-type afferents constituting the majority (for review see
Hoffmann & Stone, 1971; Stone & Dreher, 1973; Mitzdorf & Singer,
1978; Sherman & Guillery, 1996). As ganglion cells of the x-type
follow high frequency flicker less readily than y-type cells, the
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
The response of cat visual cortex to flicker stimuli 1863
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
1864 G. Rager and W. Singer
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
The response of cat visual cortex to flicker stimuli 1865
FIG. 4. Distribution of power (z-axis) in the different frequency bands (x-axis) as a function of flicker frequency (y-axis) of MUA responses recorded from area
17. (B) Contour plot of the peaks shown in (A). Peak height is expressed by the diameter of symbols. (C) Distribution of the fundamental, first and second
harmonic as a function of stimulus frequency (abscissa). (D) Ratios of the power of the first harmonic over the fundamental and of the second harmonic over
the fundamental as a function of stimulus frequency. In the surface plot, response frequencies below 2 Hz were cut off to enhance visibility of fundamental and
harmonic response components.
enhanced frequency attenuation of responses in area 17 is readily
explained by the frequency response of the retinal input. It is
surprising, however, that the differences between the two areas were
not more marked. One explanation could be that the LFP recordings,
and presumably also the multi-unit recordings, overemphasized ymediated responses in A17. This possibility is suggested by previous
FIG. 3. Cross-correlations between responses recorded from two different sites, both being located in area 17 (A,B), or one being located in A17 and the other
in A18 (C,D). Correlograms computed for the same time window from 1000 to 2000 ms are shown in (A) and (C), and the corresponding shift predictors,
computed from shifted, non-overlapping windows comprising steady state responses from 1000 to 2000 ms and 2000–3000 ms, respectively, are shown in (B)
and (D). Note the similarity between real time correlograms and shift predictors.
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
1866 G. Rager and W. Singer
FIG. 5. Spectral analysis of MUA activity in area 18 after global stimulation. Same conventions as in Fig. 4.
studies in which the contributions of x-and y-type inputs to responses
in the visual system have been assessed with current source density
analysis of LFPs (Mitzdorf & Singer, 1978; Mitzdorf, 1985). These
studies indicated that the y-system contributes to a disproportionately
large extent to LFPs because homogenous conduction velocities and
low temporal scatter of responses lead to better synchronization of
afferent volleys, and hence to better summation of synaptic currents.
The over-representation of y-mediated responses in single cell recordings probably results from a sampling bias, because in the y-pathway
cells and fibres tend to have larger diameters than in the x-pathway
(Hoffmann et al., 1972; Stone, 1973). Moreover, as our analyses were
based on auto- and cross-correlations, they overemphasize signals
exhibiting a periodic temporal structure and hence responses with
good phase locking to the stimulus. We propose that this has also
contributed to an over-representation of y-mediated responses because
these are expected to follow flicker stimuli with greater precision
than x-mediated responses (Lu et al., 1995).
Comparison of CCFs between MUA responses recorded from
different electrodes revealed that the observed synchronization of
unit discharges was essentially due to stimulus locking. The CCFs
computed for temporally shifted response segments were indistinguishable from unshifted CCFs indicating that internal, neuronal
interactions had not made a measurable contribution to the synchronization of the discharges. At first sight this is surprising, as powerful
synchronizing mechanisms have been described in the retina (Mastronarde, 1989; Neuenschwander & Singer, 1996), the LGN (Steriade
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
The response of cat visual cortex to flicker stimuli 1867
FIG. 6. Comparison of MUA and LFP responses. Activity is recorded from area 17 at 2 Hz (A) and from area 18 at 6 Hz (B). The upper trace shows the PSTH,
and the lower trace the averaged field potential together with the flash trigger pulse. Since the first trigger pulse starts the run, it is not visible on the plot.
et al., 1991; Steriade et al., 1993) and the visual cortex (Eckhorn et al.,
1988; Gray et al., 1989). Evidence indicates that these mechanisms are
based on lateral interactions within the respective structures. In the
mature retina the neuronal substrate for the interaction has not yet
been identified, in the LGN the interactions are mediated mainly by
inhibitory interneurons of the perigeniculate nucleus (Steriade et al.,
1993), and in the visual cortex by tangential intracortical connections
(Löwel & Singer, 1992; König et al., 1993). One possibility is
that these internal synchronizing mechanisms contributed to the
regularization of the steady state responses, thereby generating such
perfect phase locking that original and shifted correlograms became
similar. This interpretation is supported by the finding that it took
several flash cycles before the responses got entrained into a regular
oscillatory pattern.
The frequency response of the retino-thalamo-cortical system
Both MUA and LFP analysis revealed that the steady state responses
to flicker consisted of several components. The largest response
component reflected the frequency of the inducing flicker. The
additional components corresponded to the first, second and sometimes
even third harmonic frequency. The amplitudes of all components,
fundamental and harmonics, changed with stimulation frequency,
but these changes were not monotonous and differed for fundamental
and harmonic response components. Certain frequencies were
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
1868 G. Rager and W. Singer
The response of cat visual cortex to flicker stimuli 1869
distinguished because they elicited particularly large fundamental or
harmonic responses, or produced particularly large ratios between
the amplitudes of harmonic and fundamental response components.
Electrical activation of the MRF had a strong effect on these frequency
preferences in that it markedly enhanced oscillatory response components in the γ-frequency band.
As reviewed in the Introduction, neurons in the visual system have
the tendency to engage in oscillatory activity covering a broad
spectrum of frequencies whereby certain ranges are distinguished.
This tendency of the visual system to engage in oscillatory activity
at different frequencies suggests the possibility that there are neurons,
or networks of neurons, which behave as oscillators that are tuned to
the preferred ranges of frequencies. If so, periodic activation of
the retino-thalamo-cortical system is expected to reveal resonance
phenomena: for stimuli matching the frequency preference of the
tuned oscillators, responses should become particularly large and
their phase locking to the inducing stimuli should become particularly precise.
The present data support this view, but they indicate that the
respective oscillators are broadly tuned and strongly damped. Broad
tuning is suggested by the finding that oscillatory responses were
entrainable over a broad frequency range and that phase locking was
good for nearly all frequencies. Strong damping of the oscillators is
suggested by the fact that the oscillatory responses stopped abruptly
with the last flash of the flicker.
Two possibilities may be considered to account for the broad
frequency tuning of the system and for the preference to engage in
oscillatory responses in several distinct frequency bands. One is that
there are actually strongly damped oscillators along the retino-cortical
path which are tuned to different frequency bands and get recruited
to variable extents at different stimulation frequencies. This conjecture
is supported by evidence on diverse oscillatory mechanisms that are
tuned to different frequencies. Retinal ganglion cells have the tendency
to engage in highly synchronous oscillatory activity in the range of
60–90 Hz when activated by light stimuli (Neuenschwander & Singer,
1996). Multiple oscillatory mechanisms operating at frequencies
ranging from 0.1 to more than 40 Hz have been identified at the level
of the thalamus (Steriade et al., 1991; Amzica et al., 1992; Nunez
et al., 1992; Steriade et al., 1993; Puil et al., 1994). Finally, evidence
increases for the existence of oscillatory mechanisms at the cortical
level which sustain rhythmic activity in the β- and γ-frequency range
(Eckhorn et al., 1988; Gray & Singer, 1989; Steriade et al., 1996).
Cortical cells of both the pyramidal and non-pyramidal type have
been described that possess pacemaker mechanisms which predispose
them to engage in oscillatory activity in the λ- and θ- (Gutfreund
et al., 1995; Hutcheon et al., 1996), and β- and γ-range (Llinas et al.,
1991; Gray & McCormick, 1996). In addition, it has been shown in
simulation studies, that are supported by recordings from cortical
slice preparations, that the network of coupled inhibitory interneurons
can sustain by itself oscillatory activity in the 40 Hz range (Jefferys
et al., 1996; Traub et al., 1996).
Further support for a contribution of these oscillatory mechanisms
to the frequency response of the system comes from the finding that
MRF stimulation leads to a relative increase of responses in the high
frequency range. It is well established that the thalamic oscillators
which support low frequency oscillations (, 10–12 Hz) are gated by
central core projection systems (for a review of early literature see
Singer, 1977; Steriade & McCarley, 1990; McCormick & Bal, 1994).
When activated, these modulatory systems block the pacemakers that
support oscillations in the low frequency range (up to high α). This
gating of thalamic oscillators is likely to be one of the reasons for
the shift towards higher resonance frequencies when MRF was
stimulated. Moreover, there is now evidence that MRF stimulation
favours the occurrence of an oscillatory patterning in the γ-frequency
range of responses to moving light stimuli (Munk et al., 1996). This
indicates that central core activation predisposes cortical circuits to
engage in high frequency oscillations. The enhanced resonance in the
γ-frequency range after MRF stimulation can, thus, be accounted for
by the contribution of oscillatory mechanisms that operate at different
frequencies and are controlled differentially by central core afferents.
The experimentally determined steady state LFP responses were
compared with responses computed by linear superposition of single
flash responses. This revealed that responses to flicker could be
predicted rather well from responses to single flashes, except for the
range of frequencies for which spectral analysis had suggested
enhanced resonance. For the fundamental response component, the
predictability of the frequency response was good (. 80%) except
for a narrow range of frequencies between 9 and 13 Hz, and the
range above 30 Hz where it dropped to 80 and 60%, respectively (K.
Pawelzik, personal communication). In conclusion then, the retinothalamo-cortical system can be entrained to engage in highly regular
oscillatory activity over a broad frequency range, but exhibits nonlinearities in two distinct frequency ranges that roughly coincide with
the α-band, and the high β- and γ-band. We propose as the most
likely reason for these non-linearities that oscillatory mechanisms
tuned to the respective frequencies are driven in resonance.
Global versus local stimuli
The retina, thalamus, and in particular the cortex are characterized
by connections which allow for lateral interactions between neurons
encoding signals from spatially segregated points in the visual field.
To assess the extent to which these lateral connections contribute to
the frequency response of the system, responses to whole field stimuli
were compared with those to stimuli confined to the excitatory
receptive field. The main difference was that responses to the latter
were less attenuated at high frequencies, while responses to the
former were enhanced at frequencies below 8–10 Hz. This suggests
that lateral interactions contribute to the stabilization and synchronization of oscillatory responses at low frequencies, while they reduce
the ability of the system to synchronize at high frequencies. We
propose that the improved resonance at low frequencies is due to the
synchronizing action of lateral inhibitory connections which are
recruited by whole field stimuli both at the thalamic and cortical level.
These inhibitory mechanisms are known to contribute substantially
to the stabilization and global synchronization of low frequency
oscillations (up to 10 or 12 Hz), e.g. those which occur during slow-
FIG. 7. Power spectra of LFP responses evoked in area 17 by global stimulation. Stimulus frequency increases in steps of 2 Hz from 2 Hz to 50 Hz. (A) Surface
plot of spectra computed from steady state responses (500–1499 ms); response frequencies are cut at 100 Hz. Contour plot from the same data, conventions as
in Fig. 4. Both plots clearly show the linear relationships between stimulus and response frequencies in the fundamental, first and second harmonic, and
sometimes also in the third harmonic. (C) Distribution of power (ordinate) of the fundamental, and the first and second harmonics as a function of stimulus
frequency (abscissa). (D) Distribution of the ratios between the power of first and second harmonics over the fundamental as a function of stimulus frequency.
(E) Distribution of power (ordinate) of fundamental, first and second harmonic as function of response frequency (abscissa).
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
1870 G. Rager and W. Singer
FIG. 8. Power spectra of LFP responses evoked in area 18 by global stimulation. (Conventions as for Fig. 7.)
The response of cat visual cortex to flicker stimuli 1871
FIG. 9. Spectral analysis of LFP responses in area 17 after global visual stimulation with simultaneous MRF activation. Surface (A) and contour plots (B) are
shown together with the ratios of first and second harmonics over the fundamental (C). Same conventions as in Figs 7 and 8.
wave sleep or states of drowsiness (for review see Steriade et al.,
1990). Because of their long time constants, these interactions do not
support high frequency oscillations. Thus, recruitment of these lateral
interactions by whole-field stimuli could be one of the reasons for
the enhanced frequency attenuation with large stimuli: although there
is now in vitro evidence for the visual cortex that interacting
GABAergic neurons can sustain oscillatory activity in the γ-range
(Traub et al., 1996), the possibility should not be dismissed that
lateral interactions, in addition to providing the substrate for the
synchronization of responses, can also impede the entrainment of
large populations of neurons into synchronous oscillations. Theoretical
considerations and simulation studies have suggested that there
ought to be a subset of tangential intracortical connections with a
desynchronizing effect in order to prevent global synchronization of
cortical activity in the γ-frequency range (König & Schillen, 1990;
Schillen & König, 1994).
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
1872 G. Rager and W. Singer
FIG. 10. Spectral analysis of LFP responses in area 18 after global visual and simultaneous MRF stimulation. Surface (A) and contour plots (B) are shown
together with the ratios of first and second harmonics over the fundamental (C). Same conventions as in Fig. 9.
During states characterized by low frequency oscillations, large
and spatially contiguous arrays of neurons tend to discharge in
synchrony; hence the large amplitudes of low-frequency oscillations.
This is not so for γ-oscillations. Here, synchronous firing seems to
be confined to small and topographically dispersed clusters of neurons
that share certain functional properties, and are usually interleaved
with clusters of other neurons that either do not participate in any
synchronized activity at all or are synchronized to other partners
(Engel et al., 1991; Kreiter & Singer, 1996). Such topologically
specific temporal patterning is only possible if there are mechanisms
that prevent global entrainment.
Precision of stimulus locking
The comparison of power spectra computed from single sweeps with
those from averaged responses indicated that phase locking of the
fundamental response component was equally precise over the whole
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
The response of cat visual cortex to flicker stimuli 1873
FIG. 11. Spectral analysis of MUA (A,C,D) and LFP responses (B,E,F) from A17 evoked by flash stimuli confined to the receptive field. The contour plots in
(C) and (E), and the ratio plots of first and second harmonics in (D) and (F) correspond to the surface plots (A) and (B), respectively.
1874 G. Rager and W. Singer
The response of cat visual cortex to flicker stimuli 1875
frequency range tested. This agrees with the other evidence for
good entrainability of the various oscillatory mechanisms. Only the
harmonic response components which are more sensitive indicators
of resonance properties revealed regions of reduced entrainability in
the range of response frequencies between 12 and 30 Hz. This agrees
roughly with the frequency-dependent amplitude variations in the
averaged responses. Although the location of the troughs in these
distributions varied, there was a consistent decline in response
amplitude after the prominent peak in the α-frequency range.
Relation to perception and implications for feature binding
The result that at least a subpopulation of cortical neurons responds
with rather small latency or phase jitter to flashes, and does so over
a broad range of frequencies, agrees with the psychophysical evidence
that surprisingly small differences in the timing of stimuli can be
exploited for perceptual grouping. Stimuli that appear or disappear
simultaneously tend to be perceptually bound, but temporal offsets
as short as 8 ms are sufficient to support segregation of pattern
elements into different figures (Ramachandran & Rogers-Ramachandran, 1991; Leonards et al., 1996; but see Kiper et al., 1996). Psychophysical data indicate that in humans the signals supporting such
precise temporal judgements are conveyed by the non-colour-sensitive,
magno-cellular pathway (Leonards & Singer, 1997). Although the
magno-cellular system in primates cannot be directly equated with
the y-system in the cat, both share the property to respond reliably
and over a broad frequency range to stimuli containing temporal
transients (Schiller & Logothetis, 1990; Lee, 1996). The present
indication, derived from comparison between areas 17 and 18, that
the precisely locked responses must have been mediated mainly by
the y-system is then in general agreement with the psychophysical data.
FIG. 13. Phase locking of the response to the stimulus after global visual stimulation in area 18 without (A) and with MRF stimulation (B). The percentage of
phase locking is expressed as the ratio of the spectra of averaged responses (FFTsum) over the averaged spectra of single sweeps (FFTsingle) multiplied by 100.
FIG. 12. Spectral analysis of MUA (A,C,D) and LFP responses (B,E,F) evoked by whole-field flashes at the same recording site as in Fig. 11. Conventions are
the same as in Fig. 11.
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
1876 G. Rager and W. Singer
The present results indicate further that synchronization of cortical
responses due to stimulus locking occurred with a similar precision
(in the range of milliseconds) as the internal, feature-specific synchronization of responses to moving stimuli that is not locked to the
temporal structure of the stimuli. Since psychophysical evidence
indicates that stimulus-locked synchronization is exploited as signature
for the relatedness of stimuli, it is likely that the same also holds true
for synchronization established by internal interactions.
However, the halfwidth of the centre peak in the correlograms
changed considerably with stimulation frequency. At low stimulation
frequencies, the precision of synchronization was markedly lower
than that achieved by internal synchronization, and became comparable only at frequencies in the β- and γ-range. One possibility is that
timing information is carried by a special class of neurons that exhibit
phasic responses of equal temporal precision at all stimulation
frequencies, and that the sharpening of the responses at higher
frequencies was due to the drop-out of cells with more sustained
responses. The decrease in response amplitude at higher frequencies
is compatible with this view. This, however, raises the more general
question of how the responses are related to perception. Responses
to the first flashes in a series differed from those to later flashes, and
changing flicker frequency led to drastic changes in the amplitude
and time course of responses. Still, subjectively, early and late flashes
or flashes presented at different frequencies appear similar. This
suggests the possibility that only a particular component of the
response or only the responses of a particular class of neurons
contribute to the perception of the flashes. The relevant response
components should be those that do not change much throughout the
flash sequence and show only little frequency dependence. According
to our data, it would be the well-synchronized phasic response
components that contribute to the centre peak in the cross-correlograms, as these are equally well expressed at all flicker frequencies.
The tonic components that are pronounced at low frequencies and
are less well stimulus locked, as indicated by the broad correlation
peaks, cannot contribute substantially to flicker perception because
they disappear at higher frequencies. This interpretation is compatible
with recent psychophysical evidence that information about temporal
and non-temporal features of stimuli is conveyed and evaluated by
different systems: one that signals temporal transients with high
accuracy producing synchronized responses to simultaneous events
and asynchronous responses to temporally disjunct events; and another
that is rather insensitive to temporal gradients of stimuli and generates
sustained responses which reflect only poorly the temporal structure
of stimuli. In humans, at least, the first system could be equated with
the magno-cellular system and evidence indicates that it exploits the
temporal structure of stimuli for perceptual grouping (Leonards et al.,
1996), while the second system, which could be identified as the
colour-sensitive parvocellular system, appears to operate more independently of external timing cues and supports perceptual grouping
according to non-temporal features (Leonards & Singer, 1997).
Because the second system synchronizes less well to stimuli, it can
over-ride external timing information and bind or segregate stimulus
features independently of their temporal properties, probably relying
on internal synchronizing mechanisms. These, however, made no
measurable contribution to the synchronicity of the flash-evoked
responses investigated in this study, most likely because the uniform
stimuli lacked groupable, non-temporal features.
Acknowledgements
We wish to thank Michael Stephan at the Max Planck Institute for his
assistance in the development of evaluation programs, Patrick Faeh at the
Institute of Anatomy at Fribourg whose expertise in data processing was of
invaluable help in the computation of the figures and part of the evaluations,
and Irmi Pipacs for editing the manuscript.
Abbreviations
ACF
CCF
CCG
EEG
FFT
LFP
MRF
MUA
PSTH
RF
auto-correlation function
cross-correlation function
electrocardiogram
electroencephalogram
Fast Fourier Transform
local field potential
mesencephalic reticular formation
multi-unit activity
peri-stimulus time histogram
receptive field
References
Amzica, F., Nunez, A. & Steriade, M. (1992) Delta frequency (1–4 Hz)
oscillations of perigeniculate thalamic neurons and their modulation by
light. Neuroscience, 51, 285–294.
Basar, E. & Bullock, T.H. (1992) Induced Rhythms in the Brain. Birkhäuser,
Boston.
Bressler, S.L. (1996) Interareal synchronization in the visual cortex. Behav.
Brain Res., 76, 37–49.
Bressler, S.L. & Nakamura, R. (1993) Inter-area synchronization in macaque
neocortex during a visual pattern discrimination task. In Eeckman, F.H. &
Bower, J.M. (eds), Computation and Neural Systems. Kluwer Academic,
Boston., pp. 515–522.
Contreras, D. & Steriade, M. (1997) State-dependent fluctuations of lowfrequency rhythms in corticothalamic networks. Neuroscience, 76, 25–38.
Eckhorn, R., Bauer, R., Jordan, W., Brosch, M., Kruse, W., Munk, M. &
Reitboeck, H.J. (1988) Coherent oscillations: A mechanism for feature
linking in the visual cortex? Biol. Cybern., 60, 121–130.
Engel, A.K., König, P., Gray, C.M. & Singer, W. (1990) Stimulus-dependent
neuronal oscillations in cat visual cortex: Inter-columnar interactions as
determined by cross-correlation analysis. Eur. J. Neurosci., 2, 588–606.
Engel, A.K., König, P. & Singer, W. (1991) Direct physiological evidence for
scene segmentation by temporal coding. Proc. Natl. Acad. Sci. USA, 88,
9136–9140.
Gerstein, G.L. & Perkel, D.H. (1972) Mutual temporal relationship among
neuronal spike trains. Statistical techniques for display and analysis. Biophys.
J., 12, 453–473.
Gray, C.M., König, P., Engel, A.K. & Singer, W. (1989) Oscillatory responses
in cat visual cortex exhibit inter-columnar synchronization which reflects
global stimulus properties. Nature, 338, 334–337.
Gray, C.M. & McCormick, D.A. (1996) Chattering cells: Superficial pyramidal
neurons contributing to the generation of synchronous oscillations in the
visual cortex. Science, 274, 109–113.
Gray, C.M. & Singer, W. (1987) Stimulus-specific neuronal oscillations in the
cat visual cortex: a cortical functional unit. Soc. Neurosci. Abstr., 13,
404.0–404.3.
Gray, C.M. & Singer, W. (1989) Stimulus-specific neuronal oscillations in
orientation columns of cat visual cortex. Proc. Natl. Acad. Sci. USA, 86,
1698–1702.
Gutfreund, Y., Yarom, Y. & Seger, J. (1995) Subthreshold oscillations
and resonant frequency in guinea-pig cortical neurons: physiology and
modelling. J. Physiol. (Lond.), 483, 621–640.
Hoffmann, K.P. & Stone, J. (1971) Conduction velocity of afferents to cat
visual cortex: a correlation with cortical receptive field properties. Brain
Res., 32, 460–466.
Hoffmann, K.P., Stone, J. & Sherman, S.M. (1972) Relay of receptive field
properties in dorsal lateral geniculate nucleus of the cat. J. Neurophysiol.,
35, 518–531.
Hutcheon, B., Miura, R.M. & Puil, E. (1996) Subthreshold membrane
resonance in neocortical neurons. J. Neurophysiol., 76, 683–697.
Jefferys, J.G.R., Traub, R.D. & Whittington, M.A. (1996) Neuronal networks
for induced ‘40Hz’ rhythms. Trends Neurosci., 19, 202–208.
Kiper, D.C., Gegenfurtner, K.R. & Movshon, J.A. (1996) Cortical oscillatory
responses do not affect visual segmentation. Vision Res., 36, 539–544.
König, P., Engel, A.K., Löwel, S. & Singer, W. (1993) Squint affects
synchronization of oscillatory responses in cat visual cortex. Eur.
J. Neurosci., 5, 501–508.
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877
The response of cat visual cortex to flicker stimuli 1877
König, P. & Schillen, T.B. (1990) Segregation of oscillatory responses by
conflicting stimuli—desynchronizing connections in neural oscillator layers.
In Eckmiller, R., Hartman, G. & Hauske, G (eds), Parallel Processing in
Neural Systems and Computers. Elsevier, Amsterdam, pp. 117–120.
Kreiter, A.K. & Singer, W. (1996) Stimulus-dependent synchronization of
neuronal responses in the visual cortex of the awake macaque monkey.
J. Neuroscience, 16, 2381–2396.
Lee, B.B. (1996) Minireview: receptive field structure in the primate retina.
Vision Res., 36, 631–644.
Leonards, U. & Singer, W. (1997) Selective temporal interactions between
processing streams with differential sensitivity for colour and luminance
contrast. Vision Res., 37, 1129–1140.
Leonards, U., Singer, W. & Fahle, M. (1996) The influence of temporal phase
differences on texture segmentation. Vision Res., 36, 2689–2697.
Llinas, R.R., Grace, A.A. & Yarom, Y. (1991) In vitro neurons in mammalian
cortical layer 4 exhibit intrinsic oscillatory activity in the 10- to 50-Hz
frequency range. Proc. Natl. Acad. Sci. USA, 88, 897–901.
Löwel, S. & Singer, W. (1992) Selection of intrinsic horizontal connections
in the visual cortex by correlated neuronal activity. Science, 255, 209–212.
Lu, S.M., Guido, W., Vaughan, J.W. & Sherman, S.M. (1995) Latency
variability of responses to visual stimuli in cells of the cat’s lateral geniculate
nucleus. Exp. Brain Res., 105, 7–17.
von der Malsburg, C. (1985) Nervous structures with dynamical links. Ber.
Bunsenges. Phys. Chem., 89, 703–710.
Mastronarde, D.N. (1989) Correlated firing of retinal ganglion cells. Trends
Neurosci., 12, 75–80.
McCormick, D.A. & Bal, T. (1994) Sensory gating mechanisms of the
thalamus. Curr. Opinion Neurobiol., 4, 550–556.
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.
Mitzdorf, U. & Singer, W. (1978) Prominent excitatory pathways in the cat
visual cortex (A17 and A18): A current source density analysis of electrically
evoked potentials. Exp. Brain Res., 33, 371–394.
Munk, M.H.J., Roelfsema, P.R., König, P., Engel, A.K. & Singer, W. (1996)
Role of reticular activation in the modulation of intracortical synchronization.
Science, 272, 271–274.
Murthy, V.N. & Fetz, E.E. (1996a) Oscillatory activity in sensorimotor cortex
of awake monkeys: synchronization of local field potentials and relation to
behavior. J. Neurophysiol., 76, 3949–3967.
Murthy, V.N. & Fetz, E.E. (1996b) Synchronization of neurons during local
field potential oscillations in sensorimotor cortex of awake monkeys.
J. Neurophysiol., 76, 3968–3982.
Neuenschwander, S. & Singer, W. (1996) Long-range synchronization of
oscillatory light responses in the cat retina and lateral geniculate nucleus.
Nature, 379, 728–733.
Nunez, A., Amzica, F. & Steriade, M. (1992) Intrinsic and synaptically
generated delta (1–4 Hz) rhythms in dorsal lateral geniculate neurons and
their modulation by light induced fast (30–70 Hz) events. Neuroscience,
51, 269–284.
Puil, E., Meiri, H. & Yarom, Y. (1994) Resonant behavior and frequency
preferences of thalamic neurons. J. Neurophysiol., 71, 575–582.
Ramachandran, V.S. & Rogers-Ramachandran, D.C. (1991) Phantom contours:
A new class of visual patterns that selectively activates the magnocellular
pathway in man. Bull. Psychonomic Soc., 29, 391–394.
Regan, D. & Spekreijse, H. (1986) Evoked potentials in Vision Research
1961–86. Vision Res., 26, 1461–1480.
Roelfsema, P.R., Engel, A.K., König, P. & Singer, W. (1997) Visuomotor
integration is associated with zero time-lag synchronization among cortical
areas. Nature, 385, 157–161.
Schillen, T.B. & König, P. (1994) Binding by temporal structure in multiple
feature domains of an oscillatory neuronal network. Biol. Cybern., 70,
397–405.
Schiller, P.H. & Logothetis, N.K. (1990) The color-opponent and broad-band
channels of the primate visual system. TINS, 13, 392–398.
Sherman, S.M. & Guillery, R.W. (1996) Functional organization of
thalamocortical relay. J. Neurophysiol., 76, 1367–1395.
Singer, W. (1977) Control of thalamic transmission by cortico-fugal and
ascending reticular pathways in the visual system. Physiol. Rev., 57,
386–420.
Singer, W. (1993) Synchronization of cortical activity and its putative role in
information processing and learning. Ann. Rev. Physiol., 55, 349–374.
Singer, W. & Gray, C.M. (1995) Visual feature integration and the temporal
correlation hypothesis. Annu. Rev. Neurosci., 18, 555–586.
Steriade, M. & Amzica, F. (1996) Intracortical and corticothalamic coherency
of fast spontaneous oscillations. Proc. Natl. Acad. Sci. USA, 93, 2533–2538.
Steriade, M., Amzica, F. & Contreras, D. (1996) Synchronization of fast (30–
60 Hz) spontaneous cortical rhythms during brain activation.
J. Neuroscience, 16, 392–417.
Steriade, M., Curro-Dossi, R., Paré, D. & Oakson, G. (1991) Fast oscillations
(20–40 Hz) in thalamocortical systems and their potentiation by mesopontine
cholinergic nuclei in the cat. Proc. Natl. Acad. Sci. USA, 88, 4396–4400.
Steriade, M., Gloor, P., Llinás, R.R., Dasilva, F.H.L. & Mesulam, M.M. (1990)
Basic mechanisms of cerebral rhythmic activities. Electroenceph. Clin.
Neurophysiol., 76, 481–508.
Steriade, M. & McCarley, R.W. (1990) Brainstem Control of Wakefulness and
Sleep. Plenum Press, New York.
Steriade, M., McCormick, A. & Sejnowski, T.J. (1993) Thalamocortical
oscillations in the sleeping and aroused brain. Science, 262, 679–685.
Stone, J. (1973) Sampling properties of microelectrodes assessed in the cat’s
retina. J. Neurophysiol., 36, 1071–1079.
Stone, J. & Dreher, B. (1973) Projection of X- and Y-cells of the cat’s lateral
geniculate nucleus to areas 17 and 18 of visual cortex. J. Neurophysiol.,
36, 551–567.
Traub, R.D., Whittington, M.A., Stanford, I.M. & Jefferys, J.G.R. (1996) A
mechanism for generation of long-range synchronous fast oscillations in
the cortex. Nature, 383, 621–624.
© 1998 European Neuroscience Association, European Journal of Neuroscience, 10, 1856–1877