Neuron

Previews

between the purpose and its functional
basis, and recognize that the activation
pattern of motor cortex neurons does
two thingsit specifies for the peripheral
motor system both what to do and how
to do it.
REFERENCES
Ajemian, R., Green, A., Bullock, D., Sergio, L.,
Kalaska, J., and Grossberg, S. (2008). Neuron 58,
this issue, 414428.
Crutcher, M.D., and Alexander, G.E. (1990). J.
Neurophysiol. 64, 151163.
Evarts, E.V. (1968). J. Neurophysiol. 31, 1427.
Georgopoulos, A.P., Kalaska, J.F., Caminiti, R.,
and Massey, J.T. (1982). J. Neurosci. 11, 1527
1537.
Georgopoulos, A.P., Schwartz, A.B., and Kettner,
R.E. (1986). Science 233, 14161419.
Georgopoulos, A.P., Ashe, J., Smyrnis, N., and
Taira, M. (1992). Science 256, 16921695.
Hubel, D.H., and Wiesel, T.N. (1962). J. Physiol.
160, 106154.
Kalaska, J.F., Cohen, D.A., Hyde, M.L., and Prud-
homme, M. (1989). J. Neurosci. 6, 20802102.
Lemon, R.N. (2008). Annu. Rev. Neurosci., in press.
Published online April 4, 2008. 10.1146/annurev.
neuro.31.060407.125547.
Lennie, P., and Movshon, J.A. (2005). J. Opt. Soc.
Am. A 22, 20132033.
Moran, D.W., and Schwartz, A.B. (1999). J. Neuro-
physiol. 82, 26762692.
Pesaran, B., Nelson, M.J., and Andersen, R.A.
(2006). Neuron 51, 125134.
Scott, S.H., and Kalaska, J.F. (1997). J. Neurophy-
siol. 77, 826852.
Sergio, L.E., and Kalaska, J.F. (2003). J. Neurophy-
siol. 89, 212228.
Sergio, L.E., Hamel-Paquet, C., and Kalaska, J.F.
(2005). J. Neurophysiol. 94, 23532378.

Finding Gamma
Pascal Fries,1,2,* Rene Scheeringa,1 and Robert Oostenveld1
1F.C.Donders Centre for Cognitive Neuroimaging, Radboud University Nijmegen, 6525 EN Nijmegen, The Netherlands
2Department of Biophysics, Donders Centre for Neuroscience, Radboud University Nijmegen, 6525 EZ Nijmegen, The Netherlands
*Correspondence: pascal.fries@fcdonders.ru.nl
DOI 10.1016/j.neuron.2008.04.020

Neuronal gamma-band synchronization is central for cognition. Respective studies in human subjects fo-
cused on a visually induced transient enhancement of broadband EEG power. In this issue of Neuron,
Yuval-Greenberg et al. demonstrate that this EEG response is an artifact of microsaccades, raising the ques-
tion of whether gamma-band synchronization can be assessed with EEG.

When networks of neurons are activated,
they engage in synchronous rhythmic
activity in the gamma-frequency range
(30100 Hz) (Gray et al., 1989). This
gamma-band synchronization affects
neuronal interactions (Womelsdorf et al.,
2007) and thereby subserves several cen-
tral cognitive functions, including percep-
tual binding (Gray et al., 1989), attentional
selection (Fries et al., 2001), and working
memory maintenance (Pesaran et al.,
2002). These functions of gamma-band
synchronization have been revealed in
numerous experiments in animals, using
microelectrodes that record single neu-
rons, small groups of neurons, or the local
field potential (LFP, a sort of EEG re-
corded inside the neuropil). The LFP is
due to intra- and extracellular current
flows that can also be measured noninva-
sively as magnetoencephalogram (MEG)
or electroencephalogram (EEG).
The EEG has been used extensively in
human cognitive neuroscience, because
it is relatively cheap and easy, but
nevertheless delivers noninvasive mea-
surements of human brain activity with
millisecond temporal precision. This pre-
cision has been exploited predominantly
to study brain responses with a strict tem-
poral relation to either a sensory stimulus,
a motor response, or any other externally
accessible event. The respective event is
used to trigger the averaging of EEG
epochs to obtain the event-related poten-
tial (ERP). The underlying rationale is that
any brain response related to the event
is phase locked to it and survives averag-
ing, while anything else is noise and is re-
moved through the averaging. However,
the absence of phase locking is precisely
a characteristic feature of the neuronal
gamma-band synchronization that had
been observed in animals. The microelec-
trode recordings in animals revealed
consistently that, for example, visual stim-
uli induced synchronized rhythms that oc-
curred in each trial with a different phase
relation to stimulus onset. The variable
phase relation makes those components
disappear in ERPs, and they can only be
revealed if the spectral (frequency-wise)
power of neuronal activity is estimated
separately per trial and only then aver-
aged.
Such a power analysis in turn retains
not only the interesting gamma-band
rhythm, but also power from, for example,
small muscle artifacts. These muscle
artifacts contain power actually predomi-
nantly in the gamma band, and it is
precisely this reason why most re-
searchers prefer to low-pass filter EEG
signals around 30 Hz, eliminating many
potential artifacts but also any potential
gamma-band activity. Thus, both the

Neuron 58, May 8, 2008 2008 Elsevier Inc. 303

 Neuron

Previews

movements are controlled. Yuval-Green-

berg et al. (2008) demonstrate with
excruciating clarity that the iGBR is
a manifestation of miniature saccades.
The contraction of extraocular muscles
during miniature saccades (MSs) causes
a short spike in the EEG scalp recordings.
This spike is not actually rhythmic, just
short (a few tens of milliseconds), but in
the spectral power analysis, this corre-
sponds to a broad-band power elevation
between 30 and 100 Hz. Yuval-Greenberg
et al. can explain also the temporal
features of the iGBR out of the temporal
pattern of MSs. MSs occur with a sponta-
neous rate (roughly three per second)
already before stimulus onset. After stim-
ulus onset, this rate drops to almost zero
around 100 ms, rebounds to a peak be-
tween 200 and 300 ms, and then returns
back to baseline levels. To understand
how this explains the time course of the
iGBR, one has to realize that it is not
the slow rate of MSs that produces the
gamma rhythm, but rather, each MS pro-
duces a broad-band iGBR, and the rate of
MSs determines the strength of the iGBR.
Furthermore, the peak of MS rate be-
tween 200 and 300 ms poststimulus is
much broader than each individual MS-in-
Figure 1. Neuronal Gamma-Band Synchronization Can Be Assessed with EEG duced EEG spike. Therefore, the EEG
(A) Illustration of recording location in V1 (primary visual cortex) of an awake macaque monkey.
(B) Time-frequency analysis of LFP power recorded from the position indicated in (A).
spikes are not aligned to stimulus onset,
(C) Estimated sources of visually induced gamma-band activity as measured with MEG in a human and the iGBR is not phase locked to stim-
subject. ulation. And finally, the drop in MS rate
(D) Time-frequency analysis of MEG sensors over the source shown in (C).
around 100 ms after stimulus onset actu-
(E) Estimated sources of visually induced gamma-band activity as measured with EEG in another human
subject. ally corresponds to a drop in iGBR below
(F) Time-frequency analysis of an EEG electrode over the source shown in (E). baseline levels, which has been consis-
tently found, but hardly discussed.
Thus, both spectral and temporal iGBR
event-related averaging in ERPs and the fully aware of the risk of artifacts sneaking characteristics are well explained as MS
common low-pass filtering were blind- into the power spectra. Therefore, these artifacts. But what about the numerous
folding most of the EEG-based research issues were discussed in the initial manu- cognitive effects that have been docu-
for potential gamma-band activity. scripts, and some studies even contained mented for the iGBR? Yuval-Greenberg
Some investigators realized these direct estimates of the influence of muscle et al. could not address all of those effects
problems and took the courageous step artifacts (Tallon-Baudry et al., 1998). in one study. However, they performed an
to overcome ERP traditions, to open their Those considerations discarded an arti- experiment that is typical for iGBR stud-
filters, and to investigate non-stimulus- factual nature of the iGBR, and the studies ies, using visual stimuli in which parts
locked power at frequencies in the rather suggested that the iGBR actually could either be perceptually bound into
gamma-band range. This research fo- assessed neuronal gamma-band syn- a gestalt or not. When perceptual binding
cused on the visual modality and revealed chronization in human subjects. Indeed, was possible, iGBR was enhanced. This
that visual stimuli induce enhanced power iGBR was shown to correlate with per- iGBR finding would be in line with hypoth-
from 30100 Hz and around 200300 ms ceptual binding, priming, etc., and these eses about neuronal gamma-band syn-
after stimulus onset, a response that has cognitive modulations of the iGBR chronization, but Yuval-Greenberg et al.
often been called the visually induced seemed to provide growing evidence for show it to be fully explained by the pattern
gamma-band response, or iGBR (Tallon- it not being an artifact. of MSs in the different visual stimulation
Baudry et al., 1996; Yuval-Greenberg Yet, a paper in this issue of Neuron sug- conditions. While this does not prove
and Deouell, 2007). The involved investi- gests that those cognitive modulations of that MSs explain previously published
gators were seasoned EEG experts and the iGBR rather reveal how intricately eye iGBR effects, it does raise serious

304 Neuron 58, May 8, 2008 2008 Elsevier Inc.

Neuron
Previews
concerns. These concerns will need to be
addressed in reanalyses of existing data
and/or replications of previous studies
that will now include precise eye-position
recordings.
Yuval-Greenberg et al. explicitly do not
raise concern about neuronal gamma-
band synchronization itself. Neuronal
gamma-band synchronization has been
demonstrated directly with microelec-
trode recordings of pairs of single neurons
(Maldonado et al., 2000), pairs of neuron
groups (Gray et al., 1989), and in LFPs
(the latter reflect the precision of synchro-
nization in their amplitude) (Fries et al.,
2001) (Figures 1A and 1B). These record-
ings assess neuronal activity within only
tens to hundreds of micrometers around
the electrode tips, and electric fields
from eye muscle contractions do not af-
fect them. Importantly, gamma-band syn-
chronization assessed in this way has
a spectrotemporal pattern that is very dif-
ferent from the iGBR, but very consistent
across the different types of microelec-
trode recordings. It starts with the first
neuronal spiking response (Fries et al.,
2001), remains stable over the course of
the response, and, within a given individ-
ual, it is clearly restricted to a limited fre-
quency band of typically 1030 Hz width.
Visually induced gamma-band activity
with these characteristics has been re-
corded also in human subjects using
MEG, and source analyses located it to
early visual areas (Figures 1C and 1D)
(Hoogenboom et al., 2006). Thus, typical
neuronal gamma-band synchronization
is clearly present in the human brain,
and it should be possible to record it
also with the EEG. In Figures 1E and 1F,
we show visually induced gamma-band
activity in human EEG with a spectrotem-
poral signature and an estimated source
that are very similar to monkey LFP (Fig-
ures 1A and 1B) and human MEG (Figures
1C and 1D). This example demonstrates
that neuronal gamma-band synchroniza-
tion can be investigated with EEG, and
therefore, EEG researchers should keep
their filters wide open. The move from
ERPs to EEG spectral analysis remains
a central scientific challenge (Makeig
et al., 2004). The important correction by
Yuval-Greenberg et al. must not make
us shrink from this challenge, but rather
make us pursue it with even greater dili-
gence and caution.
Neuron 58, May 8, 2008 2008 Elsevier Inc. 305
REFERENCES
Fries, P., Reynolds, J.H., Rorie, A.E., and
Desimone, R. (2001). Science 291, 15601563.
Gray, C.M., Konig, P., Engel, A.K., and Singer, W.
(1989). Nature 338, 334337.
Hoogenboom, N., Schoffelen, J.M., Oostenveld,
R., Parkes, L.M., and Fries, P. (2006). Neuroimage
29, 764773.
Makeig, S., Debener, S., Onton, J., and Delorme,
A. (2004). Trends Cogn. Sci. 8, 204210.
Maldonado, P.E., Friedman-Hill, S., and Gray, C.M.
(2000). Cereb. Cortex 10, 11171131.
Pesaran, B., Pezaris, J.S., Sahani, M., Mitra, P.P.,
and Andersen, R.A. (2002). Nat. Neurosci. 5,
805811.
Tallon-Baudry, C., Bertrand, O., Delpuech, C., and
Pernier, J. (1996). J. Neurosci. 16, 42404249.
Tallon-Baudry, C., Bertrand, O., Peronnet, F., and
Pernier, J. (1998). J. Neurosci. 18, 42444254.
Womelsdorf, T., Schoffelen, J.M., Oostenveld, R.,
Singer, W., Desimone, R., Engel, A.K., and Fries,
P. (2007). Science 316, 16091612.
Yuval-Greenberg, S., and Deouell, L.Y. (2007). J.
Neurosci. 27, 10901096.
Yuval-Greenberg, S., Tomer, O., Keren, A.S.,
Nelken, I., and Deouell, L.Y. (2008). Neuron 58,
this issue, 429441.