Listening In on the Brain
Armed with many electrodes at once,
neurophysiologists are uncovering synchrony in neural firing that may
represent a new way of encoding information
If you try to record a symphony using a single
microphone that can pick up only the notes played by one musician at a
time, you are likely to get an extremely limited impression of the
music. Your recording will miss virtually all the symphony's melodic
and rhythmic form. That's the position some neurophysiologists say they
find themselves in when they try to understand how the brain works
using traditional methods of recording brain activity one neuron at a
Those techniques have provided--and still are providing--a wealth of
knowledge about how brain neurons encode information by varying their
firing rates. But as long as 3 decades ago, some neuroscientists, eager
to see what they were missing, started recording from multiple
electrodes at once. After a slow start, that approach is finally coming
of age. Recent results gleaned from such experiments suggest that the
brain encodes information, not just in the firing rates of individual
neurons, but also in the patterns in which groups of neurons work
One of the more tantalizing--and controversial--of these discoveries is
that neurons frequently fall into step with one another, forming
ensembles that play the same tune, as it were, firing in relative
synchrony for brief periods, before some neurons drop out of synch,
perhaps to join another ensemble. What's more, studies of systems as
diverse as the motor cortex of monkeys and the olfactory system of
honeybees indicate that these changing patterns of synchrony correlate
with specific behaviors such as a monkey's preparation to move its hand
or a honeybee's discrimination of odors.
These findings suggest that the patterns contain information, a
situation that if true "would be quite remarkable," says Eberhard Fetz,
a neurobiologist at the University of Washington, Seattle, because it
would "dramatically increase" the brain's information-encoding
capacity. But, adds Fetz, who studies synchronous neural activity in
monkeys, "a lot of people think there is no need for [a synchrony
code], because the firing rates themselves provide enough
[information-encoding] possibilities." Indeed, the case for a role for
synchrony is far from proven, says neuroscientist Ad Aertsen, of Albert
Ludwigs University in Freiburg, Germany, who designs methods for
analyzing data from multiple-electrode recordings. "There is still a
lot of need for É efforts to relate [the patterns] to behavior and see
what kind of information they might contain."
Tuning in to the ensembles
Single-electrode experiments have been so successful that
neuroscientists have had little incentive to take on the technically
challenging task of recording from many neurons at once, says
neuroscientist Moshe Abeles, of Hadassah Medical School in Jerusalem:
"Most researchers felt they could get a lot of exciting results with a
single electrode." Nevertheless, a handful of laboratories, led by
George Gerstein at the University of Pennsylvania--who began his
multielectrode experiments in the 1960s--have risen to the challenge.
The small field moved slowly at first, as researchers solved technical
problems and developed ways to handle the data. But by the 1980s, these
pioneers established that neurons often fire together for brief
periods--their action potentials mirroring each other to within a few
That synchronous firing suggested that neurons can form a functional
group, or ensemble, says Gerstein. And those ensembles are not fixed.
"In behaving [animals] or situations where you vary the stimulus, these
synchronizations occur dynamically," he says. "Assemblies come and go."
That fluid response to changing situations "was really a surprise,"
adds Abeles. It suggested that the changing patterns may encode
But if they did encode information, what was its function? A clue came
in the late 1980s from work by neurobiologist Wolf Singer and his
colleagues at the Max Planck Institute for Brain Research in Frankfurt,
Germany. Singer's team showed that neurons in the visual cortex of cats
that respond to different images on a screen, such as two bars, would
fire together when the bars moved with the same speed and direction, as
if they were part of the same object. But when the bars appeared to be
parts of different objects, the neurons fired at the same rates as
before but not in synchrony.
Those findings piqued neuroscientists' interest because they suggested
that synchronous firing may help solve the "binding problem"--the
brain's need to somehow link neural signals that are related to each
other, such as all the cues from neurons responding to different visual
aspects of an object (Science, 24 August 1990, p. 856).
More recent experiments by the Singer team suggest that synchrony may
determine not only how the brain perceives stimuli--for example, as
parts of the same object--but even whether it perceives them at all.
This conclusion comes from studies on strabismic cats, whose eyes are
misaligned like those of a cross-eyed person, so that each eye looks at
a different part of the surrounding world. The cats don't perceive both
conflicting images at once; the brain ignores the image from one eye or
the other, so the cat only perceives the image from one eye at a time.
The researchers used mirrors to reflect a different moving image on a
screen to each of a cat's eyes, and they monitored the cat's eye
movements to determine which image it perceived. When only one eye was
presented with an image, that is what the brain perceived. But the
researchers could snatch perception from that eye by presenting a rival
image to the other eye. By altering the comparative contrast of the
images, Singer's team found conditions under which a particular image
always won. Then they used electrodes to record the activity of neurons
in the cats' visual cortex that respond to signals from the eye
watching the original image, while using the rival image to control
which eye would win the battle for perception.
When they looked at just the firing rates of the neurons, they could
get no clue about whether the image was being perceived by the brain,
Singer says. The neurons' firing rates didn't change at all when
perception was shifted away. But, he adds, "there was a very clear
predictor of perception, and this was the change in synchronicity. When
it went up, we knew this eye was the winner; when it went down, we knew
this eye was the loser." That, says Singer, suggested that synchronous
firing "selectively raises the saliency" of the signal and "predisposes
that activity for further joint processing" by higher areas of the
visual system. The experiment is intriguing because it provides "an
example where synchrony does something that firing rates don't," says
Choreographing a move
Researchers studying other brain regions are finding apparent uses for
synchrony as well. In the motor cortex, where neurons fire to direct
movements, researchers have found patterns of synchronous activity that
seem to help prepare for a move. In one example, Alexa Riehle of the
CNRS Center for Research in Cognitive Neuroscience in Marseille,
France, in collaboration with Aertsen, trained monkeys to watch a
computer screen for a cue, and then, when they received a go command,
to touch the screen where the cue had appeared. In each trial, the go
command came either 600, 900, 1200, or 1500 milliseconds after the
place cue, an arrangement the team chose so that the animals would
learn to anticipate when to expect the go signal. They wanted to see
how that anticipation--"a pure cognitive signal," as Riehle calls it,
with no input from the outside world--would be represented in neural
The researchers simultaneously recorded the activity from up to seven
neurons in each monkey's motor cortex while it performed the task. The
neuronal firing rates jumped when the go signal told the monkey to
touch the screen. That was expected, as activity in the motor cortex
Firing patterns of individual neurons in
the antennal lobe of a locust show transient pairwise synchrony (open
boxes at right) in response to an odor. Such patterns help a honeybee,
shown at left with extended proboscis, recognize an odor.
MICHAEL WEHR AND GILLES LAURENT
But more interesting, Riehle says, were the patterns with which the
neurons fired relative to each other during the waiting period, when
the animals anticipated a go signal that didn't come. Although the
neurons didn't change their firing rates at the moments of
anticipation, says Riehle, "we detected a significantly higher amount
of synchronization than expected by chance" surrounding those moments.
The synchrony formed a changing and predictable pattern. Among the
seven neurons she studied in any trial, neuron number 1, say, might
briefly fire in synchrony with neuron 3, then fall out of step with 3
and join in with neuron 5.
What's more, the synchrony was most intense in those trials in which
the monkeys showed the shortest reaction times upon receiving the true
go signal, suggesting that the monkeys were highly motivated and
attentive. "We think of it [the synchrony] as ... the pretuning and
preshaping of the networks that will be involved ... in the execution
of the movement," says Riehle. As in Singer's experiments, the
synchrony seems to be performing a binding function, in this case
linking together the neurons that will need to act together to execute
Recent experiments by John Donoghue and his colleagues at Brown
University in Providence, Rhode Island, suggest that synchrony is
involved not only in binding together neurons during movement planning
but also in encoding information for actually making the movements.
Donoghue's team trained monkeys to use a mechanical drawing arm to move
a cursor on a computer screen to the location where a cue appeared.
When Donoghue's group recorded from up to 21 neurons at once in a
monkey's motor cortex while it did the task, they found that, on
average, 30% of the neurons fired in synchrony at any one time. A
mathematical analysis also showed that the synchrony patterns could
predict the direction of the monkey's arm movements. "We know [the
synchrony pattern] is carrying information about direction," says
Donoghue. "But we don't know whether that information is available to
the brain to be used for anything. All we know is it is there."
The signature of a sniff
One case in which the brain definitely seems to be using the
information encoded in synchronous firing is in the olfactory system of
insects. Gilles Laurent at the California Institute of Technology in
Pasadena and his colleagues study the part of an insect's brain called
the antennal lobe, where the neurons process smells. In locusts, the
lobe has about 1000 neurons, and about 100 of them fire action
potentials in response to any odor. Not only is information contained
in the pattern of activated neurons, Laurent says, but there is also
information in the neurons' timing with respect to one another. "It is
very clear that the temporal patterns are stimulus specific," he says.
"Knowing the precise timing of the spike gives you additional
information about the odor."
As in Donoghue's case, the existence of that information doesn't
necessarily mean the brain uses it. However, Laurent's team was in a
position to address that question directly. That's because his group
had discovered that a neurotoxin called picrotoxin abolishes the
synchronous firing of neurons in the antennal lobe in response to
odors, without altering their firing rates. Armed with the ability
specifically to get rid of the synchrony, Laurent's team was ready to
ask whether synchrony is essential to an insect's ability to
distinguish among odors.
To do that experiment, his team used honeybees, which can be trained to
recognize odors associated with a sugar-water reward. Working with
Brian Smith at Ohio State University, the researchers trained
picrotoxin-treated and control bees to associate an odor with a reward.
Bees show they have learned an odor by sticking out their tongues in
expectation of the reward. Both groups of bees did equally well.
But a difference between the groups emerged when the researchers tested
whether the bees could discriminate the familiar odor from other odors.
Both groups could do this just fine when the familiar odors were very
different from one another. But when the odors were very similar, the
picrotoxin group failed. That led Laurent to conclude that the timing
of the spikes in relation to each other is important for fine
discrimination of odors. The Laurent team's study, says Aertsen,
"really addresses the issue: Can you show that if the synchrony isn't
there, the system doesn't work."
But although "the olfactory system is certainly a specific place where
cofiring of neurons means a lot," says Hadassah's Abeles, he would like
to see more evidence in other systems that the brain is really using
the information encoded in the synchrony. Singer argues that his
binocular rivalry experiments make a good case that it is. "Since it is
such a strong predictor of perception, I can't imagine that the brain
ignores synchrony," he says. "It is inconceivable. It would be like
saying that the brain ignores firing rate."
Despite arguments such as Singer's, some researchers remain skeptical
about the function of synchronous firing in the brain. It may just be a
product of neurons changing their firing rates at the same time, argues
neuroscientist Michael Shadlen of the University of Washington. Like a
group of people who all start to run at once, their first few steps may
be nearly in unison.
Researchers who study synchrony often find no change in the firing
rates of neurons when they synchronize, but Shadlen notes that changes
in firing rate can only be seen when they are time locked to something
the researcher controls, such as the appearance of a cue on a computer
screen. That's because the only way to measure firing-rate changes
accurately is by averaging the response of neurons over many trials. If
the neurons are changing their rate in response to something the
experimenter doesn't control, such as when a monkey happens to make up
its mind, says Shadlen, that event may change with each trial, and
consequently the rate change may be invisible to the experimenter.
In that case, says Shadlen, studying synchrony would still be
important, because it can reveal mental events researchers otherwise
would have no way of detecting. "I'm saying synchrony lets us find the
rate changes, and they are saying synchrony is the code," says Shadlen.
Regardless of who is right, the study of the synchronous firing of
neurons is sure to help researchers record the music of the brain for
years to come.
P. Fries et al., "Synchronization of oscillatory responses in visual cortex correlates with perceptin in interocular rivalry," Proceedings of the National Academy of Sciences 94, 12699 (1997).
A. Riehle, S. Grün, M. Diesmann, Ad Aertsen, "Spike synchronization and
rate modulation differentially involved in motor cortical function," Science 278, 1950 (1997).
M. N. Shadlen and W. T. Newsome, "Noise, neural codes and cortical organization," Current Opinion in Neurobiology 4, 569 (1994).
M. Stopfer et al., "Impaired odour discrimination on desynchronization of odour-encoding neural assemblies," Nature 390, 70 (1997).