Rethinking How the Brain Sees Visual Features
DURHAM, N.C. -- Brain scientists will have to rethink the
current theory of how the visual processing region of the brain
is organized to analyze basic information about the geometry of
the environment, according to Duke University Medical Center
neurobiologists. In a new study reported in the June 26, 2003,
Nature, they studied the visual-processing region -- called the
visual cortex -- of ferrets, as the animals' brains responded
to complex patterns.
The results, they said, indicated that clusters of neurons
in that region do not specialize in recognizing a particular
combination of stimulus features, as previously believed.
Rather, individual clusters react to a broad range of stimulus
combinations -- combinations that can be predicted by
understanding the fundamental spatial and temporal properties
of the visual stimulus. The scientists' research was supported
by the National Eye Institute.
The visual cortex -- a layer of brain tissue at the back of
the head -- is the first area within the cerebral cortex that
processes neural signals from the eye. It performs the basic
tasks of recognizing the geometric features of a scene before
relaying that information to higher brain regions, where such
basic visual data are transformed into the conscious perception
of the visual world.
Current theory of visual cortex organization holds that in
mammals, including humans, the visual cortex consists of
overlapping "feature maps." Each map is an orderly arrangement
of neuronal clusters that represents a particular stimulus
feature, such as the orientation of edges, their direction of
motion, or their spacing. Before these new experiments were
performed, it was thought that the response properties of
neurons could be predicted by their location relative to the
places in the visual cortex where different feature maps
overlap. In this view, clusters of neurons are "specialists"
for the detection of certain combinations of visual features,
such as a set of parallel lines of a certain orientation,
spaced a certain distance apart and moving at a specific
In their experiments, Duke neurobiologists -- graduate
student Amit Basole, Assistant Professor Leonard White and
Professor David Fitzpatrick -- decided to go beyond previous
studies in which animals were exposed only to simple visual
stimuli consisting of parallel bars, or gratings, with
different spacings and moving at different speeds at a right
angle to the bars.
"Studies with gratings can tell you a lot," said
Fitzpatrick. "For example, you can get a sense of maps of
orientation if you change the orientation of the grating. And
you can also get information about how properties like spatial
frequency are mapped by changing the distance between the bars
in the grating, and mapping how that changes patterns of neural
"The underlying assumption was that, in a sense there was a
'place code' for stimulus combinations," said Fitzpatrick. "So,
a particular orientation, spatial frequency or direction would
activate a certain cluster of neurons in the cortex; and
changing the orientation, direction or spatial frequency would
shift the locus of activity in a predictable way -- one that
signified which attribute had been changed."
However, said Fitzpatrick, "these stimuli are really
limiting because you can only look at certain stimulus
combinations." To explore how the visual cortex reacted to more
complex combinations of stimuli, the researchers exposed
ferrets to patterns consisting of short line segments whose
orientation, length, direction and speed of motion could be
Said White, "With these texture patterns, we have the
ability to let different properties interact with one another
in ways that are closer to the kinds of stimulus interactions
that are often present in the visual environment." A striking
example of such interactions is the so-called barber pole
illusion, he said.
"While the barber pole is moving horizontally as the pole
spins about its axis, it creates a perception that the lines
are moving up," said White. "The perception induced by the
interaction between the orientation of the lines and the
direction of motion is the sort of phenomenon that Amit was
seeking to understand in terms of neural responses."
The researchers used a technique called optical imaging to
detect brain activity in the animals' visual cortex by shining
light of wavelengths that specifically revealed increased blood
flow to more active areas. Also, to confirm that the images
portrayed actual increases in brain activity, the researchers
also recorded electrical activity of individual neurons in
different cortical regions during exposure to the patterns.
The effects of changing the visual stimuli on the activity
patterns in the animals' brains were surprising, said
"From the prevailing view, if you kept the orientation of
the bars constant and varied the other parameters, you might
not expect to see much of a change in the maps of activity,"
said Fitzpatrick. But, in fact, we saw shifts in activity that
were much greater than we expected, and the patterns looked
identical to those that were produced by textures that had
different combinations of line orientation, direction, length
"So, this makes clear that thinking about maps in the cortex
as consisting of clusters devoted to particular combinations of
features is too simplistic when you're dealing with stimuli
that are much more like those you encounter in the visual
world," he said.
"What we're seeing is that a given spot in the cortex seems
to be integrating a number of different stimulus components.
All of these components figure into what determines the
activation of a given spot in the map."
In this new way of thinking about the visual cortex, it is
still possible to consider the clusters of neurons as
specialists; neurons in these new studies responded to complex
visual patterns with remarkable selectivity, said Fitzpatrick.
However, these findings show that what these clusters
specialize in is not the recognition of a unique combination of
stimulus features, but the detection of a narrow band of
spatial and temporal information that may be produced by a
surprising large combination of stimulus features.
The researchers plan further studies to attempt to
understand how the visual cortex is organized -- for example,
seeking to obtain faster snapshots of brain activity, to obtain
more detail in changes in brain activity. They are also working
with other colleagues to create mathematical models that might
reveal the strategy by which the brain has organized its visual