Attention is a critical factor in determining what we
perceive. At any moment we can give full attention to
only a tiny fraction of the available sensory information.
In a laboratory setting it is easy to show that attending
to a particular spatial location improves thresholds for
discrimination and speeds responses to stimuli at that
location. In some situations attention can reliably make
the difference between detecting a stimulus or missing
it entirely. Such pronounced changes in perception must
be associated with substantial changes in the way that
the brain processes sensory information. Our laboratory
is investigating how attention affects the way that individual
neurons represent visual information.
We use microelectrodes to record the activity of neurons
in the visual regions of the cerebral cortex of rhesus
monkeys. Rhesus monkeys have excellent vision, comparable
in many ways to that of humans. The monkey's visual system
has been extensively studied, and much has been learned
about its functional organization. The visual cerebral
cortex, which lies at the back of the brain, contains
dozens of discrete areas, each of which has its own representation
of the visual scene. Each area contains neurons that are
specialized for representing a particular type of visual
information. For example, some areas are specialized to
represent motion. Each neuron within them responds selectively
to stimuli moving in a particular direction, but is insensitive
to the color, size or shape of stimuli. Neurons in other
cortical areas have complementary properties, and are
specialized to represent other information, such as the
orientation of edges. Cortical areas also differ in the
complexity of the sensory information that they represent.
For example, while some cortical areas contain neurons
that respond well to any contour or edge, neurons in other
areas respond only to more elaborate shapes or patterns.
The responses of neurons within these specialized cortical
areas are determined not only by inputs coming from the
eyes, but also by top-down influences related to attention.
Neurophysiological studies from many laboratories have
shown effects of manipulating attention on the responses
of individual neurons in the visual cortex of trained
monkeys. Neuronal responses are generally stronger when
the animal pays attention to that stimulus.
Some of our recent experiments have been directed at
understanding how attention affects the quality of sensory
signals in cerebral cortex. While it is known that attention
makes sensory signals stronger, we are interested in learning
whether attention makes neuronal responses more selective.
Each cortical neuron is selective for particular stimulus
dimensions, such as color, orientation, or direction of
motion; it responds strongly only to a particular range
of stimuli that matches its sensitivity. Highly selective
neurons, which respond only to a narrow range of stimuli,
provide the most precise information about the visual
scene. We examined whether attention to stimulus orientation
affects orientation-selective neurons by restricting their
responses to a narrower range of orientations.
We trained monkeys to watch a display that contained
two stimuli: a small grating pattern and a small patch
of color. On some trials, the monkey had to report the
orientation of the grating, while on others it had to
report the color of the other stimulus. In either case
the animal had to keep its gaze fixed on a small fixation
spot at the center of the display, so that the retinal
stimulation was the same in both cases. The grating was
positioned to optimally activate the neuron that we recorded.
By changing the orientation of the grating from trial
to trial, we measured the range of orientations to which
a neuron responded under two conditions, one when the
animal was paying attention to the grating and the other
when it was ignoring the grating and paying attention
to the patch of color.
We measured the orientation selectivity of neurons in
area V4, which is an important stage in cortical analysis
of information about orientation and shape. As expected,
responses of V4 neurons were stronger when the animal
paid attention to the grating stimulus. Attention did
not, however, change the selectivity of these neurons.
Instead, responses to all orientations increased proportionately.
Thus, attention effectively increases the gain of a neuron's
response. This result suggests that what attention does
to the cortical representation of the visual scene is
roughly equivalent to adjusting the contrast on video
display. However, this adjustment is not uniform across
the whole representation; instead attention selectively
enhances those parts of the scene that are of immediate
importance.
This increase in the gain of neuronal responses suggests
that the effect of attention is limited to making sensory
responses stronger. Other experiments in our laboratory
have been exploring whether all neuronal and behavioral
effects of spatial attention might be explained in this
way. These experiments have examined whether the benefits
of attending to visual stimulus are quantitatively equivalent
to making that stimulus stronger. We trained animals to
do a task in which they had to release a lever when they
detected that a stimulus started to move in a particular
direction. We could adjust the stimulus so that the motion
was easier or more difficult to detect. The visibility
of the motion was varied from trial to trial, allowing
us to measure behavioral performance and neuronal responses
to a range of stimulus strengths. By presenting two stimuli
and directing the animal's attention to one, we could
measure the effect of attention on neuronal responses
to different motion strengths, and the animal's ability
to detect those motions. With this design we could determine
whether attention has effects that are quantitatively
equivalent to presenting a stronger sensory signal. We
asked: If a weak motion at the attended location produces
the same neuronal response as a strong motion at the unattended
location, will the animal's performance be the same in
the two cases?
We recorded the responses motion-sensitive neurons in
the middle temporal visual area (MT), which is an important
early stage in the processing of motion in visual cortex.
Although neurons in MT responded more strongly when the
animal attended to the stimulus, the effect of attention
on MT neurons was too small to account for its effect
on behavioral performance.
Because effects of attention are stronger in later stages
of cortical processing, we examined the ventral intraparietal
area (VIP) which represents a later stage of motion processing.
Neurons in VIP are direction selective, like those in
MT, but also include more elaborate response properties,
such as sensitivity to proximity. In VIP, we found that
many neurons were strongly modulated by whether the animal
was paying attention to the motion stimulus in their receptive
field. The average attentional modulation in VIP was too
strong to match the behavioral effects of attention. That
is, the changes in the neuronal responses suggested that
the animal should have done much worse than it did when
attention was directed to the incorrect location.
Although this result was unexpected, we believe it can
be explained by the fact that attentional modulations
grow stronger at successive levels of cortical processing.
Because attentional modulations differ between levels
of processing, there could be only one level that has
a correspondence between attentional modulation of neuronal
responses and behavior. We are currently exploring levels
of motion processing between MT and VI P to see if this
expectation holds up.
This work should lead to a more complete understanding
of the role of attention in creating the representations
that underlie sensation and perception, and the neuronal
mechanisms involved. Our long-term goal is to extend the
scope of this work to explore the mechanisms that transform
these representations into decisions and actions.
