In perceiving motion, as in determining color, the brain constructs a view of the world from pieces of information that can themselves be mistaken or ambiguous.
Suppose you paint an X on a piece of paper and then move that paper up and down in front of someone's eyes. Direction-selective cells in the motion-pathway layer of V1each of which sees only a small part of the scenewill respond to the diagonal orientation of each of the lines making up the X but will not register the movement of the X as a whole. How, then, is this overall movement sensed?
There must be two stages of motion analysis in the cortex, suggested Movshon and Edward Adelson, then a postdoctoral fellow at New York University (he is now a professor at the Massachusetts Institute of Technology). At the second stage, certain cells must integrate the signals regarding the orientation of moving lines and produce an overall signal about the motion of the whole object.
When Movshon presented this idea at an annual meeting of vision researchers in 1981, he was approached by William Newsome, an HHMI investigator at Stanford University, who was then a postdoctoral fellow at the National Institutes of Health. A lively three-hour dinner ensued and the two men resolved to collaborate. Together with Adelson, they would search for such cells in the motion area.
The researchers soon found that one-third of MT's cells could, in fact, signal the direction in which a hand waves through space. Later on, Albright's research group showed that MT cells can detect "transparent" motion, such as a shadow sweeping across the ground.
Then Allman and his colleagues discovered that many MT cells are able to integrate motion information from a large swath of the scene. "Even though an MT cell may respond directly to just one spot in the visual field," says Allman, "the cells have knowledge of what's going on in the region surrounding them."
Using a computer display with a background texture that looks vaguely like a leafy forest, Allman showed that some MT cells will fire particularly furiously if the leafy background moves in a direction opposite to a moving objectthe sort of visual pattern a cheetah would see when chasing an antelope along a stand of trees.
If, however, the background moved in the same direction as the moving object, the cell's firing was suppressed. The cell acted as a large-scale detector of motion contrast, performing exactly the sort of operation an animal would need to sense a figure moving through the camouflage of the forest.
While MT cells do not respond to static forms and colors, Albright has found that they will detect a moving object much more easily if its form or color strongly contrasts with its background.
"Imagine you're looking down the concourse in Grand Central Station and you're supposed to find the woman in the red dress," says Albright. "There are hundreds of surrounding people moving in different directions. Yet there's no problem at all in detecting the woman in the red dress walking along. Your visual system uses the dress's color to filter out all the irrelevant noise around it and homes in on the moving object of interest."
Suppose scientists could record from the MT cells in a laboratory monkey looking at the woman in the red dress crossing Grand Central Station. They could determine that a particular cell fired when the woman in the red dress passed through its receptive field. But how would they know that the firing of this specific MT celland not a network of thousands of other cells in the brain, of which this cell is only one nodeactually "causes" the monkey to perceive the woman's direction of movement? How could they ever get inside the monkey's mind and determine what it perceives?
Since Hubel and Wiesel's pioneering studies in the visual cortex, most visual scientists have assumed that the perception of form, color, depth, and motion corresponds to the firing of cells specialized to detect these visual qualities.
In a spectacular series of experiments conducted since the mid-1980s, Newsome, who is now a professor of neurobiology at Stanford University's School of Medicine, and his colleagues at Stanford have been directly testing this link between perception and the activity of specific neurons.
They use a device that was developed in Movshon's laboratory at NYU: a blizzard of white dots moving on a computer monitor. When all the white dots are moving randomly, the display looks like a TV tuned to a nonbroadcasting channel. However, the experimenters can gradually raise the percentage of dots moving in the same direction. When 10 percent of the dots move coherently together, their motion becomes apparent. By 25 percent, it is unmistakable.
Movshon had found that whenever a human being could detect the dots' motion at all, he or she could also tell the direction in which the dots were moving. "This means that the part of the visual pathway carrying the information used for motion detection is also carrying a label that says what direction is being detected," says Movshon." This is precisely how one would expect MT, with its columns of direction-selective cells, to encode a moving target.
Next, Newsome began to teach rhesus monkeys to "tell" him what they saw on the computer screen. When they saw dots moving downward, for instance, the monkeys were supposed to move their eyes to a downward point on the screen. Correct answers were rewarded with fruit juice. Soon the monkeys could signal with eye movements that they saw the dots move in any of six directions around the clock. And after much training on low-percentage moving dot displays, the monkeys were able to perform nearly as well as Movshon's human subjects.
Everything was in place. Newsome, Movshon, and their colleagues were ready to study the relationship between the monkeys' perception of motion and the activity of cells in particular columns of MT.
"We found, very much to our surprise," says Newsome, "that the average MT cell was as sensitive to the direction of motion as the monkey was." As more dots moved together and the monkey's ability to recognize their direction increased, so did the firing of the MT neuron surveying the dots.
If the monkeys were actually "listening" to the cells in a single MT column as they made their decision about the direction of movement of the dots on the screen, could the decision be altered by stimulating a different MT column, the researchers wondered. So they stimulated an MT "up" column electrically while the monkeys looked at the downward-moving display. This radically changed the monkeys' reports of what they saw.
"It was an unforgettable experience," remembers Newsome. "We got the first of what became known in the lab as 'Whoppers'when the effects of microstimulation were just massive. Fifty percent of the dots would move down, and yet if we'd stimulate an 'up' column, the monkey would signal up with its eyes."
The monkeys' perceptual responses no longer seemed to be driven by the direction of dots on the screen. Instead, the animals' perceptual responses were being controlled by an electric stimulus applied to specific cells in the brain by an experimenter.
These experiments, says Movshon, "close a loop between what the cells are doing and what the monkey's doing." Allman calls the finding "the most direct link that's yet been established between visual perception and the behavior of neurons in the visual cortex."
It is still possible, however, that when the dots are moving down and the experimenters stimulate an MT "up" column, the stimulation changes what the monkey "decides" without actually changing what it "sees."
"This is a key question," says Newsome. "We now know a lot about the first and last stages of this process. But we are almost totally ignorant about the decision process out there in the middlethe mechanism that links sensory input to the appropriate motor output. How does the decision get made?"
It is a burning question not only for research on the visual system, but for all of cognitive neuroscience, Newsome believes. The answer would provide a bridge from the study of the senses, where so much progress has been made, to the much more difficult study of human thought. At long last, Newsome says, "we're now poised to approach this question."
< Previous | Top of page