 |
From Vision to Action: Sensory-Motor Processing for Eye Movements
Summary: Stephen Lisberger studies the brain mechanisms that transform the motion of objects in the world, or our own motion, into accurate eye movements.
We move our eyes to facilitate excellent vision. The propensity of things to move poses a major threat to excellent vision, because it tends to cause images to move across the retina, degrading visual acuity. One source of image motion comes from the motion of objects through the world. Smooth pursuit eye movements allow us to track the visual stimuli that are created when objects move. A second source of image motion is created by our own head turns, which could drag our eyes along and thus cause images to smear across the retina. The vestibulo-ocular reflex keeps the position of gaze stable in space by generating eye movements that compensate for head motion. Our laboratory studies how these two kinds of eye movement are generated by the intact brain in behaving primates.
Smooth Pursuit Eye Movements
The basic anatomical circuits for pursuit eye movements are known. Visual inputs arise in the primary visual cortex. They are transmitted through the middle temporal visual area (MT) to the parietal and frontal cortex and, in parallel, from these three cortical areas to the brain stem and cerebellum for the assembly of motor commands. We are asking how the sensory signals created by moving visual stimuli are processed and transformed to create neural signals suitable to guide accurate smooth pursuit eye movements.
To analyze how moving images are processed in the visual parts of the brain, we have asked how signals related to target speed are transformed as they are transmitted from the extrastriate motion area MT to the pursuit system. As others have reported, the firing rate of MT neurons is tuned for the speed of moving stimuli: they give the strongest response for a given speed and weaker responses for either faster or slower speeds.
We have been analyzing the variation in neural responses and behavior as a way to understand how visual signals in MT are converted into commands for smooth pursuit eye movements. Research supported by the National Eye Institute has used Shannon's information theory to determine the time course over which it would be possible to extract information about target direction from the discharge of an MT neuron. "Information" has a formal mathematical sense in our analysis, but it conforms with the intuitive sense that the response of a neuron tells us more about the specifics of the stimulus if its discharge conveys more "information." We found that maximal information can be extracted from the first 100 ms of MT neural responses, but that individual MT neurons can discriminate different directions poorly. In contrast, analysis of the variation in pursuit eye movements has revealed that they are remarkably precise. Within the first 100 ms of the response, pursuit can discriminate directional differences as small as 3–5 degrees and speed differences of less than 20 percent of target speed. Thus, the visual system provides information about direction of target motion on the time scale that it is needed for execution of accurate motor behavior, but on a directional (and speed) scale that must be refined by pooling across multiple MT neurons.
Although pursuit is more precise about speed and direction than is any individual visual neuron, it still is imprecise to a degree. We are asking about the source of the variation. Analysis of the eye movements themselves suggested that the source of the variation is sensory, leading us to the hypothesis that the motor part of the pursuit circuit receives somewhat variable estimates of the direction and speed of a target, and tracks each incorrect estimate perfectly. We have tested this by recording from Purkinje cells in a region of the cerebellum that is involved in generating pursuit, and asking whether there is movement-by-movement covariation in the firing of the Purkinje cells and the evoked behavior. If there is, then pursuit variation arises before the site of the recording; if there is not, variation arises downstream. We find that the correlation is remarkably strong: variation in eye movement accounts for more than 35 percent of the variation in neural firing. We conclude that variation in pursuit arises upstream from the cerebellum. All Purkinje cells are receiving the same input from the sensory system and therefore are highly correlated with each other and the behavior they drive.
Because the neural circuit for smooth pursuit includes many areas of the cerebral cortex as well as the cerebellum, pursuit provides an excellent system for understanding how we learn motor skills. Two recent experiments have revealed key links in the neural process of learning. First, we have found the microstimulation in visual area MT can substitute for changes in the direction of target motion and cause precisely timed learning that mimics the learning induced by real visual stimuli. Second, neural recording and stimulation at different levels of the circuit have revealed that the site of learning is downstream from the "motor cortex" for pursuit, called the frontal pursuit area (FPA), but in or upstream from the floccular complex of the cerebellum.
Vestibulo-ocular Reflex
In primates, the performance of the vestibulo-ocular reflex (VOR) is normally nearly ideal. During rotatory head turns in darkness, the compensatory eye movements of the VOR are opposite in direction and nearly equal in amplitude to head turns. The VOR attains and maintains this excellent performance with a learning mechanism that is intact throughout life. In the laboratory, we study learning by fitting monkeys with goggles that either double or quarter the size of their visual inputs. When, for example, a monkey is wearing the doubling spectacles, visual scenes will be stabilized only if head turns in one direction are accompanied by eye rotations in the opposite direction at twice the usual speed. Initially, the VOR retains its normal response amplitude. Over several days, however, the persistent association of image motion and head turns causes the size of the VOR to increase gradually. At the end of the learning, head turns in darkness evoke an eye movement of twice the normal size.
We are exploring a number of hypotheses suggested by our earlier finding that learning occurs at specific sites in the cerebellar cortex and the vestibular nucleus. We are studying both the VOR and the responses of neurons in the VOR pathways over a much wider range of frequencies than before. We have found that learning induced during a monkey's free head turns has an effect on the responses of the VOR in the dark for sinusoidal head motions at frequencies up to 25 Hz, but not for higher frequencies. Recordings from vestibular afferents and extraocular motor neurons have provided precise predictions about the transformations done in the neural pathways for the VOR. We can model the eye movements across the full frequency range if we assume that the VOR is mediated by separate modifiable and unmodified pathways that interpose time delays of 9 and 2 ms, respectively. We are exploring the neural correlate of this observation by recording from neurons in the brain stem and cerebellum during vestibular stimuli inside and outside the range of frequencies where learning is expressed.
Last updated: September 4, 2007
|
|
|
|
