|
|
Although seeing is an end in itself, it is also a source of information for effective behavior. To successfully catch a ball or to skewer dinner on a fork we must accurately locate the target and produce an appropriate movement to its location.
Locating visual targets would be simpler if our eyes did not move: An image falling on the center of the retina (the light-sensitive structure lining the back of the eye) would indicate an object lying straight ahead; one falling to the left of center, an object lying to the right (the eye's lens reverses left and right); and so on. However, to provide clear views of stationary and moving objects, our eyes move frequently and rapidly over a large field. Thus, after a rightward eye movement, an object to the right might produce a retinal image right (instead of left) of center. Since at least the mid-nineteenth century it has been appreciated that the brain must take eye position into account, must remap the retinal image, to get the location of visual objects correct as the eyes move. Saccadic eye movements, the rapid flicks made several times a second when scanning a visual scene, place the greatest demands on the remapping mechanism. Studying visual localization around the time of saccades should reveal the capabilities, and something about the nature, of the mechanism.
We have briefly flashed small lights at various locations in front of human volunteers seated in a dark room (see Figure 1A). By monitoring the subject's eye movements, we were able to time each flash so that it would occur before, during or after a saccade; also, we could choose a light whose image would fall in a desired retinal locus. Subjects simply made saccades and then pointed with the index finger to the apparent location of each flash.
We found that the remapping mechanism was not up to the demand: A saccadic eye movement might be complete in 50 msec (see Figure 1B), when the remapping mechanism was still struggling to catch up: Remapping took about three times as long as the eye movement itself. This means that behavior based on visual localization will be in error--you may miss the ball or skewer your dinner companion--during and immediately following saccades.
 |
 |
|
Figure 1A: Subject is shown on a dental impression bite and wearing an eye position monitor, completing a trial with a positioning response. Before each trial begun, central speakers created a virtual sound source at the position of the central probe light, to help center gaze. The subject's right hand held the ready button, with which he initiated the trial. Either left or right speakers then cued his saccade. The left index finger carried a copper thimble, which, in connection with touch strips, formed a potentiometer that recorded the horizontal location of the pointing finger. |
Figure 1B: The abscissa is time in msec after beginning-of-sccade; the ordinate is position in space, with zero straight ahead of the subject's measured eye, minus to his left, and plus to his right. The dotted straight vertical line gives mean time of end-of-saccade. Each dot in the "cloud" gives the position of one pointing response and the presentation time of the probe flash the subject was attempting to localize. The heavy solid curve is a four parameter fit to these localizations. The dashed horizontal line gives the mean position of pointing responses to probe flashes long after end-of-saccade. Large leftward saccades had mean duration 64 msec, and mean magnitude 0.75 degrees right of its actual position on the line of sight. Eight msec after a saccade began, localization began to shift, with time constant =126 msec. Probes flashed long after saccades were localized 0.3 degrees right of actual position. |
These results suggest that the remapping mechanism is normally supplemented by other mechanisms, the leading candidates involving reference to other visual objects (the relative locations of retinal images of visual objects are little affected by eye movement).
We also found that the brain makes the best of its limited abilities. Saccades tend to be made to objects of interest and characteristically fall short of targets. Thus, when a saccade is made, the brain can predict the region of the retina in which the target's image will fall after the saccade (e.g., the left half of the retina following a rightward saccade). We found that this region of the retina is remapped faster, at the expense of other regions unlikely to be needed immediately.
Once we know where something is, we tend to think that reaching for it, looking at it, and saying where it is are but different expressions of our knowledge. Evolutionary considerations suggest, instead, that sensory-motor systems are more like a collection of somewhat independent mechanisms with differing information sources. To evaluate this position, we have begun experiments in which subjects use different motor systems to localize flashed targets.
