It may be hard to fathom, but our eyes make 3-5 fixations every second, connected by short, rapid eye movements known as saccades. As a result of this motion, the target object changes retinal position even though it may remain in the same spatial position. As such, the image of the world that falls on our retina is constantly changing, but we perceive the world as stationary. It is important for us to be able to detect the moving objects in space and predict and judge their position and trajectory, for instance, a moving car. Here, we study the processing coordinate system of visual motion, to see if it is coded for in retinal (relative to the eye) or spatial (relative to the world) coordinates.
To do this, we used human psychophysics, which is a method used for quantitatively measuring performance of the perceptual system using stimuli that are systematically varied. We measured the ability of nine human subjects to discriminate the direction of two moving-dot stimuli that were separated temporally. In each trial, the subjects were required to make a leftward or rightward saccade after the presentation of the first stimulus. In doing so, the second stimulus could appear in the same spatial or retinal position as the first. Using this paradigm, we found that subjects were better at detecting the differences in direction when the two stimuli were placed in the same spatial location than when the two stimuli were placed in the same retinal location. This suggests that memory for motion is likely processed and stored in a region that is spatiotopic or that automatically accounts for saccades. Using a similar task, in which a saccade was not executed, we showed that as the two stimuli to be compared were spatially displaced from each other, the distance at which performance deteriorated increased at increasing eccentricities, in a fashion which best matched the receptive field sizes of area MT. From this first set of experiments, we conclude that area MT plays an important role in the memory for motion process, and that this is carried out in spatiotopic coordinates.
As it is well established that information about visual motion is processed in area MT/V5 in both monkeys and humans, the fact that this area contributes in a memory for motion task is not unexpected. What is surprising is that it appeared to do so in spatial coordinates; MT has been traditionally thought to be a retinotopic area. However, there has been recent debate as to whether it can processes information in spatial coordinates. In the next two studies, we look closely at how area MT may be able to perform spatiotopic processing by recording extracellularly from individual neurons in three animals.
The first of these experiments had the animals performing a simple saccade task in which a spatially stable moving dot stimulus was presented in either the pre-saccadic or post-saccadic receptive field of the neuron being recorded from. The MT neurons we recorded from responded as if their receptive fields were purely retinotopic, and did not respond to stimuli presented in the pre-saccadic receptive field after the saccade and did not respond to stimulus presented in the post-saccadic receptive field before the saccade. Furthermore, when a stimulus was presented in the post-saccadic receptive field, the neurons did not start responding to it until well after the end of the saccade, as if it had just appeared on the screen. This means that the receptive field of the neurons do not shift to their post-saccadic location before or during the saccade, a process termed anticipatory remapping, but remain linked to the retina.
There is a possibility that spatial processing could come from neural remapping within MT in a more subtle fashion that does not require the receptive fields to shift ahead of time. To look at this, we performed another set of experiments where a transiently presented stimulus is present in the future receptive field of the neuron, but is no longer present when the eye movement is made. If visual responses to the stimuli in the post-saccadic receptive field are transferred from one neuron to another at the same level of processing (within MT), we would see a remapped trace of the visual signal after the eye reaches the new fixation position. Using a black disc presented on a white background for 50ms, we find that MT neurons do not remap responses to the stimulus when it is no longer present. Hence we conclude that MT neurons do not appear to have any form of remapping.
In conclusion, we have shown that motion perception involves some spatial processing, however area MT, the traditional motion processing area, has receptive fields which work strictly in retinal coordinates. These receptive fields do not shift preemptively to their post-saccadic location prior to the end of the saccade and the neurons in area MT do not appear to pass visual information to each other at the end of the saccade. We propose the presence of an automatic process which updates motion information to spatial coordinates, possibly through the transfer of information to an area which is spatiotopic or which has strong remapping activity, such as LIP or FEF. This would account for the optimal performance we see in the first experiment in spatiotopic locations, while also providing an explanation for the lack of spatial processing we see in area MT.
