its decision by making a saccadic eye movement to the appropriate target.
When viewing these displays, human observers typically see weak, coherent motion flow superimposed upon a noisy substrate of twinkling dots. The discrimination can be made easy or difficult simply by increasing or decreasing the proportion of dots in coherent motion, a value that we refer to as the coherence of the motion signal. A range of coherences, chosen to span behavioral threshold, were used in our experiments, and all stimulus conditions were presented in random order.
This task offers substantial advantages for our purposes because the sensory and motor representations underlying performance are reasonably well known. The motion signals originate in large part from columns of directionally selective neurons in extrastriate visual areas MT and MST (20). This laboratory has shown that single neurons in MT and MST are remarkably sensitive to the motion signals in our displays, that inactivation of MT selectively impairs performance on this task, and that electrical stimulation of a column of directionally selective cells can bias a monkey's decisions toward the direction encoded by the stimulated column (21–27).
Motor signals that govern the monkeys' behavioral responses (saccadic eye movements) almost certainly pass through the superior colliculus (SC) and/or the frontal eye fields (FEFs). Both structures have long been known to play key roles in producing saccades (for reviews see refs.28 and 29). Both the SC and the FEFs contain neurons that discharge just prior to saccades to well-defined regions of the visual field, termed movement fields, and simultaneous lesions of these structures eliminate most saccades (30). Electrical stimulation of either the SC or the FEFs elicits a saccade to the movement field of the stimulated neurons.
In the context of this perceptual task, therefore, we are able to state our key experimental question in a much more focused manner: how do motion signals in MT and MST influence motor structures such as SC and FEFs so as to produce correct performance on the task?
To explore the link between sensation and action, we targeted for study a specific subset of neurons in the lateral intraparietal region (LIP) of the parietal lobe that carries high-level signals appropriate for planning saccadic eye movements. These high-level signals arise early in the initial stages of planning a saccade and are therefore likely to be linked to the decision process in a revealing manner (31–34). Anatomical data suggest that LIP is an important processing stage in the context of our task: LIP receives direct input from MT and MST and projects in turn to both FEFs and SC (5,35,36). High-level signals like those in LIP also exist in SC and FEFs, and our investigation must ultimately include all three structures (10,15,37–39). We chose to begin in LIP because of its proximity to MT and MST.
The neurons of particular interest to us have been characterized most incisively in a remembered saccade task. In this task, a saccade target appears transiently at some location in the peripheral visual field while the monkey maintains its gaze on a fixation point. The monkey must remember the location of the transiently flashed target during an ensuing delay period which can last up to several seconds. At the end of the delay period, indicated by disappearance of the fixation point, the monkey must saccade to the remembered location of the target. The neurons of interest begin firing in response to the appearance of the saccade target and maintain a steady level of discharge during the delay period until the saccade is made. These neurons are spatially selective in that the delay-period response occurs only before eye movements into the movement field. Thus the delay period activity forms a temporal “bridge” between sensory responses to the visual target and motor activity that drives the extraocular muscles at the time of the saccade.‡
‡ Different investigators have suggested that the delay period activity is related to memory of the target location, attention to a particular region of visual space, or an intention to move the eyes (10, 31, 32, 34, 37, 38). We believe that current data are insufficient to take a strong stand in favor of any of these interpretations. Despite the uncertainty, the information contained in delay-period activity is sufficient to guide the eyes to a spatial target.