We examined neuronal activity in three motor cortical areas while a rhesus monkey adapted to novel visuomotor transforms. The monkey moved a joystick that controlled a cursor on a video screen. Each trial began with the joystick centered. Next, the cursor appeared in one of eight positions, arranged in a circle around a target stimulus at the center of the screen. To receive reinforcement, the monkey moved the joystick so that the cursor contacted the target continuously for Is. The video monitor provided continuous visual feedback of both cursor and target position. With those elements of the task constant, we modified the transform between joystick movement and that of the cursor at the beginning of a block of trials. Neuronal activity was studied as the monkey adapted to these novel joystick-cursor transforms. Some novel tasks included spatial transforms such as single-axis inversions, asymmetric double-axis inversions and angular deviations (also known as rotations). Other tasks involved changes in the spatiotemporal pattern and magnitude of joystick movement. As the monkey adapted to various visuomotor tasks, 209 task-related neurons (selected for stable background activity) showed significant changes in their task-related activity: 88 neurons in the primary motor cortex (M1), 32 in the supplementary motor cortex (M2), and 89 in the caudal part of the dorsal premotor cortex (PMdc). Slightly more than half of the sample in each area showed significant changes in the magnitude of activity modulation during adaptation, with the number of increases approximately equaling the number of decreases. These data support the prediction that changes in task-related neuronal activity can be observed in M1 during motor adaptation, but fail to support the hypothesis that M1 and PMdc differ in this regard. When viewed in population averages, motor cortex continued to change its activity for at least dozens of trials after performance reached a plateau. This slow, apparently continuing change in the pattern and magnitude of task-related activity may reflect the initial phases of consolidating the motor memory for preparing and executing visuomotor skills.

A fully recurrent neural network model was optimized to perform a spatial delayed matching-to-sample task (DMS). In DMS, a stimulus is presented at a sample location, and a match is reported when a subsequent stimulus appears at that location. Stimuli elsewhere are ignored. Computationally, a DMS system could consist of memory and comparison components. The model, although not constrained to do so, worked by using two corresponding classes of neurons in the hidden layer: storage and comparator units. Storage units form a dynamical system with one fixed point attractor for each sample location. Comparator units constitute a system receiving input from these storage units as well as from current input stimuli. Both unit types were tuned directionally. These two sources of information combine to create unique patterns of activity that determine whether a match has occurred. In networks with abundant hidden units, the storage and comparator functions were distributed so that individual units took part in both. We compared the model with single-neuron recordings from premotor (PM) and prefrontal (PF) cortex. As shown previously, many PM and PF neurons behaved like storage units. In addition, both regions contain neurons that behave like the comparator units of the model and appear to have dual functionality similar to that observed in the model units. No neuron in either area had properties identical to those of the match output neuron of the model. However, four PF neurons and one PM neuron resembled the output signal more closely than any of the hidden units of the model.

When a small, focally attended visual stimulus and a larger background frame shift location at the same time, the frame's new location can affect spatial perception. For horizontal displacements on the order of 1--2 degrees, when the frame moves more than the attended stimulus, human subjects may perceive that the attended stimulus has shifted to the right or left when it has not done so. However, that misapprehension does not disable accurate eye movements to the same stimulus. We trained a rhesus monkey to report the direction that an attended stimulus had shifted by making an eye movement to one of the two report targets. Then, using conditions that induce displacement illusions in human subjects, we tested the hypothesis that neuronal activity in the prefrontal cortex (PF) would reflect the displacement directions reported by the monkey, even when they conflicted with the actual displacement, if any, of the attended stimulus. We also predicted that these cells would have directional selectivity for movements used to make those reports, but not for similar eye movements made to fixate the attended stimulus. A population of PF neurons showed the predicted properties, which could not be accounted for on the basis of either eye-movement or frame-shift parameters. This activity, termed report-related, began approximately 150 ms before the onset of the reporting saccade. Another population of PF neurons showed greater directional selectivity for saccadic eye movements made to fixate the attended stimulus than for similar saccades made to report its displacement. In view of the evidence that PF functions to integrate inputs and actions occurring at different times and places, the present findings support the idea that such integration involves movements to acquire response targets, directly, as well as actions guided by less direct response rules, such as perceptual reports.