1. Monkeys were trained to grasp a rod movable in a horizontal arc (Fig. 1), and to hold the rod by angulation of the wrist in each of three positions (A,B, C). A maintained load was placed on the rod alternately to oppose flexion and extension. At a light signal, the monkey had to move to the next position in a prescribed sequence (ABCBABCBA, ETC.). The task was designed to dissociate, while holding in position, the following variables: 1) pattern of muscular activity in the forearm required to hold the wrist in position, determined by the direction of the load (flexor or extensor muscles); 2) position of the rod, and thus angulation of the wrist joint (A, B, and C); and 3) set for the direction of the intended next movement (flexor or extensor). These variables are subsequently referred to as MPAT, JPOS, and DSET, respectively. 2. After training, recordings were made of the EMG activity of muscles used in the task and of the discharge of single neurons in the motor cortex of the cerebrum and the interposed and dentate nuclei of the cerebellum. 3. While holding the wrist in position, EMG and interpositus behaved uniformly, with higher discharge frequency under load in one direction and lower discharge frequency under load in the opposite direction. This relation was relatively independent of the position held and of the direction of the intended next movement. Thus, interpositus and EMG both seemed best related to the MPAT variable, as opposed to JPOS and DSET variables. By contrast, neurons in motor cortex and in dentate fell into three categories: one category discharged in relation to the pattern of muscular activity (MPAT), a second to the position of the wrist (JPOS), and a third to the direction of the intended next movement (DSET). While MPAT neurons formed a distinct dissociated group, neurons that were best related to JPOS were often related to DSET, and vice versa. 4. A few of the MPAT neurons in interpositus and motor cortex were further studied by varying the magnitude (as well as the direction) of the loads. Both interpositus and motor cortex MPAT neurons changed firing frequency in relation to the magnitude of load, and though few neurons were thus studied, the relation seemed clearer for interpositus than for motor cortex. 5. Anatomically, the three types of neurons thus classified by firing pattern during the hold periods were intermixed in the arm area of motor cortex. In dentate and interpositus, those neurons thus related to the performance were localized to a narrow strip across the posterior part of both nuclei. Neurons apparently related to eye and drinking movements were located more posteriorly still, suggesting somatotopic representation.