Two deafferented patients and several control subjects participated in a series of experiments to investigate how accurate single-joint movements are programmed, spatially calibrated, and updated in the absence of proprioceptive information. The deafferented patients suffered from a permanent and severe loss of large sensory myelinated fibers below the neck. Subjects performed, with and without vision, sequences of forearm supinations and pronations with two temporal delays between each movement (0 s and 8 s). Overall, the lack of proprioception did not yield any significant decrease in movement accuracy when vision was available. Without vision, the absence of proprioceptive afferents yielded (1) significantly larger spatial errors, (2) amplitude errors similar to those of control subjects, and (3) a significant drift when an 8-s delay was introduced between two successive movements. Subjects also performed, without vision, a 20 degrees supination followed by a 20 degrees pronation that brought back the wrist to the starting position. On some trials, the supination was blocked unexpectedly by way of a magnetic brake. When the supination was blocked, subjects were already on the second target and no pronation was required when the brake was released. The deafferented patients, unaware of the procedure, always produced a 20 degrees pronation. These data confirm that deafferented patients were not coding a final position. It rather suggests that they coded an amplitude and translated the spatial distance between the two targets in a corresponding force pulse. Overall, the results highlight the powerful and key role of proprioceptive afferents for calibrating the spatial motor frame of reference.