It is known that proprioceptive signals from muscles, joints, and skin are involved in the execution of aimed arm movements, but their role in the acquisition of new motor behaviour is largely unknown. Previous research using deafferented patients yielded inconsistent findings: sensorimotor adaptation was found to be less, equal, or even better than in controls. The present study uses a different approach: we degraded proprioceptive signals in healthy subjects by wrist vibration, while leaving them intact in a control group. In order to induce sensorimotor adaptation, both groups executed pointing movements under a visual (Exp. A) or a mechanical perturbation (Exp. B). We found that adaptation to the visual distortion was little affected by vibration, while adaptation to the mechanical distortion was substantially impaired. We conclude that proprioceptive signals which overtly disagree with visual ones (Exp. A) are not used for adaptation, while those which largely agree with visual ones (Exp. B) are used to enhance adaptive recalibration. These results indicate that intact proprioception is needed for mechanical but not for visual adaptation, which implies that the underlying mechanisms are at least partly distinct.