The purpose of this experiment was to determine the preferred coordinate system for representation of hand orientation in 3-dimensional space. The ability of human subjects to perceive angles of the hand in 3-dimensional space (elevation, yaw, roll angles-extrinsic coordinate system) was compared to their ability to perceive hand angles relative to the proximal upper limb segments (wrist joint angles, forearm pronation-intrinsic coordinate system). With eyes closed, subjects performed a matching task in which the experimenter positioned the left arm, forearm and hand and the right arm and forearm. Subjects were then told to match an angle in one of the two coordinate systems by moving only the right hand at the wrist or the forearm as in pronation or roll matching. Absolute constant error (ACE), variable error (VE) and normalized variable error (NVE-normalized to tested range of motion) of matching were quantified for each subject for each of the six angles matched. It was hypothesized that matching angles in a preferred coordinate system would be associated with lower ACE, VE and NVE. Overall, ACE and VE were lower for matching hand angles in the intrinsic coordinate system. This suggests that the preferred coordinate system involved specification of hand angles relative to forearm and arm angles (joint angles) rather than the hand angles relative to axes external to the upper limb. However, matching of pronation angles was associated with larger VE and NVE than roll angle matching. There were no significant differences in ACE between pronation and roll matching. In a second experiment subjects with their forearms constrained at different elevations matched hand elevation and wrist flexion angles. Thus, errors in matching the angles in the non-preferred coordinate system were predictable if the subjects were biased toward matching angles in the preferred coordinate system. Trends in the data suggested that subjects preferred matching hand elevation angles but these trends were not consistent within or between subjects. Thus a preferred intrinsic coordinate system for wrist flexion matching was not observed in this experiment. We suggest that matching angles when proximal limb segments are constrained is a simpler task for the subjects (VE lower than in the first experiment) and may bias the matching toward the extrinsic coordinate system. Thus, hand orientation in 3-dimensional space may be perceived as follows: wrist flexion and abduction angles together with forearm elevation and yaw are used to specify hand elevation and yaw; these together with hand roll angle, completely specify the hand angle in 3-dimensional space.