The purpose of these experiments was to investigate whether visual perceptions of the earth-fixed vertical axis are more accurate than those of intrinsic body-fixed axes. In one experiment, nine neurologically normal young adult subjects' abilities to position a luminescent rod vertically or parallel to the longitudinal axis of the head or trunk were studied in four conditions: (1) earth-fixed--subjects stood erect with the head aligned to the trunk and visually aligned a hand-held rod to vertical; (2) earth--subjects aligned the rod to vertical as in 1, but the orientations of the head and trunk were varied in the sagittal and frontal planes on each trial; (3) head--frontal and/or sagittal plane orientation of the subject's head was varied on each trial and the rod was aligned parallel to the longitudinal axis of the head; (4) trunk--frontal and/or sagittal plane orientation of the subject's trunk was varied on each trial and the rod was aligned parallel to the longitudinal axis of the trunk. Note that in conditions 2, 3, and 4 the head and trunk were never aligned with each other. Also, each condition was carried out in normal light and in complete darkness. Perceptual errors were measured in both the frontal and the sagittal planes. The results showed that the variable errors were significantly lower when subjects aligned the rod to vertical rather than to the longitudinal axis of the head or trunk. Also, errors were similar in size in the two planes and were unaffected by vision of the surrounding environment. In a second experiment, subjects were seated and controlled the position of a luminescent rod held by a robot. They aligned the rod either to the longitudinal axis of their head or to the vertical in complete darkness, under three conditions similar to those described above: (1) earth-fixed, (2) earth, and (3) head. There was no possibility of use of kinesthetic information for controlling rod position in this experiment as in the first experiment. The results were similar to those of the first experiment, as subjects aligned the rod more accurately to vertical than to the longitudinal axis of the head. These results show convincingly that visual perceptions of earth-fixed vertical are more accurate than perceptions of intrinsic axes fixed to the head or trunk.