It has been hypothesized that the end-point position of reaching may be spe
cified in an egocentric frame of reference. In most previous studies, howev
er, reaching was toward a memorized target, rather than an actual target. T
hus, the role played by sensorimotor transformation could not be disassocia
ted from the role played by storage in short-term memory. In the present st
udy the direct process of sensorimotor transformation was investigated in r
eaching toward continuously visible targets that need not be stored in memo
ry. A virtual reality system was used to present visual targets in differen
t three-dimensional (3D) locations in two different tasks, one with visual
feedback of the hand and arm position (Seen Hand) and the other without suc
h feedback (Unseen Hand). In the Seen Hand task, the axes of maximum variab
ility and of maximum contraction converge toward the mid-point between the
eyes. In the Unseen Hand task only the maximum contraction correlates with
the sightline and the axes of maximum variability are not viewer-centered b
ut rotate anti-clockwise around the body and the effector arm during the mo
ve from the right to the left workspace. The bulk of findings from these an
d pre vious experiments support the hypothesis of a two-stage process, with
a gradual transformation from viewer-centered to body-centered and arm-cen
tered coordinates. Retinal, extra-retinal and arm-related signals appear to
be progressively combined in superior and inferior parietal areas, giving
rise to egocentric representations of the end-point position of reaching.