Research providing evidence from patients and neurologically healthy participants has demonstrated that visual perception can dissociate from visually guided actions, and that this dissociation can be removed by reducing visual feedback to a monocular view, or by completely occluding vision. Previously we have demonstrated a similar dissociation between perception and action on a rod-bisection task. The current paper examines whether manipulating the viewing conditions can also affect this dissociation. Forty-eight right-handed participants bisected five rods of different lengths, by pointing to the centre and by picking each up by the centre, under three viewing conditions: binocular viewing, monocular viewing and occluded viewing. Binocular viewing resulted in the expected perception-action dissociation: pointing bisection errors were to the right of centre and grasping bisection errors showed no bias. However, this pattern was also evident for the monocular-viewing condition, demonstrating that monocular viewing had no significant effect on bisection. In contrast, complete visual occlusion led to the elimination of the perception-action dissociation, and, in addition, the direction of the pointing errors reversed: both pointing and grasping errors were to the left of centre. These results are compared with a line-bisection task performed under similar conditions. This task resulted in consistent biases for which reducing visual feedback only influenced the extent of the error. The direction of the error was influenced by line position. These results demonstrate that theories for differential processing in the ventral and dorsal streams, used to elucidate perception-action dissociations, may not be compatible with the rod-bisection task and that online visuomotor feedback may better explain the dissociation. © 2007 Elsevier Srl. All rights reserved.