The detailed characteristics of perception/action coupling is studied using a sensori-motor pointing task. Using a graphical pen, subjects (n=6) had to point to the final location of the invisible center of simple geometrical shapes - cross, diamond, chevron -, after their movement along a circular - clockwise/anti-clockwise - trajectory ended. The target shapes could be fully visible, thus yielding a highly coherent motion percept, or presented behind vertical rectangular masks. In these latter conditions, perceived global coherence was dependent upon the visibility of the masks. Under these conditions, constant and variable errors and the spatial distribution of pointing responses indicate that: (1) Accuracy of pointing responses is better at high than at low motion coherence. (2) With fully coherent shapes, pointing accuracy is similar for a cross and a single spot - i.e. baseline condition - and worse for the diamond and chevron for which the profiles of the spatial distribution of pointing responses are different. In addition, pointing responses are biased in the direction of motion - representational momentum -, an effect which disappears at middle and low coherence. (3) At low coherence, the location of the target center is overestimated and many pointing errors occur. Overall, observers appear to adapt their motor strategies to the specific context - i.e. shape and coherence - within which they have to perform their action. These results, showing comparable, although slightly different, biases for perception and action, are discussed in the light of the proposed dichotomy of dedicated functional processes through the ventral and dorsal pathways.