Virtually exaggerating head motions when using head-coupled viewing may be beneficial. This is especially important in environments where the user is sitting, such as in desktop or "fish-tank" VR. We present an evaluation of exaggerated head-coupled perspective for 3D assembly tasks. Three levels of exaggeration were compared to determine if exaggeration was beneficial compared to a direct mapping. The results suggest that there is some user preference towards modest levels of exaggeration. However, no significant differences by the experimental conditions were detected, other than a learning effect.