To correctly position the hand with respect to the spatial location and orientation of an object to be grasped, visual information about the target and proprioceptive information from the hand must be compared. Multisensory integration models suggest that comparing visual and proprioceptive information, that are inherently encoded in retinal and joint reference frames respectively, requires visuo-proprioceptive transformations: that is the re-encoding of proprioceptive signals in a visual reference, and of the visual signals in a proprioceptive reference. Previous studies have shown that head tilt interfere with the visuo-proprioceptive transformations. It is unclear however, whether this phenomenon is related to the neck flexion or to the head-gravity misalignment. To answer to this question, we performed a behavioral virtual reality experiment in which we compared a grasping-like movement with lateral neck flexions executed in an upright seated position and while lying supine. The task requires visuo-proprioceptive transformations, because the target information is visually acquired, and the hand is sensed through proprioception only. The results show that lateral neck flexions have considerably different effects in the seated and supine posture. More precisely, the subjects’ response variability and the importance associated to the visual encoding of the information significantly increased when supine. These results, together to those of control experiments, are consistent with the idea that head-gravity misalignment interferes with the visuo-proprioceptive transformations and support the theorized central role of gravity in spatial perception.