A major issue in motor control studies is to determine whether and how we use spatial frames of reference to organize our spatially oriented behaviors. In previous experiments we showed that simulated body tilt during off-axis rotation affected the performance in verbal localization and manual pointing tasks. It was hypothesized that the observed alterations were at least partly due to a change in the orientation of the egocentric frame of reference, which was indeed centered on the body but aligned with the gravitational vector. The present experiments were designed to test this hypothesis in a situation where no inertial constraints (except the usual gravitational one) exist and where the orientation of the body longitudinal z-axis was not aligned with the direction of the gravity. Eleven subjects were exposed to real static body tilt and were required to verbally localize (experiment 1) and to point as accurately as possible towards (experiment 2) memorized visual targets, in two conditions, Head-Free and Head-Fixed conditions. Results show that the performance was only affected by real body tilt in the localization task performed when the subject's head was tilted relative to the body. Thus, dissociation between gravity and body longitudinal z-axis alone is not responsible for localization nor for pointing errors. Therefore, the egocentric frame of reference seems independent from the orientation of the gravity with regard to body z-axis as expected from our previous studies. Moreover, the use of spatial referentials appears to be less mandatory than expected for pointing movements (motor task) than for localization task (cognitive task).