We studied pointing movements to remembered visual targets in a completely darkened room with and without self-made step movements in order to investigate in which coordinate system and to what extent target representations relative to the body are updated for self-induced egomotion. A small red-light-emitting diode on the fingertip provided visual feedback about fingertip position at all times. We asked subjects to make pointing movements that started 2 s after disappearance of a visual target. In this interval of 2 s the subject did or did not make a step. The pointing errors without a step showed that subjects undershot faraway targets in a systematic way, whereas they sometimes overshot nearby targets. We found that the step causes larger pointing errors both in amplitude and direction with a bias in the direction of the step. We explored three different versions of a descriptive model in which polar coordinates were used to describe the pointing movement, and in which either Cartesian or polar coordinates were used to update target position relative to the shoulder for the step. The results suggest that incorporation of the step displacement in the new target position relative to the subject is done in a Cartesian frame of reference. Moreover, the amplitude of the step displacement tends to be underestimated by subjects.