We present a controlled experiment assessing how accurately a user can interpret the video feed of a remote user showing a shared object on a large wall-sized display by looking at it or by looking and pointing at it. We analyze distance and angle errors and how sensitive they are to the relative position between the remote viewer and the video feed. We show that users can accurately determine the target, that eye gaze alone is more accurate than when combined with the hand, and that the relative position between the viewer and the video feed has little effect on accuracy. These findings can inform the design of future telepresence systems for wall-sized displays.