An experiment was conducted to evaluate the potential for improved operator performance in a space-based telerobotic manipulation task when the operator's control interface was based around an egocentric rather than exocentric frame of reference (FOR). Participants performed three tasks of increasing difficulty using a virtual reality-based simulation of the Space Shuttle Remote Manipulation System (SRMS) under four different control interface conditions, which varied in respect of two factors, virtual viewpoint FOR (fixed versus attached to arm) and hand controller FOR (end-effector referenced versus world referenced.) Results indicated a high degree of interaction between spatial properties of the task and the optimal interface condition. Across all tasks, the conditions under end-effector-referenced control were associated with higher performance, as measured by rate of task completion. The mobile viewpoint conditions were generally associated with lower performance on task completion rate but improved performance with respect to number of collisions between the arm and objects in the environment. Increased head movement and higher number of errors in arm motion indicated that the mobile viewpoint suffered from confounding uncontrolled keyhole effects. No correlation between performance and prior 3D simulation experience was observed. There was a significant effect of gender on performance in line with results from the field. The requirement for telemanipulation interfaces to represent critical kinematic limitations in the interface emerges in discussion of origins of performance differences between conditions. The results provide support for the partial application of an egocentric telepresence control interface to space-based articulated manipulators. Different factorings of ego- and exocentric FORs in order to alleviate poor performance under the mobile viewpoint are discussed along with implications for other space-based telemanipulation applications and fruitful approaches to further studies.