A key feature of augmented reality (AR) is the ability to display virtual content that appears stationary as users move throughout the physical world (‘world-locked rendering’). Imperfect world-locked rendering gives rise to perceptual artifacts that can negatively impact user experience. One example is random variation in the position of virtual objects that are intended to be stationary (jitter’). The human visual system is highly attuned to detect moving objects, and moreover it can disambiguate between the retinal velocities that arise from object motion and self-motion, respectively. In this study, we investigated how the perceptibility of AR object jitter varies as a function of user self-motion. Using a commercially available AR HMD to display a 3D textured cube, we measured sensitivity to added jitter versus a no-jitter reference using a two-interval forced choice task. Three user motion conditions (stationary, head rotation, and walking) and three object placement conditions (floating in free space, on a desk, and against a wall) were tested in a full factorial design. We hypothesized that (1) as users move their head and eyes during self-motion, their sensitivity to jitter will decrease, due to added retinal velocity; and (2) rendering virtual objects near physical surfaces will increase sensitivity to jitter, by providing proximal veridical visual cues. Psychometric thresholds indicated that users were significantly less sensitive to jitter during self-motion than when they were stationary, consistent with hypothesis (1). Users were also more sensitive to jitter in one of the two object placement conditions, providing partial support for hypothesis (2). To generalize beyond distinct user motion and object placement conditions, we also analyzed eye tracking data. The amount of retinal slip (i.e. how much gaze drifted across the virtual object) predicted jitter thresholds better than recorded head movements alone, suggesting a retinally-driven decrease in jitter sensitivity during self-motion. These results can be used to inform requirements for AR world-locked rendering systems, as well as how these may be updated dynamically using online measurement of user head and eye movements.