The acquisition of neurophysiological data during awake, behaving animal experiments typically involves experimental sessions lasting several days to weeks. Therefore, it is important to understand natural fluctuations in behavioral performance over such periods. Here we quantified patterns of movement variability for reaches performed by two monkeys across five daily experimental sessions. The monkeys were trained to move in an immersive virtual reality (VR) environment that was designed to resemble the experimental room. Visual feedback of the limb was provided using VR avatar arms that were controlled through a reflective marker-based motion capture system. Additionally, tactile cues were provided in the form of physical reach targets. Spatial variability was characterized at early (peak acceleration) and late (movement endpoint) kinematic landmarks. We found that the magnitude of variability was generally larger at peak acceleration than at the endpoint but was relatively consistent across days and within animals. The spatial characteristics of variability were also generally highly consistent at peak acceleration both within and between animals but were noticeably less so at the endpoint. The results highlight the benefits of using early kinematic landmarks such as peak acceleration for quantifying movement variability in reaching studies involving animals.