Joint position sense (JPS) is commonly evaluated using an angle replication protocol with vision occluded. However, multiple sources of sensory information are integrated when moving limbs accurately, not just proprioception. The purpose of this study was to examine different availability of vision during an active JPS protocol at the shoulder. Specifically, the effects of four conditions of vision availability were examined for three target shoulder elevation angles (50°, 70° & 90°): vision occluded continuously (P-P); vision available continuously (VP-VP); vision occluded only during target memorization (P-VP); and vision occluded only during target position replication (VP-P). There were 18 participants ( M age = 21, SD = 1 years). We used separate repeated ANOVAs to examine the effect of condition and target angle on participants’ absolute error (AE, a measure of accuracy) and constant error (CE, a measure of directional bias). We found a significant main effect for condition and angle for both dependent variables ( p < 0.01), and follow-up analysis indicated that participants were most accurate in the VP-VP condition and least accurate in the P-VP condition. Further follow-up analysis showed that accuracy improved with higher target elevation angles, consistent with previous research findings. Constant error results were similar, as there was a prominent tendency to overshoot the target. Unsurprisingly, participants performed best at the angle replication protocol with their eyes open. However, while accuracy was reduced when vision was occluded during target memorization, it was restored during target replication. This finding may have indicated an accuracy cost due to introduced noise when transforming sensory information from a proprioceptive reference frame into a visual reference frame.