We have conducted an experiment to study the effect of an occluding surface on the accuracy of near field depth matching in augmented reality (AR). Our experiment was based on replicating a similar experiment conducted by Edwards et al. [2]. We used an AR haploscope [1], which allows us to independently manipulate accommodative demand and vergence angle of the visible image. Fifteen observers matched the perceived depth of an AR-presented virtual object with a physical pointer. Overall, observers overestimated depth by 5 mm or less in the presence of the occluder, while in the absence of an occluder they overestimated depth by 5 to 10 mm. The data from Edwards et al. [2] is normalized, and when we performed the same normalization procedure on our own data, our results do not agree with Edwards et al. [2]. We suspect that eye vergence explains these results.
