Spatialized views use visuo-spatial metaphors to facilitate sense-making from complex non-spatial databases. Spatialization typically includes the projection of a high-dimensional (non-spatial) data space onto a lower dimensional display space for visual data exploration. In comparison to 2D spatialized displays, 3D displays could potentially convey more information, as they employ all three available spatial display dimensions. In this study, we evaluate if this advantage exists and whether it outweighs the added cognitive, perceptual, and technological costs of 3D displays. In a controlled human-subjects experiment, we investigated how viewers identify document similarity in 3D network spatializations that depict news articles as points connected by links. Our quantitative findings suggest that similarity ratings for 3D network displays are similar to those obtained in a prior 2D study we conducted. With both types of displays, viewers mostly judged document similarity on the basis of metric distances along network links, as opposed to node counts or distance across the network links. However, node counts do affect similarity assessments with 3D displays more than with 2D displays. We also find no significant differences in similarity judgments whether 3D displays are presented monoscopically or stereoscopically. We conclude that any advantage of 3D displays in conveying more information than 2D displays does not necessarily outweigh their additional demands on cognitive, perceptual, and technological resources.