We propose Specimen Box, an interaction technique that allows world-fixed display (such as CAVEs) users to naturally hold a plausible physical object while manipulating virtual content inside it. This virtual content is rendered based on the tracked position of the box. Specimen Box provides the weight and tactile feel of an actual object and does not occlude rendered objects in the scene. The end result is that the user sees the virtual content as if it exists inside the clear physical box. We conducted a user study which involved a cognitively loaded inspection task requiring extensive manipulation of the box. We compared Specimen Box to Grab-and-Twirl, a naturalistic bimanual manipulation technique that closely mimics the mechanics of our proposed technique. Results show that performance was significantly faster with Specimen Box. Further, performance of the control technique was positively affected by experience with Specimen Box.