Multi-object acquisition is indispensable for many VR applications. Commonly, users select a group of objects of interest to perform further transformation or analysis. In this paper, we present three multi-object selection techniques that were derived based on a two-dimensional design space. The primary design dimension concerns whether a technique acquires targets through point-based methods (selecting one object at a time) or volume-based methods (selecting a set of objects within a selection volume). The secondary design dimension examines the mechanisms of selection and deselection (cancel the selection of unwanted objects). We compared these techniques through a user study, emphasizing on scenarios with more randomly distributed objects. We discovered, for example, that the point-based technique was more efficient and robust than the volume-based techniques in environments where the targets did not follow a specific layout. We also found that users applied the deselection mechanism mostly for error correction. We provide an in-depth discussion of our findings and further distill design implications for future applications that leverage multi-object acquisition techniques in VR.