We explore verbalization as a means for quick-and-dirty 3D shape exploration in early design. Our work stems from the knowledge gap that the fundamental principles necessary to operationalize speech as a viable means for describing and communicating 3D digital forms do not currently exist. To address this gap, we present a case study on 3D scene modeling within the context of interior design. For this, we implemented a constrained workflow wherein a user can iteratively generate variations of seed templates of objects in a scene through verbal input. By using this workflow as an experimental setup, we systematically study four aspects of speech-based shape exploration, namely, (a) design-in-context (creating one shape with respect or in relation to the other), (b) order independence (sequence of parts preferred in speech-based shape exploration), (c) multiscale exploration (study how speech allows overview-then-detail modifications), and (d) semantic regions of interest (effectiveness of speech for modifying regions of a given object). We finally present an observational study with six participants selected from diverse backgrounds to better understand shape verbalization.