This paper explores the use of a guessability study to examine child-defined gestures with Kinect. Applying a Wizard-of-Oz approach, gestures were elicited from six children (age 3-8) through a series of 22 task stimuli including object manipulation, navigation-based tasks, and spatial interaction. Gestures were video recorded, transcribed, and coded by three researchers employing an inductive, qualitative method of analysis. Five themes emerged from the data: (1) the influence of 2D touchscreens on children's interactions in 3D, (2) the role of contextual cues in designing a stimuli set, (3) individual preferences for dominant styles of interaction, (4) different approaches children employ to simulate the same object path, and (5) and allocentric versus egocentric approaches for manipulating objects on screen. While we did not achieve strong consensus among all of the gestures produced by children in our study, our results provide a basis for further refinement of the stimulus set and methodology used for future work examining child-defined gestures for whole-body interfaces.