The boundary between subject and object is becoming ever-the-more blurred by the creation of new types of computational objects. Especially when these objects take the form of robotic creatures do we get to question the powerful impact of the object on the person. Couple this with the expression of internal, unspoken experience through the making of non-speech sounds and we have a situation that demands new thoughts and new methodologies. This thesis works through these questions via the design and study of syngvab, a robotic marionette that moves in response to human non-speech vocal sounds. I draw from the world of puppetry and performing objects in the creation of syngvab the object and its stage, showing how this old tradition is directly relevant for the development of non-anthropomorphic, non-zoomorphic robotic creatures. I show how this mongrel of an object requires different methodologies of study, pulling from actor-network theory to examine syngvab in a symmetric manner with the human participants. The results of a case study interaction with syngvab support the contention that non-speech sounds as drawn out by a robotic creature are a potent means of exploring and investigating the unspeakable. syngvab, along with syngvaa, were my Master’s thesis projects at the MIT Media Lab.