Participants were asked to try to describe the sound of their own voices as they speak, so that what’s describing is what’s being described and vice versa. In addition to being recorded with a microphone, these descriptions were transcribed using Google’s voice-to-text API. Each participant was then filmed listening back to his/her respective recording. The original audio is synced up to the video so that we hear exactly what the participants are listening to in real time.
Listening to oneself speak makes speaking hard enough, but speaking about what one hears is all but impossible, since as soon as one stops to think, the object of description disappears. The absurdity of the participants’ words appear to be a measure for how close they come to the limit of reversibility (simultaneously speaking and hearing, subject and object): the more fluent their speech, the more off-topic they become; the harder they try to listen, the less they can say (and the less they have to listen to). It seems the sonic characteristics that differentiate each voice from another fail to contribute to or be captured by signification, yet it’s not as if an accurate description of the voice is necessary to know how it sounds. It goes without saying that the sound of one’s voice is instantiated in each word one uses to describe it, regardless of how accurate the description is. Therefore, the impossibility of the task is meant not to reveal an imminent failure, which in this case is predictable, but rather to condition a performance that generates spontaneous expressive locutions.