What can we learn about experiences of hearing speech sounds in a voice from auditory illusions? I will address this question by looking at four cases of such illusions: (1) the temporal induction illusion in speech, (2) the phantom words illusion, (3) the McGurk effect, and (4) the voice-over translation illusion. I will argue that (1)-(4) jointly support the claim that the mechanisms responsible for speech sound and voice perception are (at least to some degree) constructive in the following sense. Hearers' expectations, assumptions and background knowledge about the upcoming speech signal can steer the perception of speech sounds and the process of tracking the source that is intentionally producing them. I will also discuss one interesting consequence that this proposal has for the epistemology of spoken language understanding.
Presented at the Fifth PLM Workshop on Delusion in Language and Mind, in Amsterdam, October 23---24, 2020