Events

Coda devoicing and perceptual similarity

Languages resolve voiced obstruents in codas by devoicing but not by any other phonological means (i.e. nasalization, deletion or epenthesis). For example, the underlying /ab/ can become [ap], but not *[am], *[aba] or *[a]. Steriade (2001/2008) claims that (i) speakers maximize the similarity between inputs and outputs, assuming that (ii) devoicing yields an outcome that is most similar to the original form.
In this talk, I present my recent attempts to test the premise of this hypothesis (clause (ii) above).