| Abstract: | There is a rather vast phonetic literature showing that final devoicing in languages such as German, Catalan or Dutch is not complete: there are subtle acoustic differences between devoiced and underlyingly voiceless consonants, and these differences can also be perceived. Some authors have claimed that this constitutes an argument against 'formal phonology' based on a discrete alphabet of symbols. This article argues that this is not the case, and in particular that a theory based on ideas of Containment (Prince and Smolensky 1993) and Turbidity (Goldrick 2000, Revithiadou 2006) predicts incomplete devoicing, exactly because it is a formal theory of phonology, albeit one using representations which are a little more sophisticated than those used in SPE. |
