Lecture: Sara Mackenzie

Restricted Structure Preservation in Stratal Optimality Theory 

This talk investigates the role of structure preservation within the framework of Stratal Optimality Theory (e.g. Kiparsky 2000) through an analysis of German dorsal fricative assimilation. The principle of structure preservation (e.g. Kiparsky 1985) prohibits the creation of allophones during the course of operations in the lexical phonology. Although structure preservation has largely been rejected within Optimality Theory, previous work has shown that processes which are both neutralizing and non-structure-preserving result in a ranking paradox in a single, parallel OT evaluation (e.g. Krämer 2006). This has been presented as an argument that such processes must apply at the word or phrase level in a Stratal model of OT (Bermúdez-Otero 2007, Mackenzie 2016).  

The lexical phonology literature, however, includes numerous cases of purely allophonic processes that appear to apply early in the lexical phonology (e.g. Harris 1990). This talk considers German dorsal fricative assimilation as one such case. In German, [x] and [ç] are in complementary distribution with [x] occurring after back vowels and [ç] occurring elsewhere. The back variant of the fricative does not occur when a morpheme boundary intervenes between the fricative and a preceding back vowel, resulting in well-known surface contrasts such as [kuxən] ‘cake’, [ku-çən] ‘little cow’. These data have been argued to provide a counterexample to structure preservation as they require the allophonic process to occur early in the lexical phonology (e.g. Hall 1989). If assimilation is motivated by constraints which penalize marked feature sequences, a ranking paradox similar to that demonstrated in analyses of neutralizing and non-structure-preserving processes arises. Instead, this talk argues that purely allophonic processes occurring at the earliest lexical level are motivated by constraints which require rich output specifications. This approach is integrated with a model of contrastive specifications in which a hierarchy of featural faithfulness constraints maps the rich base to contrastively specified outputs (e.g. Dresher 2009).