We are having our final talk of the year: Who: Aaron Dinkin (U of T) What: "Gradience, allophony and chain shifts" -- see abstract below. When: Friday, April 4th at 2:30pm sharp (<-- 1073="" be="" by="" end="" followed="" font="" in="" location="" lounge.="" note="" of="" party="" sid="" smith="" talk="" that="" the="" time="" unusual="" where:="" will="" year="">--> Gradience, allophony and chain shifts / Aaron Dinkin A modular feedforward architecture of phonology (Bermúdez-Otero 2007) implies that the entities that undergo chain shifting are not phonemes per se, but the intermediate representations that are the outputs of discrete phonological rules. I test the predictions made by this model by examining the role of gradiency in the Northern Cities Shift and the Southern Shift. In Upstate New York (Dinkin 2009), the Hudson Valley and Inland North regions have distinct systems of /æ/ allophony, whereby the Hudson Valley possesses two phonologically distinct allophones of /æ/ (prenasal and non-prenasal), while the Inland North has prenasal and non-prenasal /æ/ as part of a single gradient distribution. I argue that this phonological difference explains the Hudson Valley's failure to acquire the raising of (non-prenasal) /æ/, while it acquires other Northern Cities Shift vowel features. Meanwhile, in the Southern Shift, the modular feedforward architecture suggests that the monophthongization of /ay/ must have been a gradient process which pre-voiced and pre-voiceless /ay/ were both participating in from the initiation of the shift. Analysis of Southern speakers from Labov et al. (2006) suggests that this is the case, and that discrete allophonic differences in monophthongization between pre-voiced and pre-voiceless /ay/ are a secondary development (for those speakers for whom they exist).