Why Should Markedness Constraints be Relative? -- Four Case Studies in Tone Sandhi Directionality

Abstract

UC Berkeley Phonology Lab Annual Report (2009) Why should Markedness Constraints be Relative? - Four Case Studies in Tone Sandhi Directionality 1 Te-hsin Liu liu.tehsin@berkeley.edu Abstract This paper presents a novel analysis on tone sandhi directionality, introducing a family of relative markedness constraints, i.e. M IN M OD (minimize modulation) and M IN A MP (minimize amplitude of contour tones), to account for the apparently ungoverned traffic of sandhi operations. A closer look at the data leads to the conclusion that tone sandhi directionality is not output-driven but directionality-driven. This observation echoes Hyman & VanBik’s analysis on Hakha-Lai, according to which output-based generalizations fail to capture the input-output relations (2002). Keywords: conflicting directionality, tone sandhi, Optimality theory. 1. Introduction The increasing interest in conflicting directionality has heightened the need for phonological theories to account for such complex and apparently unpredictable phenomena. The material targeted can range from segmental elements (i.e. palatalization in Japanese mimetic words), stress placement (i.e. stress patterns in Selkup) to directional tone sandhi (i.e. Hakha Lai, some Chinese dialects). The present study focuses on the third pattern of conflicting directionality, tone sandhi directionality. In some Chinese dialects, sandhi rules must apply from left to right to derive the outputs for some trisyllabic sequences and from right to left to account for the others. One common point among them is that morphosyntactic structures play no role, a left-branching string and a right-branching string generating the same surface form. The research presented in this paper, supported by Fyssen Foundation, is partly based on a talk given jointly with Joaquim Brandao de Carvalho at the Seventeenth Manchester Phonology Meeting held on May 28-30, 2009. I am grateful to Larry Hyman as well as the audience of the conference, especially Stuart Davis, Andrew Nevins, Marc van Oostendorp and Danial Hall, for their precious comments and feedback. All errors are of course my own responsibility.