[Author Login]
Title:Segmental unmarkedness versus input preservation in reduplication
Authors:Moira Yip
Comment:43pp. This paper will appear in Lombardi, Linda (ed.), Segmental Phonology in Optimality Theory. Cambridge University Press
Abstract:Segmental unmarkedness versus input preservation in reduplication

Moira Yip

University of California, Irvine and University College London

Reduplication in many Chinese languages differs in two ways from the types

of reduplication that are most discussed in the literature. First, it is

not obvious which piece of the output is the base and which is the

affixed copy, and instead the outputs look more like compounds of a word

with itself. Second, and the focus of this paper, the copies are

imperfect, with various segments from the input being replaced by fixed

segments, [l] in onsets, [i] in nuclei, and [ ] or [ ] in codas. I shall

argue that these segmental replacements are the unmarked segments for

these syllabic positions, an instance of the emergence of the unmarked

(TETU), as discussed in Alderete et al. (1998). In non-reduplicative

morphology, Faith-IO preserves marked input segments. In reduplication,

I propose that each input segment has two output correspondents, one of

which is thus free to succumb to markedness pressure so long as the other

remains intact. All faithfulness relations hold between input and output,

in line with a converging body of work including Sherrard 1997, Yip 1998,

Inkelas and Zoll 1999, Struijke 2000.

After a survey of the data, I begin by laying out the basic analysis of

reduplication as a response to two constraints, Alliterate and Rhyme. I

then show how the ranking of these with respect to segmental markedness

constraints gives rise to the segmental changes. If Rhyme >> Markedness >>

Alliterate, marked segments survive in the rhyme but onset segments

become the unmarked [l]. If Alliterate >> Markedness >> Rhyme, onset

segments survive but rhyme segments become unmarked [i] or glottal stop.

I continue by discussing the particular choice of unmarked segments, and

propose a set of markedness constraints. The final two sections discuss

other reduplicative forms from secret languages where the replacement

segments do not appear to be the most unmarked ones. In the first case

the defining characteristic of secret languages, a high-ranked constraint

requiring the output to be distinct from the input, forces an increase in

markedness. In the second case, conflicting markedness constraints

interact with the secret language constraint to produce surface marked


[This paper will appear in Lombardi, Linda (ed.), Segmental Phonology in

Optimality Theory. Cambridge University Press.]
Type:Paper/tech report
Article:Version 1