Dissimilation by Correspondence in Sundanese

The Agreement-By-Correspondence framework (=‘ABC’) is a theory of agreement developed (by Walker 2000a,b, Hansson 2001, and especially Rose & Walker 2004) to explain long-distance consonant harmony: patterns where non-adjacent consonants agree with each other, but do not interact with the other material that intervenes between them. In ABC, the basis for this agreement is Surface Correspondence (‘SCorr’): correspondence relationships that hold between different consonants of the same output form. The arrangement of these correspondences affects the input-output mapping because there are constraints that take them into consideration when assessing violations. Three types of constraints assess violations based on surface correspondences, and they play distinct roles in ABC analyses of consonant harmony. First, there are the CORR constraints, which assign violations when output consonants with some shared feature(s) are not in surface correspondence with each other. Their role is requiring certain consonants to correspond with each other, thereby picking out the class of interacting consonants in a harmony pattern. Next, there are the CC-IDENT constraints. These constraints assign violations when consonants that are in surface correspondence with each other disagree in some feature(s). They demand featural agreement among correspondents, which provides the impetus for long-distance assimilation. Finally, there are Structural SCorr constraints; they assign violations for correspondence between consonants that are in some particular morphological or prosodic configuration. In a harmony system, these constraints limit the extent of agreement, by allowing correspondence only under certain conditions. The interaction of CORR, CC-IDENT, and Structural SCorr constraints can be seen in a simple ABC analysis of a consonant harmony system, such as Kikongo nasal consonant harmony (1) (see Rose & Walker 2004, Walker 2000b for more details). The basic generalization in Kikongo is that sonorants in the stem agree for nasality; this causes suffixes with /l/ (1a) to assimilate to [n] when the root contains a nasal consonant (1b). However, coda nasals are inert: they do not trigger the same agreement (1c).