|Comment:||Has minor corrections from published version, can be cited as in the Proceedings of the 29th Boston University Conference on Language Development, edited by Alejna Brugos, Manuella R. Clark-Cotton, and Seungwan Ha, Cascadilla Press, Somerville, MA. 2005.
|Abstract:||This paper shows that a simple extension of the Biased Constraint Demotion Algorithm (Prince and Tesar 2004) results in the construction of stratified grammars (Ito and Mester 1999). Phonological structures are sometimes restricted to particular sets of words, such as loanwords. To capture such generalizations, Ito and Mester (1999) propose that faithfulness constraints applying to subsets of the lexicon are interspersed between markedness constraints. Three learnability problems present themselves: 1. How does a learner create lexically specific constraints for exceptions to phonotactics? 2. How do the markedness constraints get in the right order? 3. How do the faithfulness constraints get interspersed correctly? To address 1, this paper proposes that when a learner encounters a form that requires an adjustment to the grammar, it makes the initial conservative assumption that this adjustment is specific to that word. More formally, in terms of Tesar and Smolensky (1998) et seq., when Error-Driven Constraint Demotion produces a Mark-Data pair, faithfulness constraints preferring the winner are indexed to the lexical item in question. With this one assumption, Biased Constraint Demotion automatically yields answers to problems 2 and 3. In applying lexically specific constraints to instances of gradient phonotactics, this paper also suggests an approach to the problems raised by Frisch, Pierrehumbert and Broe (2004).