A rich array of different types of correspondence constraints (CCs) may be found in contemporary optimality theory (OT) literature. Descriptive coverage is impressive. But some CCs implicitly entail radical and probably unacceptable revisions in OT. We use elementary model theory to expose the problem. We develop an extensible metalanguage (a kind of modal logic) for making precise the content of OT phonological constraints, and specify a class of structures with which to give our language a model-theoretic intepretation. Our structures embody standard OT assumptions about phonological candidates. We show how to state the content of markedness, input-output faithfulness, base-reduplicant faithfulness, and paradigm uniformity constraints. OO CCs, however, emerge as not just complicated to restate in these terms but actually impossible. The radical difference is that OO constraints appeal to correspondence between distinct forms. We illustrate using a description of truncation from Laura Benua's influential dissertation, which inspired much work on OO constraints. We note that asking what an OO constraint says about candidates makes no sense: OO CCs determine properties not of finite individual candidates, but of (potentially infinite) candidate sets in which some members may not even be grammatical. Similar remarks hold for sympathy analyses and targeted constraints.
@UNPUBLISHED{PottsPullum03LSA,
| AUTHOR = | "Potts, Christopher and Geoffrey K.~Pullum", | |
| TITLE = | "Model theory and output--output correspondence", | |
| NOTE = | "Paper presented at the Linguistic Society of America, Annual Meeting. Atlanta, GA (January 5--9)", | |
| YEAR = | "2003", | |
| annote = | "Date and time of session: 01/04/03, 9:00 a.m.; Title of session: Phonology: models and evidence", |