|Title:||Floating phonotactics: Infixation and reduplication in Tagalog loanwords|
|Comment:||from 1996; equations in Appendix B suffered under conversion from an ancient word-processor|
|Abstract:||When a phonological constraint is unviolated in a language, but no lexical items exist for which the constraint is relevant, the constraint's ranking cannot be determined. Language contact can introduce loanwords which, if faithfully parsed, violate the constraint, thus exposing its ranking. Tagalog has many English and Spanish loanwords which, in the loaning language, contain segments and sequences foreign to Tagalog. Field data presented here show that constraints against the foreign segments and sequences ("new" constraints) are variably ranked with respect to established constraints (phonotactic constraints which were active in Tagalog prior to the loanwords' introduction, and the constraints which enforce uniformity of allomorphs and faithfulness to the lexical entry). That is, the new constraints may be obeyed despite violations of established constraints, or may be violated in order to obey the established constraints.
Constraints relevant to Spanish loanwords, which have been in Tagalog longer, tend to be ranked lower, suggesting they have drifted downward. Variable ranking is natural for a constraint whose ranking was, until recently, indeterminate. Older loanwords from Chinese and Sanskrit demonstrate that similar variability existed at the time of their adoption into Tagalog, although individual words have since become fossilized in the absence of strong, continuing language contact. Part II of the thesis presents an algorithm for imperfect learning of variable data, which results in a speech community's tolerating lower and lower ranking of a variably-ranked constraint over time.