=========================================================================
UMASS LINGUISTICS COLLOQUIUM                                 MAY 7, 2004
=========================================================================

Reinterpretation of Contrast Across Language


Ellen Broselow
Stony Brook University (SUNY at Stony Brook)

Speakers confronted with structures that are not found in their native
language often employ systematic adaptation patterns which can be
described by means of ranked constraints. In many cases, however, these
rankings are not obviously motivated by the data of either the native or
the foreign language. Nor do these rankings necessarily correspond to what
have been claimed to be universal default rankings; for example, in
borrowings into Fijian, English [b] is realized as [mb] (Kenstowicz 2003),
in violation of the putative universal ranking IDnas >> IDvoi (Steriade
2001). Furthermore, the rankings associated with foreign word modification
may be antithetical to the rankings needed to describe native patterns; in
Shona, for example, the quality of vowels inserted in borrowed words is
dependent on nearby consonant and vowel quality, while vowels inserted to
repair minimality violations in native vocabulary are always the default
[i], regardless of phonetic context. Facts like these, as well as the
tendency for greater tolerance of some foreign structures over other
(equally foreign) structures, have led to the assumption of different
subgrammars associated with different lexical strata, each with their own
set of rankings.

Given the general assumption that constraints are universal but rankings
are language-specific, we need to explain how speakers would have arrived
at these particular subgrammar rankings --- not an easy task when the only
evidence for the rankings rests in the modification patterns themselves. I
will argue that many of the patterns used to motivate particular loanword
rankings do not in fact reflect the rankings of the production grammar;
rather, they reflect listeners' attempts to interpret the acoustic data
in terms of their native language system of contrast.