Positional Neutralization - Linguistics - University of California ...

uncovered such an effect. It is also possible that the phrase-boundary selected for
comparison with the word-boundary was insufficiently high in the hierarchy to trigger the
increased strengthening effect. Further work in this direction would do well to vary the
sentential contexts used, in order to control for possible problems of this type. One last
problem may also have been in the choice of real-word stimuli, which, although carefully
controlled, may nonetheless have introduced sufficient variation into the results to make
detection of this level of durational asymmetry difficult. The studies that detected the
hierarchical strengthening patterns all used single repeated stimuli and nonsense words
representing each phonetic environment under investigation. Such a strategy could prove
more effective in Turkish as well.
We have seen so far that Turkish, but not English, exhibits a pattern of domain-
initial strengthening affecting the vowels of initial syllables. Minimally then, we can say
that domain-initial strengthening is implemented to differing degrees in different
circumstances on a language-specific basis. This was shown in the earlier experiments as
well (i.e. Keating, Cho, Fougeron and Hsu 1999), in service of the point that insofar as
domain-initial strengthening varies on a language-specific basis, it cannot be relegated to
the level of “universal phonetic implementation”, and must receive some representation
in the grammars of the languages in question.
322