Introduction.
Vedic Sanskrit provides strong evidence for the phenomenon of
ambisyllabicity(1), the idea that the
consonantal interlude between two vowels can belong simultaneously to
both of the adjacent syllables. Despite several studies purporting to
demonstrate ambisyllabicity in modern Germanic languages, including
Kahn's classical description of consonantal allophony in English (1976), the existence of
ambisyllabicity has been seriously questioned. In this paper I shall
support the existence of ambisyllabicity, arguing that in the language
of the Rigveda, the oldest extensive literature in Sanskrit, the first
consonant of an intervocalic onset cluster is ambisyllabic, at least
after a short vowel. Since accent is not a conditioning factor, this
is a rather different environment from the stress-sensitive phenomena
described in Germanic, and so the data are not susceptible to Kiparsky's (1979) foot-based
reanalysis. It will also be seen that the Sanskrit data are not
easily tractable by Selkirk's
(1982) resyllabification-based alternative to ambisyllabicity,
either.
Among those who support ambisyllabicity, the details of its
representation are a matter of some debate. Borowsky et al.
(1984) conjectured that there is no distinction between geminate
and ambisyllabic consonants: it is a matter of a language's phonetic
implementation to determine whether consonants shared by adjacent
syllables will be pronounced as short (ambisyllabically, as in
English) or as long (as geminates, as in Italian). On the other hand,
Kang (1991) argued that the two
phenomena contrast in Korean, and so the two must have different
structures. It turns out that Vedic Sanskrit has both ambisyllabicity
and long consonants, but they are in complementary distribution. This
argues for Borowsky et
al.'s conjecture, although with a more fine-grained phonetic
implementation than a single language-wide setting.
[Forward to Background: The reality of
ambisyllabicity.]
[Up to table of contents]