466 MICHAEL SILVERSTEIN
nailing; second, the fashioning of a theory of inherent phonetic content of
phonemes that is necessarily both substantive and functional, like that of
the Prague School theorists. The Saussurean advance, never actually taken
by Saussure himself, but only by such successors as Leonard Bloomfield
(the first explicit Saussurean in America), provides a strictly distributional
framework for describing the inventory of units and their "tactics," their
rules of combination into syntagmatic sequences in (perhaps) hierarchical
constituencies. As many commentators on this understanding have pointed
out, Chao (1934) perhaps the first, there is no uniqueness of solution to the
problem of phonemicizing the forms of a language, just with these assump
tions of a distributional structure. What is nonetheless necessary is a theory
of the substantive nature of phonemic units, that is, of the so-called
"phonetic" content of possible units in possible systems of relative distribu
tion. A rich enough theory, it is assumed, will determine a unique, or at
least a best, solution to phonemicization.
The Praguean theory, and all its successors, have this in common,
whether or not they admit it. The two necessary requirements for a
phonemic theory are presupposed and joined in the fundamental formal-
functional hypothesis on the nature of phonemic coding of phonetic charac
teristics of speech: phonemes function in all languages to build up syllables
of relatively similar, cross-linguistically comparable phonetic shapes, which
similarity can be expressed as an extensional metalanguage of sound. That
segmental phonemes function in syllables says that there is an integral unit
into which all words have to be discretely segmentable so that we can state
the distributional facts of phonological structure. Whether we state them
with strict surface forms or postulate "agglutinating analogues," as Floyd
Lounsbury (1953:13) so astutely called the underlying forms of morpho-
phonology, is not at issue; in the latter case we need transformational
phonological rules to yield the surface distributions from the underlying,
idealized morphological ones. Notice then that the term "function" is a
purely distributional one from this point of view. That syllables are of rela
tively similar, cross-linguistically comparable phonetic shapes says that it
must be necessary to show why the syllable skcqwIx "my two hips" as a type
in Chinookan is no different from the syllable spilz as a type in English,
given their inventories of phonemes; i.e., we must have a language-inde
pendent theory of possible syllable types (and how they combine), and
hence a way of talking about them. That this way of talking about syllables
should be an extensional metalanguage of sound, says just that the inten-