What are the disadvantages of Element Theory in phonology? In other words, why is the theory of binary features still commonly held, even though aspects of it, such as its binarity, are untenable? Are they untenable??
Add a comment
|
1 Answer
Element theory differs from SPE theory (including autosegmental versions prior to Clements 1985) in two fundamental ways. Elements are in a way more abstract w.r.t. phonetic definitions, and they are privative. Additionally, the claim is that any single element is a well-formed segment and can be pronounced as such. Subsequent to Clements 1985, the difference between the theories has largely disappeared. Privativity was introduced partially via the node / feature distinction in geometric theories (nodes are labels with no features), and partly in response to radical underspecification (the ternary theory with +, - and 0). Following the logic of Lombardi (dissertation), it was realized that [+X] can be represented as X dominated by Y (dominated by Z) and [-X] can be represented as terminal Y dominated by Z: spreading of [-X] simply requires spreading Y, which wipes out any subordinate specification of X. There being no factual issue at stake, it has become more a matter of style or deeper-seated theoretical principles (such as a requirement to minimize the number of dominating nodes allowed in the theory). Since there are no clear and compelling arguments for or against either theory, we have reached a stalemate.
A classical example of the move towards more abstract features is seen in the adoption of "labial" which encompasses bilabial and labiodental in consonants, and rounding in vowels and glides. SPE theory required [round] for round segments, and [+anterior,-coronal] for labial consonants – Post-Sagian feature geometry, especially Unified Features theory, requires just Labial. The Parallel Structures Model (PSM) is even more abstract, in positing very few features which have only minimal phonetic content, and is just a step away from Radical Substance Free Phonology which does not assume phonetic feature definitions at all. Element theory started with just a very few elements and they employed relational combinations to derive the same phonetic outputs: that is, feature theory converged on ET.
The closest theory pair that I can see is Element Theory and PSM. They differ in what the primitives are, and I think they differ in some minor bit inter-translatable formal ways, but the one main difference between PSM and ET is that in PSM, you sill need a language-specific interpretation device that tells you what a segment with a single feature is. In ET, |A| is [a]: that theory retains a predefined universal phonetic core.
One of the motivations for ET is it's program of reifying crosslinguistic probability distribution in representations: uncommon vowels are structurally more complex. Since PSM and RSFP are strictly theories of computation, probability distribution is a non-issue. So to put the matter somewhat differently, if you want your theory of grammar to encode Markedness Theory, then you would like ET. If you want a simpler theory which shifts the burden of explanation to grammar-external factors (various phonetic factors shaping the primary data that are the inductive base for grammar-learning), PSM / RSFP would be more appealing.
