Time span:
Project starts:
01.08.2005

Project ends:
31.12.2007

Brief description and goals:
Using Selective Global Comparison to Investigate the Nature of Phonological Features

DRAFT

In this project, we will use the selective global comparison methodology to investigate the nature of phonological features, their relationships to one another, and their phonetic variation. We will do this by exploring the make up of segmental inventories as seen through the lens of a highly constrained and explicit feature theory.The goals are:

1) to provide in-depth studies of a selection of phonological grammars using contrastive segment inventories and a range of phonetic and phonological facts and principles,
2) to test the limits of feature combinations and the relationship between features and their phonetic manifestations,
3) to assess current models of phonological features and segment-internal representations, and
4) to develop a more satisfactory (explicit, restrictive, economical, explanatory, etc.) model.

Why features and segment inventories?
Until the early 1990s with the advent of Optimality Theory (Prince and Smolensky 1993), the major questions within phonology dealt with the nature of phonological features, the relationship among them,
segmental and supra-segmental structure (Autosegmental Phonology, Feature Geometry, Syllable Theory, Moraic Theory), rules, constraints, and the interactions among rules and constraints.

With the introduction of OT and the explicit rejection of phonological and morpho-phonological rules, the focus shifted almost exclusively to one of constraint interactions (often not even questioning the makeup of the constraints themselves). Despite the fact that there are still many unanswered questions regarding the nature of phonological features and how they relate to one another, and despite the fact that
“OT is agnostic with respect to structure” (Lombardi...), the majority of researchers using OT as a tool either reject segment-internal representation outright (e.g. Padgett 2002) or simply ignore questions of feature theory and use features conveniently inherited from pre-OT work without worrying about their validity, relationships, or how they fit into the overall grammar of the given language under investigation.

Although the insights brought to the table by OT must be acknowledged, the featural and structural concerns that were of such import to prior work did not disappear with the rejection of phonological rules - as the recent work of Clements, Dresher, Rice, Avery, and the schools of Dependency Phonology and Government Phonology attests.

Some hypotheses/predictions our restrictive feature theory makes1:
• Small consonant inventories are expected to show phonetic dispersion in a way similar to that found
in vowel inventories.
• Phonetic variation is more likely along a dimension that is not phonologically specified, and a dimension that is phonologically specified with a feature is more likely to resist phonetic variation.
• Phonetic enhancements should only occur along a dimension that does not involve a contrastive
feature.
• The phonetic realization of a particular set of phonological features can differ by context depending
on the contrasts found in that context and a range of articulatory/perceptual influences.
• Phonetically predictable characteristics are not phonologically specified unless they form part of a
contrastive opposition or are made use of in a phonological process.
• Processes/interactions/alternations not involving a contrastive feature are phonetic in nature (too strong?).
• There are no synchronic phonological unnatural classes. What seem like phonetically unnatural
classes are coherent from a phonological feature perspective.
• The presence of a phonologically complex segment implies the presence of a minimally different
phonologically less complex segment. This includes segments composed of only single features.
• Phonological segment acquisition should proceed from less complex to more complex as measured by structural complexity. Phonologically simpler segments should arise before phonologically complex segments.

1 Note that many of these hypotheses have direct impact on important claims in the current literature - e.g. Steriade, Kirchner, Flemming, Boersma, Dresher, Clements, etc.
• Very similar (identical) surface inventories can have very different phonological structures across
languages.
• Very similar (identical) surface segments can have very different phonological structures across languages.
• Consonants and vowels should phonologically interact only along a dimension for which they both
have a contrastive feature (too strong?).