Feature Classes



Jaye Padgett

UC Santa Cruz





This paper and the next investigate the status of feature classes like 
 Place and Laryngeal in featural phonology. In a departure from Feature
 Geometry, I argue that constraints mentioning such classes are
 gradiently violable in the sense given within Optimality Theory.  The
 result is what I call partial class behavior: under compulsion of a
 constraint mentioning an entire feature class, only a proper subset of
 the relevant features spread, for reasons of markedness, locality or
 whatever. This idea leads to feature class generalizations that are
 not visible to Feature Geometry.



ROA-112 strengthens the evidence for a class Color (cf. Odden's 1991
 [back]/[round]), arguing that the copatternings of [back] and 
 [round] harmonies in Turkic and other languages represent instances
 of Color harmony, formerly unrecognized only because Color spreading
 is often gradiently violated, leading to partial class behavior. This
 paper also argues for an understanding of segment transparency as
 segment participation, following work of Smolensky and McCarthy.
 

ROA-113 focusses on nasal place assimilation and has two goals. The
 first is to demonstrate the robust existence of partial class
 behavior in nasal place assimilation, and so further motivate the
 ideas outlined above. The second is to better understand nasal
 place assimilation itself. This latter goal leads to proposals for
 the employment of a release feature in phonology (following Steriade)
 and for very general spreading constraints.



ROA-113 supersedes ROA-112 in some technical details, and the two
 papers should be read together.