Onset Sonority Constraints and Subsyllabic Structure
 

Jennifer L. Smith 

UNC Chapel Hill 

April 15, 2003



Often, abstract formal structure and functionally grounded markedness constraints are seen as distinct, or even opposing, approaches to phonological analysis.  However, this paper argues that certain functionally grounded markedness constraints must actually be defined with reference to formal phonological structure.
 



In some languages, liquid onsets are prohibited.  This pattern seems to have a functionally grounded account based on the sonority scale, consistent with the cross-linguistic tendency to avoid high-sonority onsets (implemented in OT as the *MARGIN/X constraint family of Prince & Smolensky 1993).  But if a prohibition on liquid onsets is to be given a sonority-based (functionally grounded) account, then a liquid-onset ban should imply a glide-onset ban:  glides are even higher in sonority.  Some languages do in fact ban glides along with liquids, but there are languages that do not.
 



The proposal advanced here is that *MARGIN/X constraints are defined in terms of formal syllable structure.  True-onset glides (daughters of the syllable node) violate *MARGIN/GLIDE, but nuclear onglides in rising diphthongs (daughters of a mora) do not.  A language allowing syllable-initial glides while banning liquid onsets is now predicted to have nuclear onglides, not true-onset glides.  Support for this distinction is found in a comparison of two dialects of Campidanian Sardinian (Bolognesi 1998) that have liquid-onset restrictions in initial syllables.  Crucially, the ability to have initial glides correlates with the ability to have rising diphthongs in other contexts.
 



This proposal has two important implications.  First, even though adopting ranked and violable constraints sometimes allows us to simplify our assumptions about formal phonological structure, there is still a role for such structure in our understanding of sound patterns in language. Second, a functionally grounded constraint is not necessarily one that is created directly from functional considerations. It can also be a formally defined constraint that is compatible with functionally determined criteria (as argued by Hayes (1999)).