Positional Constraints in Optimality Theory and Harmonic Grammar

The ability of weighted constraints to model certain patterns that elude ranked constraints has often been taken to be typologically problematic (e.g., Prince & Smolensky 1993/2004, Legendre, Sorace & Smolensky 2006). This paper argues that the greater power of weighted constraints can actually be advantageous, permitting a smaller set of more general constraints to be employed in the analysis of attested languages. With limitations on the set of constraints, the predicted typology is restricted beyond what is possible in OT where more, and more specific, constraints are necessary. These principled revisions to CON mean that the languages generated by Harmonic Grammar need not be a superset of those generated by Optimality Theory (see also Pater [this volume], Potts et al. 2010).