Restrictions on direction of voicing assimilation: an OT account

In Lombardi (1996) I propose a set of constraints within Optimality Theory (Prince and Smolensky 1993) that account for patterns of obstruent devoicing and assimilation phenomena in the world's languages. Rerankings of these constraints produce only patterns in which voicing assimilation is regressive. However, there are cases of progressive voicing assimilation. I will argue in this paper that the constraints of Lombardi (1996) make a correct prediction: that voicing assimilation will always be regressive unless additional constraints are active. While many languages have regressive voicing assimilation in all clusters, I will show that cases of progressive assimilation always involve additional restrictions to special circumstances and thus show the action of additional constraints.