Constraints on Historical Change: The History of Stress in Welsh

The representation of historical change poses a challenge for Optimality Theory. On one hand, the basic tenets of Optimality Theory (Prince and Smolensky 1993) urge us to think non-derivationally about synchronic phonology. On the other hand, historical phonological change can be seen as inherently derivational — we have an input (the original form) and an output (the new form), and a mapping between the two (the diachronic rule). According to the traditional neo-grammarian approach as laid out by Paul (1920) and summarized by Kiparsky (1965), phonological change is a combination of sound change (Lautwandel) and imperfect learning (abweichende Neuerzeugung). As Kiparsky points out, the fundamental insight here is that, “[s]ound change is brought about by the joint operation of innovation and subsequent restructuring.” In the generative framework, this is easily accomplished by means of rules. In the simplest case, a change is merely the addition or deletion of a rule, or the re-ordering of two pre-existing rules in the grammar. A more complex version of this would be that change involves a whole range of competing forms which eventually get “weeded out” in favor of the preferred form. This can be represented as follows (Vennemann 1988):