|Title:||Between-Language Frequency Effects in Phonological Theory|
|Authors:||Andries W Coetzee|
|Comment:||Ms from 2002|
|Abstract:||Many phonological properties are not evenly distributed in the languages of the world. For instance, only 25% of the languages in the UPSID database have back unrounded vowels in their phonemic inventories, and even fewer (10%) have front rounded vowels. Optimality Theory is very successful in differentiating between possible and impossible patterns (through factorial typology). However, in classic OT there is no straight-forward way of accounting for frequency differences amongst the possible systems. I argue that the central aim of linguistic theory is to account for significant linguistic generalizations. In so far as these uneven distributions of phonological properties are significant, they therefore fall within the scope of linguistic theory. An adequate theory of phonology must be able to account for these inter-language distributions. I show that Optimality Theory can easily be extended so that it can also account for these inter-language distributional patterns. Two specific extensions to the classic Optimality Theory model are necessary:
(i) The OT-model that Anttila (1995, 1997) developed to account for within language variation and frequency effects is extended so that it can also make predictions about between-language frequency effects.
(ii) A new family of constraints, Preference Constraints, is introduced. These constraints evaluate constraint rankings rather than candidates. They penalize rankings that would result in patterns that are attested less frequently crosslinguistically. In this way Preference Constraints help to account for the fact that the output patterns associated with these rankings are observed less frequently.
By these two extensions to Optimality Theory the coverage of the theory is extended to also include cross-linguistic frequency distributions.
I also argue that the introduction of Preference Constraints obviates the need for fixed markedness hierarchies (such as the place markedness hierarchy). The extension to Optimality Theory argued for in this paper therefore also has implications for how markedness is expressed in the theory.