| This thesis explores the issue of segmental representations in Optimality Theory. I make two claims relating to this issue. First, I argue that input segmental representations may vary from language to language, and that these differences influence the patterning of different segment types in the phonology. Second, I argue that segmental markedness is evaluated in terms of representational complexity: the more complex a representation is, the more marked it is. Markedness relations are thus encoded in the segmental representations, so that marked segments involve more structure than unmarked segments.;The variation that is found in segmental markedness across languages is expected under the proposed theory, since representations are constructed in response to contrasts in the inventory. Since segmental representations are influenced by contrast, variation in inventories entails variation in segmental representations.;The representations that I propose are built in response to the competing considerations of structure minimization and the need to make contrasting specifications representationally distinct. Structure is only added to a specification under the pressure of contrast. Since additional structure means a more complex representation, and markedness is linked to complexity, the markedness status of a particular segment may vary depending on the number and type of contrasts in an inventory. My proposal entails that input representations play a more central role in the phonology than is currently assumed in the Optimality Theory literature. I argue that, despite claims to the contrary, all cross-linguistic variation cannot be made to fall out from output constraint ranking alone. Instead, phonemic inventories are necessarily defined in the input, and therefore there is no universal set of inputs. |
