The main goal of this dissertation is to provide a generative account of phonological opacity within a framework built upon direct mapping in the synchronic grammar without abstract intermediate representations, as in standard Optimality Theory (OT; Prince and Smolensky 1993/2002). Such a framework predicts that certain types of opacity cannot be synchronically productive. I take this prediction seriously and develop an analysis in which opacity is shown to arise from the interaction of sound change and a strong version of Prince and Smolensky's principle of lexicon optimization, in which the underlying lexicon is optimized by becoming more faithful to the surface pronunciation. This interaction results in a progressive encoding of sound changes directly into the evolving lexicon, mirroring the stepwise effect of multistratal derivations, but diachronically rather than synchronically, preserving direct mapping.
The specific theoretical framework used in this dissertation is Faithfulness, Dispersion, and Markedness in OT (FDM-OT), which differs from standard OT by offering a functional account of sound change and synchronic phonology through the interaction of faithfulness (along the lines of McCarthy and Prince 1995), dispersion (generalized from Dispersion Theory (Flemming 1995, Padgett 1997, and Ní Chiosáin and Padgett 2001)), and universally ranked articulatory markedness constraints. Grounded in cognition, acoustics, and articulation, FDM-OT explains and predicts phonological patterns with fewer arbitrary or abstract stipulations than are required by competing theories.
I analyze three well-known instances of opacity from Polish, the language of focus. Additionally, I provide analyses of opacity in refined Low German, Turkish, and Tuyuca, arguing that all cases of opacity fit into the following typology: (i) synchronically unproductive opacity, which fails to apply to nonce forms and to lexical exceptions but is still pervasive in the lexicon due to lexicon optimization; (ii) morphologically conditioned opacity, which may be synchronically productive, but only at particular morphological boundaries because the relevant affixes have allomorphs created by lexicon optimization that are encoded with historically opaque alternations; and (iii) transparent 'opacity', which can be reanalyzed transparently because original opaque analyses lacked sufficient phonetic detail or access to certain recent theoretical advances, such as FDM-OT's dispersion constraints.
[PDF 1up] [PDF 2up.]
The 1up version has equal margins on the left and right for printing one page per sheet of paper, while the 2up version has mirrored margins that are balanced when printed two pages per sheet. Contact me at firstname.lastname@example.org if you have any questions about my dissertation.