Galician geada: In defense of underspecification in Optimality Theory

In Galician geada a voiced velar fricative or approximant surfaces as a voiceless fricative in all contexts except post-nasally, e.g., pega [péxa] 'magpie' (non-geada dialects [péγ{phonetic}a]) vs. longo [ló{eng}go] 'long'. Although geada seems to be a rather basic phonological problem, existing analyses are complex. This paper shows that the analysis becomes quite straightforward once the connection of geada with voiced obstruent alternations is recognized. The proposed account obviates the need to resort to controversial mechanisms such as constraint conjunction and sheds light into the allophonic realizations of voiced obstruents. One crucial aspect of voiced obstruent allophones in Galician is their predictability and their underspecification for continuancy. Underspecification is argued to be compatible with Optimality Theory, as long as the underspecified input is not stipulated, but derived from the constraints and constraint ranking. In Galician the phonological system reveals the need for a three-way contrast in obstruents: [-continuant], [+continuant] and underspecified [continuant]. The findings about voiced obstruents have consequences for NC heterosyllabic clusters, showing that there is no voicing requirement at stake, rather a continuancy one, according to which voiced obstruents must be [-continuant] after nasals. Only voiced obstruents agree in continuancy with nasals, because only voiced obstruents are underspecified for continuancy. Constraint reranking accounts for dialectal variation.