Syllabification

Abstract

Abstract One of the first problems confronted by students of Armenian is the discrepancy between the orthography, which contains consonant clusters of as many as ten members, and the pronunciation, which displays relatively manageable clusters and any number of schwas, almost none of which are reflected in the orthography. A typical example is the word for ‘brilliant’, which is spelled <plsthrplsthr> but pronounced [pa.las.thar.pa.las.thJr]. One immediately wonders whether the orthography directly represents underlying lexical representations which are subsequently altered by rules of epenthesis, or alternately that the schwas present in surface pronunciations are already present in underlying representations, and the orthography is underdetermined. If schwas are present in underlying representations we expect them to show the same distribution as other vowels in the Armenian inventory, and to defy attempts to predict their occurrence from principles of epenthesis. Conversely, if schwas are not present in underlying representations we expect their occurrences to be predictable from underlying representations by simple and universal principles. I argue in this chapter that the latter hypothesis is correct, and demonstrate that a modified version of the syllabification algorithm postulated as a component of Universal Grammar by Dell and Elmedlaoui (1985) accounts for the occurrence of Armenian schwas in a principled and elegant manner.