Ph.D. Dissertation, Romance Linguistics, UCLA
One of the main arguments for the inclusion of syllables in phonological theory is that syllable-based analyses shed light on the interrelatedness of multiple phenomena. This dissertation investigated the role of the syllable in Italian phonology in three domains through a series of experiments with native speakers of Standard Italian: (i) native speaker intuition of consonant cluster syllabification, (ii) definite article allomorphy (il vs. lo), and (iii) segment duration. The primary goal was to test the predictions of syllable-based analyses. The secondary goal was to test the claimed convergence of multiple phenomena on the same syllable structure.
The consonant cluster syllabification experiments found that native speakers do not use knowledge of sonority distance constraints, sonority sequencing constraints, or the maximum onset principle when confronted with the task of string division. Instead, two strategies for string division were identified: (i) the word-based syllables strategy (Steriade, 1999), and (ii) the phonotactic-constraint satisfaction strategy. The co-existence of these two strategies was reflected in increased response variability in cases where the strategies conflicted.
The syllable-based analysis of masculine singular definite article allomorphy (il vs. lo) was not supported by the experimental data. A syllable-independent phonotactic analysis was proposed in which the same constraints that govern word-internal phonotactics are ranked above the *lo constraint [il is the default]. It was also found that article allomorph selection is only marginally predictable before non-native clusters.
The segment duration experiments found that the maintenance of contrastive length has far reaching effects on the duration of both consonants and vowels. Speakers only produced categorical segment duration differences in the phonotactic context where length is contrastive. Consonant length is contrastive and differences in consonant duration were categorical. Vowel length is not contrastive and differences in vowel duration were gradient. No evidence for open-syllable vowel lengthening was found.
The cross-experimental comparison found that variability in one domain (e.g. allomorphy), does not correspond to variability in the others (e.g. string division, duration). Therefore, the syllable-based analysis was not able to simultaneously generate predictions in multiple domains.