Committee: Lisa Davidson (chair), Gillian Gallagher (chair), Maria Gouskova, Frans Adriaans, and Tara McAllister Byun.
This dissertation provides evidence from phonetics, phonology and learning to shed light on patterns of contrast neutralization and contrast enhancement in phonological systems. I examine the case of place contrasts among sibilants, which exhibit various patterns in the context of the high front vowel. Specific questions to be addressed include: (i) how to properly characterize sibilants phonetically and what articulatory and perceptual considerations apply to sibilant contrasts in vocalic context and (ii) the phonology of sibilant contrast distribution and neutralization versus enhancement of sibilant contrasts before the high front vowel.
In this dissertation, a number of extant sibilant inventories will be examined using articulatory and perceptual methods in order to establish some of the phonetic factors that must necessarily be taken into consideration in order to better understand patterns of contrast enhancement and neutralization. Two of the languages that this dissertation focuses on are Mandarin and Polish, as these languages exhibit distinct patterns for enhancing contrast in sibilant inventories. In Chapter 2, I start by establishing the phonetic characteristics of Mandarin syllabic approximants that occur after dental and retroflex sibilants instead of the high front vowel (e.g. /Sr Şr ci/, but /*si *Şi ci/) using articulatory and acoustic methods. To examine the phonetic basis for the attested pattern, Chapter 3 shows that the cues to the contrast are impoverished in the high front vowel context compared to other vowels in an acoustic and perception study looking at the [s Ş c] sibilant contrast in Polish. Once the phonetic properties of sibilant contrasts have been established, I go on in Chapter 4 to identify relevant factors underlying the possible modifications to the sibilant contrast, ranging from contrast neutralization to an extreme level of enhancement. In particular, it will be shown that varying degrees of enhancement through modifications of the neighboring vowel are attested and that the pattern with a more extreme modification is less common in typology. Building upon the inventory-based model (Flemming 2006), I show that faithfulness to the inventory is another key aspect of sibilant typology. The central idea of this framework is to capture the intuition that despite some contextual variance it is desirable that phonetic surface forms of a sound are maintained relatively consistent across contexts. In this regard, extreme enhancement (e.g. [r] as in Mandarin) is disfavored, since it differs remarkably from the canonical form of a sound in the inventory (e.g. /i/). Having identified a formal bias against low-ranked faithfulness in a grammar, in Chapter 5 I provide evidence of its influence in learning using an artificial grammar paradigm.