UC Santa Cruz

Debuccalization is a weakening phenomenon whereby various consonants reduce to laryngeals. Examples include Spanish s-aspiration (s becomes h word-finally) and English t-glottalization (t becomes glottal stop syllable-finally). Previous analyses of debuccalization view it as a lenition process that deletes or manipulates formal phonological features. This dissertation frames debuccalization based on the articulatory gestures involved rather than features. In the default case, debuccalization processes delete oral gestures and leave behind laryngeal gestures. This is captured in an Optimality Theoretic grammar by the low ranking of faithfulness to oral gestures, and the high ranking of faithfulness to laryngeal gestures. The motivation for changing the consonant is a markedness constraint which favors more articulatorily efficient consonants. When this constraint outranks oral faithfulness, consonants debuccalize.

This constraint system allows us to account for many cases of debuccalization well, but it highlights other cases where the laryngeal gestures of the debuccalized consonant and the fortis version are not identical---cases which show supplementary gestures. These supplementary gestures require an explanation, and three competing analyses are considered: perceptual faithfulness, dissimilation, and neutralization avoidance. The perceptual faithfulness analysis claims that the laryngeal gesture is changed to make the resulting sound more similar to the underlying sound. The dissimilation analysis states that the laryngeal gesture is changed to make it less like its neighboring sounds. Finally, the neutralization avoidance analysis claims that laryngeal gestures must change to avoid neutralization with other phonemes.

I first demonstrate that all three analyses are able to account for the case study of Indonesian k-debuccalization (k becomes glottal stop in coda position). Moreover, all three are consistent with the principles of Optimality Theory and have plausible motivations. As such, experimental data is used to provide evidence for or against the competing analyses. A perceptual experiment was designed to verify the perceptual faithfulness and dissimilation analyses, evaluating the predictions each analysis makes regarding perceptual similarity. Some evidence is found for the perceptual faithfulness account of Indonesian. An artificial grammar learning experiment was also performed, which was used to test the possibility of a learning bias of avoiding neutralization in phoneme inventories, and the evidence from this experiment provides support for the neutralization avoidance analysis.