Similarity in phonology: Evidence from reduplication and loan adaptation
This dissertation concerns the role of similarity in phonology, specifically with respect to two processes: onset simplification in reduplication, and onset simplification in loanword adaptation. These two processes have previously been considered two distinct realms, with different explanations proposed for each. However, I highlight a commonality across the two phenomena. In reduplication and in loanword adaptation, obstruent + sonorant onset clusters exceptionally permit skipping (i.e., deletion or failure to copy the second cluster member) and intrusion (i.e., insertion of a vowel between the two cluster members).
I propose that this exceptional behavior of obstruent + sonorant onsets stems from a general principle: that phonological processes occur more freely when the result of the process sounds quite similar to the original form. Obstruent + sonorant onset clusters are more vulnerable than other clusters to skipping and intrusion---triggered by phonotactic constraints against consonant clusters---because correspondence is evaluated according to the standard of perceptual similarity, and an intact obstruent + sonorant onset is minimally distinct from the result of skipping or intrusion.
The claim, then, is that obstruent + sonorant clusters sound more like their counterparts affected by skipping or intrusion, than other clusters sound like their comparably affected counterparts. That is, pra sounds more like pa or pira than sta sounds like sa or sita. I examine a variety of evidence in support of this claim: linguistic evidence, from alliteration and a large corpus of English-language puns; and direct experimental evidence, from a discrimination task.
With this evidence in hand, I formalize an analysis in which similarity plays a direct role in the grammar, in the form of context-sensitive correspondence constraints which penalize correspondence between more similar forms less severely than correspondence between less similar forms.