Speech production planning affects variation in external sandhi [PhD Thesis]
Oriana Kilbourn-Ceron
July 2017

Phonological variation is common in many alternations, especially in processes where the target and the trigger of the alternation are in different words—external sandhi processes. Much previous work on external sandhi has addressed the morpho-syntactic locality conditions that restrict these cross-word processes, but they are also often sensitive to phonetic and usage factors like pauses, speech rate, lexical frequency, and speech style. What has not been explored in previous work is why these factors are consistently associated with external sandhi. This thesis pursues the hypothesis that patterns of external sandhi variation are shaped by online speech production planning constraints, which can mediate the effect of both grammatical and non-grammatical factors. I investigate the Production Planning Hypothesis (PPH) proposal that the narrow window of phonological encoding can block application of external sandhi processes—if the triggering context is not within the same planning window as the target of a process, it cannot apply. The size of the speech planning window is variable, and has been shown to be influenced by many of the factors associated with phonological variation. The predictions of the PPH are tested in case studies of three different external sandhi processes. These studies also contribute to a general understanding of the relationship between variability and syntactic, prosodic, and lexical factors. The first study investigates the effect of word boundaries, prosodic position, and phonetic pauses on variation of high vowel devoicing (HVD) in Tokyo Japanese. Statistical modeling of HVD patterns in a corpus of spontaneous speech suggests that these factors jointly affect HVD, and that position within a prosodic phrase modulates the effect of speech rate and lexical frequency on HVD. It is proposed that two distinct processes underlie HVD, one that is sensitive to the segmental content of the upcoming word (interconsonantal HVD), and another that is triggered by strong prosodic boundaries (phrase-final HVD). Under this view, part of the variation can be explained under the PPH as production planning effects on interconsonantal HVD. In a second case study, two factors previously associated with external sandhi variation are tested directly: syntactic structure, and lexical frequency. Both of these factors have also been shown to affect speech planning and therefore, according to the PPH, should affect external sandhi. In a production experiment, we examine the effect of a clause boundary on realization of word-final coronal stops in North American English. Clause boundaries are found to have a gradient inhibitory effect on flapping, beyond the effect of associated final lengthening. The PPH explanation for this effect is that a clause boundary induces a delay while high-level planning of the next clause takes place, so segmental details of a potentially flap-triggering word will rarely be available. In contrast, a word that is within the same clause is much more like to be planned within the same window. Lexical frequency can also affect the time course of speech planning. Higher lexical frequency is associated with faster retrieval, so the PPH predicts that the realization of a word-final coronal stop will be related to the frequency of the word that follows it. A higher frequency following word will be retrieved more quickly, and be more likely to trigger flapping. The relationship between lexical frequency and coronal stop realization is examined in a corpus of American English, and a positive correlation is found between frequency and flapping, as predicted. The third case study extends testing of PPH predictions to a non-reductive external sandhi process: liaison in French. Frequency and also predictability are used as proxies for upcoming word availability. The PPH predicts that the correlations should be positive, just as for flapping, since both flapping and liaison rely on the knowledge that the upcoming word starts with a vowel. Examination of two syntactic contexts suggests that increased following word predictability, measured by both local (conditional probability) and global (lexical frequency), increase the likelihood of liaison application. Finding the same effect for the qualitatively distinct processes of flapping and liaison lends support to the PPH proposal that accessibility of the word containing the triggering information for sandhi constrains application of the process. The PPH offers a unified account of variation in external sandhi related to both grammatical and non-grammatical factors. In addition, the PPH makes many new, testable predictions for future work: the size of the window for phonological encoding should be correlated with the application of external sandhi, with less sandhi applying when the window is more narrow.
Format: [ pdf ]
Reference: lingbuzz/003601
(please use that when you cite this article)
Published in: PhD Thesis, McGill University
keywords: speech production planning, corpus phonology, phonological variation, french liaison, high vowel devoicing, japanese, flapping, external sandhi, phonology
Downloaded:963 times


[ edit this article | back to article list ]