| dc.description.abstract | This thesis provides an account of the phonology and phonetics of intonation in Derry City English using the Autosegmental-Metrical (AM) framework (Ladd, 2008). It focuses on the phonological inventory and phonetic features of pitch events in different metrical and lexical contexts and in different sentence modes. It investigates if, despite the dominance of nuclear rises (L*H) as opposed to the high pitch accents (H*) found in other varieties of English, there is evidence for a special status of the H tone in DCE. It looks for evidence of a phonological register tier (a register tier hypothesis) which nIE speakers exploit to show phonological contrasts indicated in other English varieties via falling and rising nuclear contours.
A corpus of read speech was collected from 11 speakers of Derry City English (6F, 5M, mean age = 40, SD 9.9). All utterances (n = 1427) were analysed and annotated in Praat (Boersma and Weenink, 2002) following the IViE labelling guide (Grabe 2001) with a few adjustments. The f0 contour and annotations for each utterance were processed to generate data tables for all subsequent phonological and phonetic analyses. Statistical analysis was conducted in R (R Core team, 2022), using mixed-effects models (Bates et al., 2015; Chung & Rabe-Hesketh, 2013).
Two major analyses were conducted. The first looked at the intonation of unmarked declarative utterances under changes in anacrusis, foot size, and lexical boundaries. The second analysed intonation across sentence modes. A third short analysis using phonetic f0 turning points and pitch contour resynthesis (Rodgers 2020) was conducted to investigate some phenomena from the main analyses.
The analysis of metrical and lexical effects found that L*H was more likely in prenuclear (PN) position when the foot was longer or when the word-final syllable was later in the foot. H* was more common in shorter feet and when the stressed syllable was early in the foot. This indicated that the L tone was apt to be deleted when there was less syllabic space in the foot, suggesting that the H tone has a privileged status in the PN position. Nuclear pitch accents were resistant to lexical and metrical effects, and all were L*H. The phonetic analysis indicated that tonal targets in PN pitch accents were more vulnerable to metrical and lexical effects than those in nuclear pitch accents. Nuclear peak alignment was strongly affected by foot size; however, when analysed as a proportion of voicing in the final foot, it was remarkably stable. Stress clash effects were found, with the PN L*H rise being truncated to accommodate an upcoming nuclear L, and the nuclear L target being aligned slightly later. The nuclear L*H rise was subject both to compression and truncation depending on the foot size, with the two strategies working in tandem to maintain the proportional alignment of the peak relative to the voiced material in the foot. The availability of a truncation strategy to maintain the prenuclear H target while the prenuclear L target was subject to deletion offered more evidence of the special status of the H tone.
The phonological analysis of intonation and mode used two labelling systems, one with a register tier and one without. BGLMM analyses using both labelling systems indicated the register-tier analysis was superior to a non-register tier analysis, and that high register in nuclear PA was most likely in declarative questions, especially among male speakers. Nuclear pitch-accent-only IPs were likely in yes-no and declarative questions, especially among female speakers. The likelihood of either high nuclear register or nuclear-PA-only IPs was very high in declarative questions. A dual strategy was employed for the phonetic analysis, using LME models with and without register tier effects. Differences in f0 scaling as an effect of mode were noticeably smaller in models with register tier effects, but even with the register tier analysis, paralinguistic effects of mode were still evident. f0 scaling of low register L and H targets and high register L and H targets was well distinguished, further supporting the register tier hypothesis.
The Phonetics-first analysis provided evidence that an ambiguous >H* PA can be both an H* with a long plateau or an L*H in which the L has been suppressed, that downstepped !H* pitch accents may require anticipatory lowering (L_), and that apparently earlier tonal alignment of nuclear L targets in declarative questions may be a side effect of physiological constraints on the vocal folds.
The research reported here contributes both to the body of knowledge on the intonation of northern Irish English and to the description of intonation as conceptualised within the AM approach. It is hoped that the investigation into the possibility of a phonological register tier, even if not confirmatory, can help shed light on some otherwise challenging intonational phenomena. | en |
