The effect of accent on acoustic cues to stop voicing and place of articulation in radio news speech

Heejin Kim, Jennifer Cole, Hangseok Choi, Mark Hasegawa‐Johnson2004 · DOI

TL;DR

This paper investigates how phrasal accent affects the acoustic cues that distinguish between voiced/voiceless stops and stops made at different places of articulation. They found that accent enhances voicing contrasts through "strengthening" but has a more uniform, less contrast-enhancing effect on place of articulation cues.

What Problem Does This Paper Address?

Previous research on how stress and accent influence speech sounds has largely relied on highly controlled laboratory speech. This leaves a gap in understanding how these prosodic features affect pronunciation in more natural, unscripted speech, where speakers have more freedom in choosing phrasing and prominence. Specifically, this study aims to uncover how phrasal accent impacts the acoustic cues for two key phonological distinctions: voicing (e.g., /p/ vs. /b/) and place of articulation (e.g., /p/ vs. /t/).

Methods

The researchers analyzed speech from four announcers in the Boston University Radio News corpus, specifically focusing on naturally occurring "lab-news" recordings. They extracted stop consonants (/p, b, t, d, k, g/) from medial positions within intonational phrases and word-initial, pre-vocalic positions. These tokens were categorized as either "accented" or "unaccented" based on hand-labeled prosodic markings.

Four acoustic features were measured for each stop consonant:

  • Voice Onset Time (VOT): The time between the release of a stop consonant and the onset of vocal fold vibration.
  • Fundamental Frequency (F0): The basic pitch of the voice.
  • Closure Duration (CD): The length of time the vocal tract is closed during the stop consonant.
  • Burst Amplitude (BA): The intensity of the burst of noise released at the end of a stop consonant.

To analyze the effects of accent, they employed a 2-way ANOVA, examining interactions between "Accent" (accented vs. unaccented) and the phonological categories (Voicing or Place of Articulation). Significant interactions were interpreted as evidence of "paradigmatic strengthening" (where contrast is enhanced by opposite changes in acoustic cues) or "combined strengthening" (where both categories change in the same direction but to different degrees, also enhancing contrast). A lack of significant interaction was interpreted as "syntagmatic strengthening" (a uniform change across categories, not necessarily enhancing contrast).

Datasets

The study utilized the lab-news portion of the Boston University Radio News Corpus. This corpus consists of news stories read by professional announcers in a radio style, recorded in a studio. The data was hand-labeled with prosodic markings following the ToBI standard. The researchers analyzed data from four specific speakers (F3A, F2B, M1B, M2B) out of six available.

Key Insights and Ideas

  • Accent Matters in Natural Speech: The study confirms that phrasal accent significantly influences acoustic cues in naturalistic radio news speech, not just in controlled lab settings.
  • Voicing Contrasts Get a Boost: Phrasal accent appears to actively enhance the acoustic differences between voiced and voiceless stops. This is achieved through "strengthening" mechanisms that make the contrast more pronounced.
  • Place of Articulation Cues are More Robust: While accent does affect cues for place of articulation, the effect is more uniform across different places. This suggests accent's primary role might not be to sharpen place distinctions, but rather to signal prominence.
  • Paradigmatic vs. Syntagmatic Strengthening: The paper introduces and applies the concepts of paradigmatic and syntagmatic strengthening to explain how accent affects acoustic cues. Paradigmatic strengthening, where cues change in opposite directions for contrasting sounds, was observed for voicing, leading to enhanced contrast. Syntagmatic strengthening, a more uniform change, was more characteristic of place of articulation cues.

Key Results

  • Voicing Cues:
    • Significant effects of accent were found on VOT, F0, and CD for most speakers.
  • VOT and F0 were generally greater for voiceless stops than voiced stops.
  • CD was generally higher for voiced stops.
  • Accent typically increased VOT and F0 for both voiced and voiceless stops, often with a greater increase for voiceless stops, enhancing the contrast.
  • CD also increased under accent, sometimes more for voiced stops, contributing to contrast enhancement.
  • In some cases (F2B, M1B for VOT), accent caused opposite changes (increase for voiceless, decrease for voiced), a clear example of paradigmatic strengthening.
  • Place of Articulation (POA) Cues:
    • Significant effects of accent were found on VOT and CD for all speakers.
  • VOT, CD, and BA showed significant effects of POA for all speakers.
  • Unlike voicing, interactions between Accent and POA were generally not significant, indicating a more uniform effect of accent across different places of articulation. This points towards syntagmatic strengthening as the dominant pattern for POA cues.

Limitations and Future Work

  • Limited Speakers: The study analyzed only four speakers, which may limit the generalizability of the findings.
  • Specific Corpus: The findings are based on radio news speech, and it would be beneficial to investigate other genres of natural speech.
  • Focus on Stops: The analysis was limited to stop consonants. Future work could explore the effects of accent on other consonant types and vowels.
  • Further Investigation of Interaction Patterns: While the paper categorizes effects as paradigmatic or syntagmatic, a deeper dive into the nuances of interaction patterns across all speakers and features could be valuable.
  • Broader Acoustic Features: Exploring additional acoustic features beyond the four examined could provide a more comprehensive understanding of accent's impact.

Summary updated

Abstract

In previous research evidence for the effects of stress and accent on phonetic variation is based on laboratory speech. In the present paper, we report on a study of the effects of accent on the acoustic cues for stop voicing and place of articulation in the speech of four announcers from the Boston University Radio News corpus. The results show that there are significant effects of accent on VOT, F 0 and Closure Duration for voicing contrasts and significant effects on VOT and Closure Duration for place contrasts. In addition, comparison of the patterns of accentual effects reveals that the effect on voicing cues has a pattern of paradigmatic strengthening and combined strengthening, resulting in enhancement of voicing contrasts while syntagmatic strengthening appears to be the main effect on acoustic cues for place of articulation.
BibTeX
@misc{kim2004effect,
  author = {Heejin Kim and Jennifer Cole and Hangseok Choi and Mark Hasegawa‐Johnson},
  title = {{The effect of accent on acoustic cues to stop voicing and place of articulation in radio news speech}},
  year = {2004},
  doi = {10.21437/speechprosody.2004-6},
}
Figure