Dialectal and generational variations in vowels in spontaneous speech

Robert Allen Fox, Ewa JacewiczInterspeech 2012

TL;DR

This paper investigates how American English vowels are changing across generations and dialects, focusing on spontaneous speech. Researchers found evidence of ongoing sound changes, not just in vowel positions but also in their spectral dynamics, suggesting that sound change is a complex process involving more than just static vowel pronunciations.

What Problem Does This Paper Address?

Linguists have long studied sound change, but the methods have evolved. Historically, analysis relied on historical texts. More recently, sociolinguists have incorporated social factors, while phoneticians have focused on controlled experimental settings. This paper bridges these traditions by examining sound change in spontaneous speech, a more naturalistic but less controlled environment. The core challenge is understanding how regional dialectal variations in vowels are evolving between older and middle-aged generations, particularly in the context of established sound change patterns like the Northern Cities Shift (NCS). The authors aim to confirm previously observed changes in citation-form and read speech using spontaneous speech data, while also exploring the role of spectral dynamics (how vowel sounds change over time) in these shifts.

Methods

The researchers analyzed acoustic properties of six lax vowels (/ɪ, ɛ, æ, ɑ, ɔ, ʌ/) from spontaneous speech recordings of 36 adult female speakers from southeastern Wisconsin (WI) and central Ohio (OH). These speakers were divided into two generations: older (68-90 years old) and middle-aged (35-50 years old).

The key analytical techniques involved:

  • Acoustic Measurement: Formant frequencies (F1 and F2) were measured at five equidistant time points within each vowel token.
  • Positional Analysis: Mean formant values at the midpoint (50%) of the vowel were plotted to visualize vowel positions in acoustic space, allowing for comparison of dialectal and generational differences.
  • Spectral Dynamics Measurement: A measure called "trajectory length" (TL) was calculated. This quantifies the total formant movement over specific vowel sections (e.g., 20-35% of duration) and overall, to assess the dynamic changes within a vowel. This is a novel aspect compared to traditional analyses that often focus only on static midpoint values.

Datasets

The study utilized a corpus of spontaneous speech recordings collected from 36 adult female speakers (18 from southeastern WI, 18 from central OH) across two generations. The recordings consisted of 5-10 minute talks or conversations on various personal topics. This dataset is a continuation of previous work by the authors, building on existing corpora of regional American English.

Key Insights and Ideas

  • Spontaneous Speech Confirms Sound Change: The study successfully confirmed previously observed dialectal and generational vowel changes in spontaneous speech, validating its use for studying sound change.
  • Beyond Static Vowel Positions: Sound change isn't just about where a vowel "lands" in acoustic space (its midpoint), but also how it gets there. The spectral dynamics (formant movement) are also changing, particularly in younger generations.
  • Regional Differences Persist: While both regions show generational shifts, the nature of these changes differs. Central Ohio shows a new chain-like rotation of lax vowels, while southeastern Wisconsin exhibits elements of the Northern Cities Shift but with some key differences, suggesting ongoing dialectal divergence.
  • Spectral Dynamics Reduction: Younger speakers generally exhibit reduced formant movement (lower TL) in front vowels, consistent with previous findings in read speech. This suggests a potential simplification or regularization of vowel production over generations.
  • Challenges of Spontaneous Speech Analysis: The authors highlight the inherent difficulties in analyzing spontaneous speech, such as the loss of fine experimental control over stress and phonetic context, which can complicate the interpretation of acoustic data.

Key Results

  • Central Ohio: Middle-aged speakers (OH_A2) showed positional changes in vowels like /æ/, /ɑ/, and /ɔ/, indicating a new sound change, potentially a merger of /ɑ/ and /ɔ/, and fronting of /ɪ/ and /ɛ/.
  • Southeastern Wisconsin: While older speakers (WI_A4) showed some NCS-related changes, younger speakers (WI_A2) exhibited more drastic positional shifts, including retraction of most vowels, but notably, the characteristic fronting and raising of /ɑ/ (a key NCS feature) was absent, and /a/ appeared to be backing and raising towards /ɔ/.
  • Spectral Dynamics: A statistically significant reduction in trajectory length (TL) was found for the vowel /æ/ between older and middle-aged speakers across both regions, with WI_A2 showing the most drastic change. Reductions were also observed in /ɪ/ and /ɛ/, though less consistently. Interestingly, /ɔ/ showed an increase in formant movement in younger speakers, along with a directional change, possibly related to the low back merger.

Limitations and Future Work

  • Experimental Control: The authors acknowledge the significant challenge of maintaining experimental control in spontaneous speech, making it harder to isolate generational changes from variations in stress, speaking rate, and consonantal context. This can lead to more subtle acoustic differences that are harder to interpret.
  • Statistical Significance: Due to the variability in spontaneous speech, achieving statistical significance for subtle differences in spectral dynamics can be difficult, requiring further research and more sophisticated statistical modeling.
  • Future Directions: The paper suggests that future research should explore larger speech units (syllables, words) to better characterize dynamic information. Further work is needed to understand the precise mechanisms driving these changes and the role of perceptual factors. The authors also note that their findings for female speakers are consistent with patterns observed in male speakers and will be reported elsewhere.

Summary updated

BibTeX
@inproceedings{fox2012dialectal,
  author = {Robert Allen Fox and Ewa Jacewicz},
  title = {{Dialectal and generational variations in vowels in spontaneous speech}},
  booktitle = {Interspeech},
  year = {2012},
}
Figure