Speech rate effects on linguistic change

Alexsandro R. Meireles, Plínio A. BarbosaInterspeech 2009

TL;DR

This paper investigates how speaking speed influences linguistic change, specifically focusing on the reduction of unstressed vowels in Brazilian Portuguese. The authors found that faster speech rates lead to greater vowel reduction, which can eventually alter how words are perceived and even change their underlying structure over time. This work highlights the dynamic interplay between articulatory phonetics and language evolution.

What Problem Does This Paper Address?

For decades, linguistics has grappled with understanding how languages change. While historical linguistics has long studied these shifts, the precise mechanisms driving them, especially at the phonetic level, have been debated. Traditional linguistic theories often treated language as a static system, overlooking the role of real-time speech production. This paper addresses this gap by exploring how variations in speech rate, a fundamental aspect of spoken language, can act as a catalyst for linguistic change. Specifically, it examines the phenomenon of vowel reduction in Brazilian Portuguese, where unstressed vowels in words with antepenultimate stress (stress on the third-to-last syllable) can be reduced or even disappear, leading to a perception of penultimate stress (stress on the second-to-last syllable). The authors aim to demonstrate that increased speech rate is a key factor driving this process, potentially explaining historical shifts in word stress and structure.

Methods

The study employs an Articulatory Phonology (AP) framework, which views speech as a dynamic system of gestures. The core methodology involves analyzing speech recordings of Brazilian Portuguese speakers under three controlled conditions: normal, slow, and fast speech rates.

Key techniques include:

  • Semi-automatic segmentation and analysis: Using the Praat software, researchers segmented speech into vowels, consonants, and syllables, and measured their durations.
  • Speech rate calculation: Speech rate was quantified by dividing sentence duration by the number of phonetic syllables, yielding syllables per second.
  • Statistical analysis: One-way ANOVAs were used to detect significant differences in speech rates among speakers, dialects, and genders, as well as to compare vowel durations. A Friedman non-parametric ANOVA was employed to analyze the rate of duration decrease for vowels in different stress positions.
  • Articulatory Phonology interpretation: The findings are interpreted through the lens of AP, focusing on gestural overlap. Increased speech rate is hypothesized to lead to greater temporal overlap between articulatory gestures, resulting in vowel reduction and potential perceptual shifts in word stress.

Datasets

The study utilized a custom-built database consisting of six specific words: "abóbora" (pumpkin), "análise" (analysis), "cócega" (tickle), "fôlego" (breath), "fósforo" (phosphorus), and "sábado" (Saturday). These words were embedded in the carrier sentence "Digo baixinho" (I say quietly). Each sentence was recorded ten times by each speaker at each of the three speech rates (normal, slow, fast). This dataset was designed to include words already undergoing variation (e.g., "abóbora" to "abobra") and words without expected variation, as well as varying word lengths (three- and four-syllable words) to observe their effects on lexical reorganization.

Key Insights and Ideas

  • Speech rate as a driver of linguistic change: The study provides strong evidence that increased speech rate is not just a stylistic variation but a potential engine for diachronic linguistic change, specifically in the realm of vowel reduction and stress shifts.
  • Gestural overlap explains reduction: The AP framework offers an elegant explanation: faster speech leads to increased temporal overlap of articulatory gestures, causing unstressed vowels to shorten and reduce, which can eventually lead to their perceptual loss.
  • Dialectal and gender influences: Linguistic change is not uniform. The study reveals that speakers from different dialects (Mineiro vs. Paulista) and genders exhibit distinct speech rates and reduction patterns, suggesting these sociolinguistic factors modulate the impact of speech rate.
  • Word-specific reduction patterns: Not all words are affected equally. The research highlights that certain words (like "abóbora") show greater reduction in medial post-stressed vowels compared to others (like "fôlego"), which helps explain why some historical changes are more common than others.
  • Beyond discrete segments: The paper advocates for dynamic systems approaches to linguistics, moving beyond static, discrete phonetic segments to understand the continuous, fluid nature of speech and its evolution.

Key Results

  • Faster speakers, faster reduction: Speakers with higher average speech rates (syllables per second) were more likely to produce reduced lexical forms, suggesting they are more prone to linguistic change.
  • Mineiro speakers are faster: Speakers from the Mineiro dialect (average 7.5 syllables/sec) spoke faster than those from the Paulista dialect (average 6.4 syllables/sec).
  • Men speak faster than women: Male speakers (average 5.8 syllables/sec) spoke faster than female speakers (average 5.2 syllables/sec).
  • Vowel reduction is key: The study found that the medial post-stressed vowel (MPS) tends to reduce more than the final post-stressed vowel (FPS) at faster speech rates, particularly in words like "abóbora" and "cócega." This differential reduction is hypothesized to be the mechanism behind the perceived shift from antepenultimate to penultimate stress.
  • Quantitative evidence of reduction: Table 1 shows that for a specific speaker (DF, MG), the MPS vowel duration decreased by up to 54% (for "abóbora") from normal to fast rates, while the FPS vowel decreased by only 22% in the same word.

Limitations and Future Work

  • Perceptual studies needed: While the study provides strong evidence for articulatory changes driven by speech rate, the authors acknowledge that perception studies are crucial. Future work should investigate how listeners reinterpret these reduced forms and how this perceptual shift contributes to actual linguistic change.
  • Individual variation: While dialect and gender were explored, the authors note that individual factors also play a role in speech rate and reduction strategies, suggesting further research into these nuances.
  • Generalizability: The study focused on a limited set of words and a specific dialect of Brazilian Portuguese. Further research with a broader range of linguistic phenomena and languages would be beneficial to confirm these findings.
  • Consonant vs. Vowel compressibility: The paper suggests consonants are less compressible than vowels, but a more in-depth analysis of consonant reduction and its role in change could be a future direction.

Summary updated

BibTeX
@inproceedings{meireles2009speech,
  author = {Alexsandro R. Meireles and Plínio A. Barbosa},
  title = {{Speech rate effects on linguistic change}},
  booktitle = {Interspeech},
  year = {2009},
}
Figure