Measuring basic tempo across languages and some implications for speech rhythm

Gertraud Fenk-Oczlon, August FenkInterspeech 2010

TL;DR

This paper proposes a novel, language-inherent metric for measuring "basic tempo" – the intrinsic speed of a language – by counting syllables per intonation unit. This metric, applied to 51 languages, reveals significant correlations with syllable complexity and traditional rhythm classifications, suggesting that syllable structure is a key determinant of perceived speech tempo.

What Problem Does This Paper Address?

The common perception that some languages are spoken faster than others has long been a subject of linguistic inquiry. However, accurately measuring this "basic tempo" in a way that reflects inherent linguistic properties, rather than just individual speaking habits or cultural norms, has been challenging. Existing metrics often rely on time-consuming chronometry and are susceptible to significant inter- and intra-individual variations in speech rate. This paper aims to bridge this gap by proposing a method that is less dependent on direct time measurements and more rooted in the structural characteristics of languages, allowing for large-scale cross-linguistic comparisons.

Methods

The authors propose a new metric for measuring basic tempo: the number of syllables per intonation unit. An "intonation unit" is defined as a simple declarative sentence, roughly corresponding to a breath group or a short utterance (around two seconds), which is considered a universally comparable rhythmic unit. The "pulse" within this unit is the syllable.

The procedure involved:

  1. Data Collection: Native speakers of 51 diverse languages translated 22 simple declarative sentences.
  2. Syllable Counting: Informants counted the number of syllables in their translated sentences.
  3. Syllable Complexity Measurement: The number of phonemes per syllable was also determined.

This approach bypasses direct time measurements, focusing instead on structural counts that are argued to be less affected by individual speaking styles.

Datasets

The study utilized a dataset of 51 languages from various language families (19 Indo-European, 32 Non-Indo-European). The core material consisted of 22 simple declarative sentences translated by native speakers. This sample is an expansion of a previous study involving 26 languages.

Key Insights and Ideas

  • Intrinsic Tempo is Linked to Syllable Structure: The paper strongly suggests that a language's "basic tempo" is not an arbitrary feature but is deeply intertwined with its syllable structure. Languages with simpler syllable structures tend to have more syllables per intonation unit, leading to a higher intrinsic tempo.
  • Tempo vs. Information Rate: A higher intrinsic tempo (more syllables per unit time) does not necessarily mean a higher information transmission rate. Languages with faster intrinsic tempos might achieve this by using simpler syllables, not by packing more semantic information per syllable.
  • Rethinking Rhythm Classes: The findings challenge the strict dichotomy of "stress-timed" vs. "syllable-timed" languages. Instead, these classifications might represent points on a continuum, with syllable structure playing a crucial role in determining a language's position.
  • Cross-Linguistic Robustness: The negative correlation between syllable complexity (phonemes per syllable) and the number of syllables per clause proved to be remarkably robust across different language samples and over time.

Key Results

  • Syllables per Clause Variation: The study found a considerable variation in the number of syllables per simple declarative sentence across the 51 languages, ranging from 4.64 (Thai) to 10.96 (Telugu).
  • Negative Correlation: A significant negative correlation (r = -0.73) was found between syllable complexity (phonemes per syllable) and the number of syllables per clause. This means languages with more complex syllables tend to have fewer syllables per sentence, and vice versa.
  • Correlations with Rhythm Metrics: The proposed metric showed significant correlations with established metrics for distinguishing stress-timed and syllable-timed languages (e.g., percentage of vowels and standard deviation of consonantal intervals). Languages traditionally classified as stress-timed tended to have a lower intrinsic tempo (fewer syllables per intonation unit) than syllable-timed or mora-timed languages.
  • Mora-timed languages (e.g., Japanese, Telugu, Hawaiian) exhibited the highest number of syllables per intonation unit, likely due to their simpler syllable structures.

Limitations and Future Work

  • Sample Representativeness: While the sample of 51 languages is extensive, it's noted that many informants were students, and the basic requirement was proficiency in English or German for translation. The representativeness of this specific informant pool for each language's full spectrum of speech might be a consideration.
  • Definition of Intonation Unit: The use of simple declarative sentences as a proxy for intonation units is a simplification. Real-world intonation units can be more complex and variable.
  • Future Work: The authors suggest that their findings have implications for a systemic typology of languages, emphasizing the role of syllable structure in understanding perceived differences in speech rhythm and tempo. Further exploration into how syllable structure interacts with prominence and inter-stress intervals is also implied.

Summary updated

BibTeX
@inproceedings{fenkoczlon2010measuring,
  author = {Gertraud Fenk-Oczlon and August Fenk},
  title = {{Measuring basic tempo across languages and some implications for speech rhythm}},
  booktitle = {Interspeech},
  year = {2010},
}

Loading related content...

Figure