Languages in China link climate, voice quality, and tone in a causal chain

The study explores the hypothesis that language sound systems are ecologically adaptive, focusing on the relationship between climate and tonal languages. Previous research suggested a correlation between desiccation and the absence of lexical tone, based on the premise that dry air compromises voice quality and constrains the use of phonemic tone. However, this hypothesis lacked direct evidence from natural speech analysis. This research aims to address this gap by investigating the causal chain linking humidity, voice quality, and the number of tones using a large, standardized Chinese speech database (YuBao). The YuBao database, covering 997 language varieties across diverse Chinese climates, provides a unique opportunity to test the hypothesis. The study’s importance lies in providing a direct experimental test of the full causal chain, moving beyond previous correlational analyses and offering insights into the ecological adaptation of human speech.

Prior research explored the correlation between ecological factors and language sound systems. Studies found links between reduced air pressure and ejectives, and between climate and sonority classes. Everett et al. (2015) suggested a statistical association between ambient desiccation and the absence of lexical tone, arguing that dry air affects vocal fold physiology, hindering precise pitch control needed for complex tone systems. While this idea garnered general agreement on language's ecological adaptability, debates arose concerning the specific hypothesis about desiccation and tonality. Critics questioned the magnitude of desiccation's impact on voice quality and the assumption that tonal languages inherently rely on precise laryngeal pitch control. Existing studies lacked sufficient direct evidence from natural speech data, limiting the conclusiveness of the findings, leading to discussions on the role of geographical distribution and language contact.

This study leverages the China’s Language Resources Protection Project (YuBao) database, a large-scale standardized collection of audio and video recordings of 997 language varieties across China. Each language is represented by approximately 1200 sound files, comprising a total of 1,174,686 recordings. The study utilizes continuous variables: humidity (obtained from the WheatA database), jitter and shimmer (acoustic measures of voice quality extracted using Praat), and the number of tones (obtained from the Chinese Academy of Social Sciences). Data was collected from one native male speaker aged 55–65 for each language variety. The researchers employed linear mixed-effects models for voice quality analysis (jitter and shimmer) and generalized linear mixed-effects models with a Poisson distribution for analyzing the number of tones. Linguistic family was included as a random effect to account for non-independence of data points within language families. Analyses were conducted using base R and the lme4 package. Separate analyses were also performed for each linguistic group to explore within-family patterns.

The analysis revealed a significant negative correlation between humidity and voice quality (jitter and shimmer), supporting the hypothesis that drier environments lead to poorer voice quality. Locations with lower humidity exhibited higher jitter and shimmer values. This effect was statistically significant (jitter: χ²=160.68, df=1, p<0.0001; shimmer: χ²=42.58, df=1, p<0.0001). While the overall analysis did not show a significant direct effect of jitter/shimmer on the number of tones, within-family regressions revealed significant effects within Sino-Tibetan and Austroasiatic language families. Sino-Tibetan languages with more tones exhibited better voice quality (lower jitter). The study also found a significant nonlinear relationship between humidity and the number of tones. Higher humidity was associated with a higher number of tones, particularly within Sino-Tibetan languages. The Altaic language group, characterized by non-tonal languages and high jitter/shimmer, was contrasted with other groups to strengthen the argument.

The findings provide strong support for the causal chain linking humidity, voice quality, and the number of tones, particularly within the Sino-Tibetan family. The results demonstrate that desiccation impacts voice quality in natural speech contexts, leading to reduced efficiency of vocal fold vibration, and this, in turn, seems to be related to a reduction in the number of tonal distinctions. This corroborates previous research suggesting an ecological link between climate and tonal features, while providing direct experimental evidence and addressing limitations of past studies. The influence of humidity appears more pronounced in drier conditions, whereas in humid regions, the effect of further humidification on vocal fold vibration may be less significant. The study also considers potential confounding factors such as syllable structure, arguing that the observed correlations are not simply artifacts of these factors. Historical contingencies such as language contact and migration patterns are also acknowledged, but the systematic effect of humidity is seen as a significant mediating factor.

This study provides the first robust empirical support for the causal chain linking humidity, voice quality, and the number of tones in languages. The use of a large, standardized speech database allowed for a direct test of the hypothesis, addressing previous methodological limitations. The findings confirm that drier climates impact vocal fold vibration efficiency, affecting tone systems through the mediation of voice quality. Future research could focus on expanding the dataset to include more extremely dry regions and investigating the role of humidity in global geo-phonetic correlations, requiring the development of global, high-quality speech databases. The study offers a compelling case study of how ecological factors can influence human behavior through physiological mechanisms.

The study’s sample is primarily focused on Chinese languages, limiting the generalizability of the findings to other language families and geographic regions. The focus on male speakers aged 55–65 may also limit the applicability of the findings to other demographic groups. While the study accounts for language family and considers potential confounding factors, other sociocultural or historical factors might influence the relationships observed. The limited number of languages from extremely dry regions could also affect the interpretation of the humidity-voice quality relationship. Further investigation with a more diverse and globally representative dataset is required for broader conclusions.