Temperature and Language Sonority

Code and data for Temperature shapes language sonority: Revalidation from a large dataset.

Usage

The following 5 steps can be run separately as the output of each step is already provided in this repository (see Data below). Steps 1 and 2 require a local storage of the ASJP dataset and the FLDAS dataset, but you can skip these two steps so you do not need to download full datasets.

1. Extract geometry and sonority data from ASJP

Run python get_sonority.py [raw_path], where [raw_path] is the path to raw folder in the local ASJP dataset (e.g. python get_sonority.py C:/ASJP/raw/). Results will be saved as sonorities.csv, phones.csv, word_structures.csv, and word_lengths.csv in the data folder.

2. Extract temperature data from FLDAS

Run python get_temperature.py [FLDAS_path] to extract monthly temperature data of all doculects in sonorities.csv, where [FLDAS_path] is the path to FLDAS_NOAH01_C_GL_M.001 folder of the local FLDAS dataset (e.g. python get_temperature.py C:/FLDAS/FLDAS_NOAH01_C_GL_M.001/). Results will be saved as data/temperatures.csv.

Run python get_temperature_global.py [FLDAS_path] to extract global monthly mean temperature data. Result will be saved as temperature_global.csv.

3. Plot global distribution of temperature and sonority

Run python plot_global.py. Plot will be saved as figure/global.png.

4. Combine and process temperature and linguistic data

Run python process.py. Results will be saved as data.csv, data_genus.csv, data_family.csv, and data_macroarea.csv in the data folder.

5. Generate distribution and correlation plots, and more

Run corresponding code blocks in process.r in R.

6. Compare vowel length solutions

Run python test_vowel_length_solutions.py [raw_path]. Results will be saved as data/vowel_length_solutions.csv. Then, run code block of “Plot correlations between vowel length solutions” in process.r to plot correlations.

Data

All extracted data files are in the data folder.

Temperature Data

• temperatures.csv: Monthly temperature (1982–2022) for each filtered doculect
• temperature_global.csv: Global mean annual temperature over 41 years (180° W–180° E, 60° S–90° N)

Linguistic Data

• sonorities.csv: Mean sonority index (MSI) of each filtered doculect. We adapted 5 methods to calculate MSI from ASJP codes:
  • index0: Parker’s scale, from Sonority in The Blackwell Companion to Phonology
  • index1: Fought’s scale, from Sonority and climate in a world sample of languages: Findings and prospects
  • index2: List’s scale, from Sequence Comparison in Historical Linguistics
  • index3: Clements’s scale, from The role of the sonority cycle in core syllabification in Papers in Laboratory Phonology
  • index4: Sonorant index (here obstruent = 1; sonorant = 2)
  • index5: Vowel index (here consonant = 1; semivowel = 2; vowel = 3)
  • index6: List’s scale, calculated using LingPy tokens2class()
• phones.csv: Extracted phones from all doculects
• word_structures.csv: Word structures statistics of all doculects, characterized by C (= consonant) and V (= vowel) symbols
• word_structures_grouped.csv: Word lengths statistics of all doculects
• vowel_length_solutions.csv: MSI results under three vowel length solutions

Combined Data

• data.csv: Data for each filtered doculect, with temperature data and linguistic data combined
  • WL: Mean word length
  • Index0 to Index6: MSIs in 7 methods
  • T: Mean annual temperature
  • T_max: Max of 41-year mean monthly temperatures
  • T_min: Min of 41-year mean monthly temperatures
  • T_sd: Standard deviation of monthly temperatures over 41 years
  • T_diff: Mean annual range of temperature
  • Index0_trans, etc.: Transformated above data
• data_genus.csv: Data for each language “genus” classified by WALS
• data_family.csv: Data for each language family classified by WALS
• data_macroarea.csv: Data for each macroarea (North America, South America, Eurasia, Africa, Greater New Guinea, and Australia)

Figures

All saved figure files are in the figures folder.

• global.png (also converted into global.pdf): Global distribution of MATs and MSIs
• distribution.pdf: Distribution of MATs and MSIs grouped by macroarea
• correlation.pdf: Relationship between MSI and MAT
• correlation_by_family.pdf: Relationship between MSI and MAT of the top 25 largest families
• word_length.pdf: Relationship between mean word length and MSI or MAT
• word_length_by_family.pdf: Relationship between MSI and mean word length of the top 25 largest families
• range.pdf: Relationship between mean annual range and MAT
• vowel_length_solutions.pdf: Relationship between vowel length solutions