README.md

A-tradeoff-between-acoustic-and-linguistic-feature-encoding-

Codes used in the paper "A tradeoff between acoustic and linguistic feature encoding in spoken language comprehension" eLife2023;12:e82386 DOI: https://doi.org/10.7554/eLife.82386

• Generate .csv files that contains word and phoneme onsets of stories

1. Run WebMaus forced aligner to generate .TextGrid files
2. Run extract_phonemes_timing.py
3. Run extract_word_timing.py
4. Run word_phoneme_bridge.py

• Generate Cohor Model, add word frequencies and word entropy by using GPT2 model

1. Run Clean_freq_file.py for both Dutch and French
2. Run Cohort_model.py for both Dutch and French
3. Run Word_entropy_GPT.py for both Dutch and French
4. Manually correct for the missing word frequency values by replacing NaN values with the mean frequency
5. Run Word_entropy_low_vs_high.py
6. Run Generate_High_Low_entropy_cont_arrays.py

• Generate Predictors

1. Run make_gammatone.py
2. Run make_gammatone_predictors.py
3. Run make_word_predictors.py

• Generate TRF models

1. Run estimate_trfs.py from Scripts_for_publication/TRFs/Generate_TRF_models/sub001/ (In utils_TRF.py set Sources_saved = False for the first time you run the TRF model so it saves the source reconstructed signal for the next TRF models. Then set it to True)

• Accuracy Analysis

1. Run Whole_brain_accuracies_basic_models.py, Word_entropy_effect_acoustic_features.py and Word_entropy_effect_phoneme_features.py to generate .csv files.
2. Run LMM in R for further analysis

• TRF Weights Statistical Analysis

1. Run ANOVA_weigths_all_features.py and ANOVA_weigths_phoneme_features.py for Part 1 and Part 2.
2. Run Visualize_brains_all_phoneme_features.py to generate contrast on brain surface and the statistical analysis report.

• Generate TRF graphs

1. Run TRF_LH.py and TRF_RH.py for Part 1 and Part 2