This project proposes a run-wise analysis of the relationship between spoken-audio amplitude and brain activity during Le Petit Prince listening in Chinese-, English-, and French-language fMRI data. Preprocessed functional images will be paired with the corresponding audio sections. For each run, short-time root-mean-square (RMS) audio amplitude will be sampled at the fMRI repetition time, standardized, and convolved with a hemodynamic response function to create a predictor of audio intensity. A first-level general linear model will estimate the association between this predictor and BOLD signal while accounting for motion, censored volumes, and baseline trends. Design matrices will be generated with AFNI’s 3dDeconvolve, and models will be fit using 3dREMLfit to account for temporal autocorrelation. The analysis will provide run-level estimates of audio-amplitude-related activity and residual time series for quality assessment and potential subsequent analyses. Before interpreting results, the audio-to-run mapping, synchronization offsets, censoring conventions, and regressor construction will be verified. No inferential results are reported here; the workflow is a proposed analysis pipeline.