Envelope-Based Metrics Reveal Rhythmic Development Tendencies in Brazilian L2 English Learners

Speech rhythm in second language (L2) development remains underexplored, despite previous reports indicating its impact on intelligibility, comprehensibility, and perceived degree of foreign accent (Munro & Derwing, 2001; Silva Jr. & Barbosa, 2019). While most studies rely on interval-based rhythm metrics, envelope-based approaches such as spectral analysis of the envelope (Tilsen & Johnson, 2008) and Empirical Mode Decomposition (EMD; Tilsen & Arvaniti, 2013) provide alternative tools to capture rhythmic periodicities in the amplitude envelope of speech. This study investigates speech rhythm development in Brazilian learners of English (EN-L2), comparing them with monolingual native speakers of English (EN-L1) and of Brazilian Portuguese (BP-L1). We analyzed 1,060 speech chunks (1–2.5s) extracted from controlled readings. Filtered vocalic energy envelopes were extracted via Voice Activity Detection and decomposed into Intrinsic Mode Functions (IMFs). Three envelope-based rhythm metrics were analyzed: Spectral Band Power Ratio (sbpr₁), Spectral Centroid (scntr₁), and the IMF₂/IMF₁ power ratio (imf_ratio₂₁). Mixed-effects models assessed the influence of language and chunk duration while accounting for speaker variability. Contrary to expectations, results indicated no significant cross-linguistic differences between BP-L1 and EN-L1. The most robust effect was observed in L2 learners, who gradually amplified the stress-level timescale over time. By the latest stage, learners exhibited rhythm patterns even more stress-dominant than either L1 baseline group, consistent with prosodic hyper-articulation (Cho, Lee & Kim, 2011; Smiljanić & Bradlow, 2009). A key contribution of this study lies in the analysis of between-group differences in correlations between envelope-based rhythm metrics. This relational analysis reveals that while L1 speakers exhibit relatively stable patterns of spectral energy coordination, L2 learners show evolving patterns of co-variation that complement the analysis of groupwise differences observed in individual metric values. These results point to the suitability of envelope-based approaches to the study of L2 rhythm development, suggesting that learners undergo a non-linear process of adjustment and redistribution of acoustic energy along different timescales.