Dynamic formant trajectories in L2 English: evidence from Arabic-speaking adolescent learners

• L2 English learners show reduced formant dynamics and slower spectral transitions. • Trajectory measures (VL, TL, FV) and angular displacement index spectral–temporal reduction. • fPCA reveals system-wide vowel space compression and F1–F2 modulation decline. • Trajectory-based cues outperform static midpoints in capturing within-speaker variation. • Findings support the SLM-r: L2 reorganization affects dynamic structure, not just targets. Previous research on Arabic-speaking learners of English (L2) has often focused on adults and relied on static formant measurements, underscoring the need for dynamic approaches that better capture the temporal characteristics of L2 vowel production. Grounded in the revised Speech Learning Model (SLM-r), this study analyzes the formant trajectories of seven English vowels in CVC contexts, produced by 65 adolescent female learners whose L1 is Palestinian Arabic. Static vowel targets were descriptively compared to normative values from a corpus of L1 English speakers (N = 436; Lee et al., 1999), with age, gender, and CVC context matched across datasets; dynamic analyses were conducted on the L2 learner data only. Formant trajectories were modeled as multidimensional time series to assess Vowel-Inherent Spectral Change (VISC), using three trajectory-based measures—Vector Length (VL), Trajectory Length (TL), and Formant Velocity (FV)—alongside angular displacement as an additional cue. A multi-method approach combining Linear Mixed-Effects modeling (LME), Functional Principal Component Analysis (fPCA), and trajectory analysis identified group-level patterns of vowel-space compression, slower formant transitions, and reduced spectral–temporal modulation. fPCA showed joint reductions in F1 and F2 dynamics, reflecting a centralized, flattened vowel space with limited temporal modulation and diminished contrast. These outcomes suggest system-level reorganization of the L2 vowel space, shaped by L1-based constraints. Trajectory orientation emerged as a secondary cue when integrated with other time-varying measures, with directional characteristics providing limited but detectable acoustic distinctiveness. Classification models comparing dynamic, static, combined, and direction-only features showed that time-resolved spectral cues most consistently supported within-speaker classification and were more sensitive to spectral variation. Trajectory and fPCA analyses extend the empirical scope of the SLM-r by highlighting that equivalence classification may involve not only reduced acoustic category separation but also a reorganization of the spectral–temporal geometry of the L2 vowel system. These findings offer a signal-level perspective on vowel restructuring as it unfolds acoustically in underrepresented L2 learners and broaden our understanding of the acoustic structure of L2 vowels across diverse learning environments.