earticle

영어

This study investigates how Korean EFL learners produce English prosodic stress, focusing on the contrast between compound nouns and adjective–noun phrases. While traditional acoustic analyses have revealed persistent difficulties in L2 stress acquisition, they often fail to account for how learners integrate multiple acoustic cues in real-time speech production. To address this limitation, the current study integrates two complementary inferential approaches: traditional group comparisons (e.g., t-tests) and logistic regression modeling for predictive classification. twenty-six Korean university students were recorded producing compound–phrasal minimal pairs. Measurements of vowel duration, pitch (F0), and intensity were extracted and compared across stress types. In addition to identifying significant group-level differences, a logistic regression model was used to evaluate whether learners’ combined acoustic cues could reliably predict their intended stress pattern. Results show that while learners exhibited partial differentiation—most notably through intensity—they often failed to coordinate cues systematically. The regression model revealed that only some learner productions could be accurately classified based on their acoustic cues, indicating that many learners did not consistently coordinate duration, pitch, and intensity to mark stress. Particularly, intensity emerged as the most significant cue in group-level comparisons, while logistic regression complemented this by revealing how learners inconsistently integrated cues to signal intended stress. These findings may inform methodological approaches in L2 prosody research and suggest potential implications for pronunciation instruction and the development of AI-assisted feedback tools targeting suprasegmental features.