This study explored the initial stages of adult second language (L2) learning with a special focus on the acquisition of the target language sound system. The aim was to analyze the longitudinal development of Spanish stop voicing contrasts in an immersion learning context. Native English‐speaking late learners of Spanish provided production data for Spanish stops on a weekly basis throughout a domestic immersion program. The data were analyzed using Bayesian multilevel models and generalized additive mixed models. The analyses revealed phonetic learning in voiced stops over the course of the immersion program, though learners’ production was inconsistent and did not fall within native ranges for most stop segments. The results suggest that L2 phonetic category formation can occur at an early stage of development and follows a nonlinear trajectory that is subject to individual and segment‐specific variability. Furthermore, L2 phonetic representations are unstable during the early stages of learning.

中文翻译：

---

精细语音细节的纵向发展：在家庭浸泡程序中停止生产

这项研究探索了成人第二语言（L2）学习的初始阶段，并特别关注目标语言声音系统的获取。目的是在沉浸式学习环境中分析西班牙语停止语音对比的纵向发展。西班牙语为母语的英语为母语的后期学习者，在整个家庭沉浸程序中，每周都会提供西班牙语站点的生产数据。使用贝叶斯多级模型和广义加性混合模型分析数据。分析显示，在浸入式程序的整个过程中，在语音停止中进行语音学习，尽管学习者的产生不一致，并且在大多数停止阶段都没有落在本机范围内。结果表明，L2语音类别的形成可以发生在开发的早期阶段，并且遵循非线性轨迹，该轨迹受个体和段特定的可变性的影响。此外，L2语音表示在学习的早期阶段是不稳定的。