The present study investigates the acquisition of the [±realis] features encoded in Chinese negation by L1-English and L1-Korean learners through the framework of the Feature Reassembly Hypothesis. Two modes of grammaticality judgment tests (written and aural) were administered to 90 L1-English and 92 L1-Korean learners ranging from elementary to advanced Chinese proficiency levels, along with 15 native controls. Our study's findings in general support the proposals of the Feature Reassembly Hypothesis. The results indicate that L2 learners with lower Chinese proficiency tend to correspond their L1 features encoded in negation with similar L2 items, leading to target-deviant performances at the early stages of learning. L1-English and L1-Korean learners with increased Chinese proficiency assemble the [+realis] and [−realis] features with mei and bu, respectively. However, they face difficulties in identifying the [±realis] features represented in certain licensing contexts. Based on the results, we argue that the Feature Reassembly Hypothesis is strengthened when variables that constrain feature detectability and reassembly, such as L1 influence, L2 proficiency, input frequency and consistency, and task modality, are considered.

本研究通过特征重组假说的框架，研究了母语为L1的英语和母语为L1的韩国学习者对中文否定编码的[±realis]特征的获取。对90名L1-英语和92名L1-韩语学习者（从基础水平到高级汉语水平）以及15个本地对照进行了两种语法判断测试（笔试和听觉）。我们的研究结果总体上支持“特征重组假说”的建议。结果表明，汉语水平较低的第二语言学习者倾向于将他们的第一语言特征与相似的第二语言项目进行否定对应，从而导致在学习的早期阶段表现出目标偏差。具有较高汉语水平的L1英语和L1韩国学生将[+ realis]和[−realis]功能组合为分别是mei和bu。但是，他们在识别某些许可上下文中表示的[±realis]功能时遇到困难。根据结果​​，我们认为，当考虑约束特征可检测性和重组的变量（例如L1影响，L2熟练程度，输入频率和一致性以及任务模态）时，特征重组假说就得到了加强。