Scholarlink

<p>In late-closure garden-path sentences such as “<i>Whenever the boys hug Tom smiles happily</i>”, the locally ambiguous noun phrase (“<i>Tom</i>”) is initially analysed as the direct object of the embedded verb (“<i>hug</i>”) and is later revised as the subject of the matrix clause once the disambiguating verb (“<i>smiles</i>”) is encountered. The present study investigated whether active dependency formation affects the resolution of this ambiguity in three self-paced reading experiments. The experimental sentences contained wh-trace and/or PRO-controller dependencies, which were designed to introduce biases towards the subject analysis. The results showed that, although active dependency formation did not prevent the object analysis, it did reduce the revision cost at the disambiguating verb. This reduction was more pronounced when multiple unresolved dependencies were present than when only a single unresolved dependency was present. We discuss the implications of these findings for our understanding of how dependencies are established and how ambiguities are resolved during real-time sentence processing.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Active dependency formation and ambiguity resolution in real-time sentence processing

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月7日

查看全文
  • Hiroki Fujita,
  • Masaya Yoshida

摘要

In late-closure garden-path sentences such as “Whenever the boys hug Tom smiles happily”, the locally ambiguous noun phrase (“Tom”) is initially analysed as the direct object of the embedded verb (“hug”) and is later revised as the subject of the matrix clause once the disambiguating verb (“smiles”) is encountered. The present study investigated whether active dependency formation affects the resolution of this ambiguity in three self-paced reading experiments. The experimental sentences contained wh-trace and/or PRO-controller dependencies, which were designed to introduce biases towards the subject analysis. The results showed that, although active dependency formation did not prevent the object analysis, it did reduce the revision cost at the disambiguating verb. This reduction was more pronounced when multiple unresolved dependencies were present than when only a single unresolved dependency was present. We discuss the implications of these findings for our understanding of how dependencies are established and how ambiguities are resolved during real-time sentence processing.