When people read temporarily ambiguous (“garden-path”) sentences, the forward movement of their eyes is often interrupted by regressions. These regressions are usually followed by rereading some portion of the previously read text. proposed the Selective Reanalysis Hypothesis (SRH), which proposed that readers regress to critical choice points in the syntactic phrase marker of garden-paths where misparses had occurred, and furthermore, then reanalyzed the syntactic structure to arrive at a correct parse in most cases. A considerable amount of more recent work, however, suggests that readers often do not derive a correct parse or interpretation from such sentences. If these more recent observations are accurate, perhaps rereading is not necessarily strategic, controlled, or predictable. The current study consists of two large-scale eye-tracking experiments designed specifically to examine where and how much people reread garden-path sentences, and whether rereading influences comprehension accuracy. A variable text-masking paradigm was employed to restrict access to portions of garden-paths and non-garden-paths during rereading. Scanpath analyses were used to determine whether some or all participants targeted syntactically critical parts of previously read text. Comprehension questions probed final interpretations. In short, readers often misinterpreted the garden-paths, and no rereading measures predicted better comprehension. Furthermore, scanpath analyses revealed considerable variation across and within readers; only small percentages of trials conformed to structurally-based predictions. Taken together, we fail to find support for structurally strategic rereading. We therefore propose that rereading of these sentences is more often “confirmatory” than “revisionary” in nature.

中文翻译：

---

沿着花园小路重走：非选择性重读，不重新分析

当人们阅读暂时模棱两可（“花园小路”）的句子时，他们的眼睛向前移动常常会被回退打断。这些回归之后通常会重新阅读之前阅读过的文本的某些部分。提出了选择性重新分析假说（SRH），该假说建议读者回归到发生错误解析的花园路径句法短语标记中的关键选择点，然后重新分析句法结构以在大多数情况下获得正确的解析。然而，最近的大量研究表明，读者常常无法从这些句​​子中得出正确的解析或解释。如果这些最近的观察是准确的，也许重读不一定是战略性的、受控的或可预测的。目前的研究包括两项大规模的眼球追踪实验，专门设计用于检查人们在何处重读花园小径句子以及重读的程度，以及重读是否会影响理解准确性。采用可变文本屏蔽范例来限制重读期间对花园路径和非花园路径部分的访问。扫描路径分析用于确定部分或全部参与者是否针对先前阅读文本的语法关键部分。理解问题探讨了最终的解释。简而言之，读者经常误解花园小径，并且没有任何重读措施可以预测更好的理解。此外，扫描路径分析揭示了读者之间和内部的巨大差异。只有一小部分试验符合基于结构的预测。总而言之，我们未能找到对结构性战略重读的支持。因此，我们建议重读这些句子本质上更多的是“确认”而不是“修正”。