A central aspect of language comprehension is that hearers integrate incoming linguistic content both with the rest of the current sentence and the larger discourse context. Presuppositions crucially interact with both intraand inter-sentential context in intricate ways, which makes their study especially useful in this regard. We present a series of experiments investigating the time-course of interpreting presuppositions in online comprehension and the impact that so-called presupposition projection in cases where presuppositions appear in embedded environments has on this. We find immediate delays in eye tracking reading times when the presupposition of German wieder (‘again’) is not supported by the context, but only for unembedded occurrences of wieder. Further evidence from a rating experiment and a stops-making-sense study supports our interpretation of this result to the effect that global presuppositions of embedded presupposition triggers are not immediately available in processing. A second reading time experiment explores the effects of embedding further by providing presuppositional support in different locations in contexts with a more complex structure involving conditionals. We find longer reading times when the support is more distant, measured in terms of the number of projection steps posited by Discourse Representation Theory (DRT). Altogether, the results suggest that presupposition projection is a cognitively effortful process, and are thus consistent with theoretical accounts that reflect this in terms of the complexity of the representations involved in the different types of contexts, while other accounts that are more neutral in this regard need to be supplemented by additional assumptions or alternative explanations for the observed effects. On the more general level of discourse processing models, these results suggest that there is even more structure relevant to cognitive processes at a level between the surface representation and the purely semantic level (e.g., the commonly assumed level of a text-base) than previously assumed.

中文翻译：

---

在线处理中的预设投影

语言理解的一个中心方面是，听众将输入的语言内容与当前句子的其余部分以及较大的语篇上下文相结合。预设以复杂的方式与句子内和句子间的上下文至关重要地相互作用，这使得它们的研究在这方面特别有用。我们提出了一系列实验，研究在线理解中解释预设的时程以及在嵌入式环境中出现预设的情况下所谓的预设投影的影响。当上下文不支持德语wieder的预设（“再次”）时，我们发现眼动跟踪阅读时间会立即延迟，而仅针对未嵌入的wieder出现。评级实验和制止感研究的进一步证据支持我们对这一结果的解释，即嵌入的预设触发器的全局预设在处理中不会立即可用。第二阅读时间实验通过在涉及条件的更复杂结构的情况下在不同位置提供预设支持来进一步探索嵌入的效果。根据话语表征理论（DRT）提出的投影步骤数，我们发现支撑距离更远时，阅读时间更长。总而言之，结果表明预设投影是一个认知上努力的过程，因此，这与在不同类型的情境中所涉及的表示形式的复杂性方面反映出这一点的理论说明是一致的，而在这方面较为中立的其他说明则需要补充其他假设或观察结果的替代解释。在话语处理模型的更一般的层面上，这些结果表明，与表面上的表征和纯粹的语义层面（例如，基于文本的普遍假设的层面）之间的层面相比，与认知过程相关的结构甚至更多。假定。