Language comprehension and production are generally assumed to use the same representations, but resumption poses a problem for this view: This structure is regularly produced, but judged highly unacceptable. Production-based solutions to this paradox explain resumption in terms of processing pressures, whereas the Facilitation Hypothesis suggests resumption is produced to help listeners comprehend. Previous research purported to support the Facilitation Hypothesis did not test its keystone prediction: that resumption improves accuracy of interpretation. Here, we test this prediction directly, controlling for factors that previous work did not. Results show that resumption in fact hinders comprehension in the same sentences that native speakers produced, a finding which replicated across four high-powered experiments with varying paradigms: sentence-picture matching (N=300), self-paced reading (N=96), visual world eye-tracking (N=96), and multiple-choice comprehension question (N=150). These findings are consistent with production-based accounts, indicating that comprehension and production may indeed share representations, although our findings point toward a limit on the degree of overlap. Methodologically speaking, the findings highlight the importance of measuring interpretation when studying comprehension.

语言理解和生产通常被假定使用相同的表示，但恢复对这种观点提出了一个问题：这种结构是有规律的生产，但被认为是非常不可接受的。针对这一悖论的基于生产的解决方案从处理压力的角度解释了恢复，而促进假说则表明恢复是为了帮助听众理解。先前声称支持促进假说的研究并没有测试其关键预测：这种恢复提高了解释的准确性。在这里，我们直接测试这个预测，控制以前工作没有的因素。结果表明，复工实际上阻碍了对母语者产生的相同句子的理解，这一发现在四个不同范式的高功率实验中重复：句子-图片匹配（N = 300）、自定进度阅读（N = 96）、视觉世界眼动追踪（N = 96）和多项选择题（N = 150）。这些发现与基于生产的解释一致，表明理解和生产确实可能共享表征，尽管我们的发现指向重叠程度的限制。从方法论上讲，研究结果强调了在研究理解时衡量解释的重要性。