Abstract We examine the distinction between animacy, a conceptual property, and agency, a conceptual and linguistic property involved in semantic role assignment. Are these two notions—agency and animacy—interchangeable in the mind/brain of the comprehender or are they distinct yet correlated? If the latter, does brain behavior show sensitivity to this distinction? Through a self-paced reading and an fMRI experiment we investigate the comprehension of sentences containing aspectual verbs (e.g., begin, finish) in composition with animate vs. inanimate-denoting subjects (e.g., The noblewoman/prologue began the book). Aspectual-verb sentences are ambiguous independently of subject denotation: whereas sentences with inanimate subjects engender a constitutive/subpart reading along one of multiple possible dimensions (e.g., the spatial vs. informational construal of the book), sentences with animate-denoting subjects can also engender a constitutive/subpart reading along one of several possible dimensions, (The story about) the noblewoman began the book (informational dimension), in addition to the agentive reading along the eventive dimension, The noblewoman began (doing something with) the book. So, both animate and inanimate subject-denotations can license the construal of the complement as a structured individual along mutually-exclusive conceptual dimensions (spatial, informational, eventive, etc.). During real-time comprehension such a situation calls for ambiguity resolution, and therefore computational cost, to settle on one interpretation. Therein lies the relevance of this construction: the final reading of the sentence depends on whether or not the subject can be interpreted as agentive independent of animacy. Results show that controlling for animacy, all aspectual-verb sentences (1) engender longer reading times, consistent with shared increased compositional cost and (2) exhibit overlapping cortical activity: left angular gyrus and left inferior frontal gyrus, consistent with the distinction between concept composition on the one hand, and ambiguity resolution processes on the other. This converging behavioral and neurological pattern thus suggests that sentence interpretation need not be determined by perception of animacy per se but by a more fundamental conceptual parameter that has, as by-product, agency but that more locally results from the control relationship between two participants of the situation denoted by the sentence. Finally, (3) sentences with animate-denoting subjects (> inanimate-subject counterparts) additionally recruited the posterior cortex; an activation pattern consistent with the construal of animate-denoting individuals as “default” controllers in the agentive reading. This suggests in turn a de-coupling of animacy and agency such that the former is an entity-level attribution that supports but does not determine an eventive interpretation, and the latter is the output of a control asymmetry evaluation in an eventive representation connecting two otherwise independent entities. Altogether, the findings are consistent with a neurocognitive model of sentence semantic comprehension whereby meaning construal and composition are carried out within a parametrized conceptual space that is only secondarily modulated by independent conceptual factors such as animacy.

中文翻译：

---

主体生命力和未指明的意义：概念和皮质基础

摘要 我们研究了生命力（一种概念属性）和能动性（一种涉及语义角色分配的概念和语言属性）之间的区别。这两个概念——能动性和生命力——在理解者的头脑/大脑中是否可以互换，或者它们是否不同但相互关联？如果是后者，大脑行为是否对这种区别表现出敏感？通过自定进度的阅读和 fMRI 实验，我们调查了对包含有生命与无生命的主题（例如，贵妇人/序言开始这本书）的组合中包含体格动词（例如开始、结束）的句子的理解。Aspectual-动词句子是模棱两可的，独立于主语外延：而带有无生命主语的句子会产生沿着多个可能维度之一的构成性/子部分阅读（例如，空间与空间）。书的信息解释），带有动画表示主题的句子也可以沿着几个可能的维度之一产生构成性/子部分阅读，（关于）贵妇人开始这本书（信息维度），除了代理阅读之外事件维度，贵妇人开始（用）这本书。因此，有生命的和无生命的主语外延都可以将补语解释为一个结构化的个体，沿着相互排斥的概念维度（空间、信息、事件等）。在实时理解过程中，这种情况需要解决歧义，因此需要计算成本，以解决一个解释。这就是这种结构的相关性：句子的最终阅读取决于主语是否可以被解释为独立于生命力的施事者。结果表明，控制生命力，所有方面动词句子（1）产生更长的阅读时间，与共同增加的成分成本一致，（2）表现出重叠的皮质活动：左角回和左额下回，与概念之间的区别一致一方面是组合，另一方面是歧义解决过程。因此，这种趋同的行为和神经学模式表明，句子解释不需要由对生命力本身的感知决定，而是由一个更基本的概念参数决定，作为副产品，代理，但更多的是由句子表示的情况的两个参与者之间的控制关系产生的。最后，（3）带有表示有生命的主语（> 无生命的主语对应物）的句子额外招募了后皮层；一种激活模式，与将有生命的个体作为代理阅读中的“默认”控制器的解释一致。这反过来又表明了生命力和能动性的脱钩，使得前者是实体级归因，支持但不决定事件解释，后者是事件表示中控制不对称评估的输出，连接两个否则独立实体。共，