A central pursuit of cognitive neuroscience is to find neural mechanisms of cognition, with research programs favoring different strategies to look for them. But what is a neural mechanism, and how do we know we have captured them? Here I answer these questions through a framework that integrates Marr's levels with philosophical work on mechanism. From this, the following goal emerges: What needs to be explained are the computations of cognition, with explanation itself given by mechanism-composed of algorithms and parts of the brain that realize them. This reveals a delineation within cognitive neuroscience research. In the premechanism stage, the computations of cognition are linked to phenomena in the brain, narrowing down where and when mechanisms are situated in space and time. In the mechanism stage, it is established how computation emerges from organized interactions between parts-filling the premechanistic mold. I explain why a shift toward mechanistic modeling helps us meet our aims while outlining a road map for doing so. Finally, I argue that the explanatory scope of neural mechanisms can be approximated by effect sizes collected across studies, not just conceptual analysis. Together, these points synthesize a mechanistic agenda that allows subfields to connect at the level of theory.

中文翻译：

---

认知神经科学中神经机制的批判视角：走向统一。

认知神经科学的一个中心任务是寻找认知的神经机制，研究项目倾向于采用不同的策略来寻找它们。但什么是神经机制，我们如何知道我们已经捕获了它们？在这里，我通过一个将马尔的层次与机制哲学工作相结合的框架来回答这些问题。由此，出现了以下目标：需要解释的是认知的计算，解释本身是由算法和实现算法的大脑部分组成的机制给出的。这揭示了认知神经科学研究的一个轮廓。在前机制阶段，认知的计算与大脑中的现象相关联，缩小了机制在空间和时间中所处的位置和时间。在机制阶段，确定了如何从填充预机械模具的零件之间有组织的相互作用中产生计算。我解释了为什么转向机械建模有助于我们实现目标，同时概述了这样做的路线图。最后，我认为神经机制的解释范围可以通过研究中收集的效应大小来近似，而不仅仅是概念分析。这些要点共同综合了一个机械议程，允许子领域在理论层面上相互联系。