Mindreading mindreading: Using multivariate pattern analysis to decode the neural basis of mental state ascription.

读心术（Mindreading），也被称为心智化（mentalising）或心智理论（Theory of Mind），是一种思考他人所见、所知、所想、所想和意图的能力，被认为是社交互动和沟通的基础。读心术在幼儿和婴儿中得到了广泛的研究，已知自闭症和精神分裂症患者会受到损害，最近开始在成人中研究其认知和神经基础。使用脑成像的研究，以及检查脑损伤和脑刺激的影响，一致认为与不涉及理解其他思想的类似任务相比，某些大脑区域明显参与了读心术，从而导致它们构成了“社交”或“读心术”的大脑网络。通常认为这些大脑区域代表读心信息。然而，最近的另一种说法是，读心信息在其他地方，也许是在“语义”大脑区域，而读心大脑网络只是控制对这些信息的访问。这些观点之间的测试对于我们理解大脑如何支持读心术至关重要，并且对长期争论的关于我们如何恐惧的理论解释具有影响。读心术的基本单位是将特定代理（例如，Mariam）到特定内容（例如，...我以为盒子里有一只兔子。我们将要求参与者使用呈现人物图片的任务和描述他们正在思考的简单场景来简要地记住这些信息。因此，参与者将被引导记住类似于“Mariam认为盒子里有一只兔子，但我知道真的有一个钟”的东西，特定的代理和特定的内容在不同的审判中有所不同。如果“读心的大脑区域”主要参与了这些信息的表征，那么就有可能从这些区域的大脑活动模式中解码出正在思考的主体（在本例中为玛丽雅姆）和他们正在思考的内容（兔子）。然而，如果“读心脑区”实际上只将注意力直接集中在代表“语义”脑区的信息上，那么就应该有可能从“语义”脑区的活动中解码施事和内容信息，而不是“读心”脑区。这项工作是由最近的发展，任务评估读心术和分析技术解码信息从细粒度的模式，人类大脑活动。我们的任务提出了大量试验的读心问题，在这些试验中，施事和内容是系统变化的。三个实验将集中在内容（实验1），代理+内容（实验2），以及内容信息随时间的协调（实验3）。实验1和2将使用功能磁共振成像，它提供了良好的空间分辨率，而实验3将使用脑电图，它提供了良好的时间分辨率。这三个实验都将采用先进的技术对大脑活动数据进行多变量分析，以检查我们是否真的可以从“读心术”或“语义”大脑区域解码关于谁在想什么的信息。我们的研究将显示，社交脑的区域是否真的编码了读心术所需的信息，并确定这些信息被激活的顺序。这些结果将对大脑如何支持读心术，以及读心术如何通过脑损伤或自闭症和精神分裂症等疾病中的非典型脑功能而被破坏产生影响。

Mindreading (also known as mentalising, or Theory of Mind) is the ability to think about what others see, know, think, want and intend, and is thought to be a fundamental basis of social interaction and communication. Mindreading has been widely studied in young children and infants, is known to be impaired in people with autism and schizophrenia, and more recently its cognitive and neural basis has begun to be studied in adults. Research using brain imaging, and examining the effects of brain injury and brain stimulation, converges on the view that some brain regions are distinctively involved in mindreading compared with similar tasks that do not involve understanding other minds, leading to suggestions that they comprise a "social" or "mindreading brain network". It is typically assumed that these brain regions represent mindreading information. However, a recent alternative suggestion is that mindreading information is represented elsewhere, perhaps in "semantic" brain regions, and that the mindreading brain network merely controls access to this information. Testing between these views is critical for our understanding of how the brain supports mindreading, and has implications for long-debated theoretical accounts about how we mindread.The fundamental unit of mindreading is a representation that relates a particular AGENT (e.g., Mariam) to a particular CONTENT (e.g., ...thinks there's a rabbit in the box). We will ask participants to hold such information briefly in mind using tasks that present pictures of people and simple scenarios that depict what they are thinking. Thus, participants will be led to hold in mind something like "Mariam thinks there's a rabbit in the box, but I know there's really a bell", with the particular AGENT and the particular CONTENTS varying from trial to trial. If "mindreading brain regions" are centrally involved in representing such information, then it should be possible to decode the AGENT who is thinking (Mariam in this case), and the CONTENT they are thinking about (Rabbit) from patterns of brain activity in these areas. However, if "mindreading brain regions" actually only direct attention to information represented "semantic" brain regions then, it should be possible to decode AGENT and CONTENT information from activity in "semantic" brain areas, but not "mindreading" areas. The work is made possible by recent developments in tasks for assessing mindreading and analytic techniques for decoding information from fine-grained patterns of human brain activity. Our tasks present mindreading problems for a large number of trials over which AGENT and CONTENT are systematically varied. Three experiments will focus on CONTENT (Experiment 1), AGENT+CONTENT (Experiment 2), and the co-ordination of CONTENT information over time (Experiment 3). Experiments 1 and 2 will employ fMRI, which gives excellent spatial resolution, while Experiment 3 will use EEG, which gives excellent temporal resolution. All three experiments will employ advanced techniques for multivariate analysis of brain activity data to examine whether we can indeed decode information about who is thinking what from "mindreading" or "semantic" brain regions. Our studies will show whether regions of the social brain actually encode information necessary for mindreading, and determine the order in which such information is activated. These results will have implications for how the brain supports mindreading, and how mindreading may be disrupted through brain injury, or atypical brain function in disorders such as autism and schizophrenia.