Putting the individual into face recognition: Bridging theory and application

这个项目将研究一项基本社交技能的神经基础--从面孔识别我们认识的人的能力。我们每天都会认识我们的亲戚、朋友和同事几十次，但令人惊讶的是，人们对我们是如何做到这一点的知之甚少。我们认为，其中一个原因是之前的研究没有承认熟悉面孔识别的特殊性。我们都有自己独特的一组熟悉的面孔，这些面孔往往只与其他面孔部分重叠。在这里，我们使用允许结合个人的、特殊的熟悉度的方法来解决熟悉面孔识别的问题。考虑到特殊的熟悉度将导致更可靠的人脸识别措施，并将使我们能够发展创新的理论和方法重点。以前的研究在很大程度上已经确定了熟悉的面孔在大脑中的什么位置以及何时被识别。在这里，我们将首次系统地研究已知面孔的视觉外观是如何存储在大脑中的。此外，我们还将研究一个具有实际意义的问题--即使参与者有隐藏这类知识的动机，也能可靠地检测出他们对一张脸的熟悉程度。在第一步，我们将解决同一张脸在不同条件下(例如，由于光线、视角或化妆的变化)可能看起来非常不同的问题。那么，大脑如何可能识别出这张熟悉的面孔呢？我们发展了两种潜在的互补理论观点来解释这一现象。第一种观点认为，我们应该把一张脸的“要点”储存在我们的记忆中，也就是说，在任何情况下都能观察到的信息。第二种认为，除了这种抽象的表征，我们可能对一张脸有许多具体的记忆--类似于在不同场合拍摄的“快照”--因此关于一张脸在特定遭遇中是什么样子的信息。我们的项目将为“要点”和“快照”观点提供决定性的证据。在第二条线索中，我们将研究不同面孔的神经表征是如何组织起来的，以实现有效的识别。我们认识成千上万的面孔。那么，怎么可能不经常把它们弄混呢？也许是因为我们所知道的不同面孔的组织方式，只允许“最佳匹配”被激活，而其他可能看起来相似的面孔被禁止。这一想法是基于人脸识别的计算机模型，但很少在人类观察者身上进行测试，其神经基础完全未知。我们现在已经开发了一系列新颖的实验研究来填补这一空白。最后，尽管之前的认知神经科学研究大多局限于检查群体中的人脸加工，即通过折叠多个参与者的数据，但对于任何潜在的现实应用来说，能够在单个参与者中检测到熟悉性是至关重要的。在应用情况下，知道10-20名参与者平均表现出某种大脑反应是没有用的，因为通常需要关于单个证人或嫌疑人的证据。因此，该项目将开发一种新的神经测量方法来检测个体的熟悉度。此外，为了具有实际意义，即使参与者试图掩盖熟悉性，这种措施也需要可靠地发挥作用--例如，避免在刑事调查中牵连共谋者。因此，我们将测试对企图欺骗的稳健性。总体而言，通过考虑特殊的熟悉程度，并将研究重点从“何时何地”转移到“如何”，从群体到个人层面，这个项目将在人类大脑如何识别熟悉的面孔方面产生创新的发现-这是一个具有很高理论重要性的问题。此外，我们的成果将有助于解决一个具有重大实际意义的问题。

This project will examine the neural basis of a fundamental social skill - the ability to recognise the people we know from their faces. We recognise our relatives, friends, and colleagues dozens of times every day, but surprisingly little is known about how we achieve this. We propose that one reason for this is that previous research has not acknowledged the idiosyncrasy of familiar face recognition. We all have our own unique set of familiar faces, and these tend to overlap only partly with others. Here, we address the problem of familiar face recognition using methods that allow incorporation of individual, idiosyncratic familiarity. Taking idiosyncratic familiarity into account will result in substantially more reliable measures of face recognition and will allow us to develop an innovative theoretical and methodological focus. Previous work has largely established where in the brain and when in time a familiar face is recognised. Here, we will, for the first time, systematically examine how the visual appearance of known faces is stored in the brain. In addition, we will examine a question of practical importance - the reliable detection of familiarity with a face, even when participants are motivated to conceal such knowledge.In a first strand, we will tackle the problem that the same face can look very different in different conditions (e.g. due to changes in lighting, viewing angle, or make-up). How then is it possible for the brain to recognise it as the same familiar face? We have developed two potentially complementary theoretical views to explain this phenomenon. The first suggests that we store the "gist" of a face in our memory, i.e. the information that is commonly observed in all circumstances. The second suggests that, in addition to this abstract representation, we may have many specific memories of a face - similar to "snapshots" taken on different occasions - and thus information about what a face looked like in a particular encounter. Our project will provide decisive evidence for the 'gist' and 'snapshot' views. In a second strand, we will examine how neural representations of different faces are organised to allow for efficient recognition. We know literally thousands of faces. How then is it possible not to constantly mix them up? Maybe because the different faces we know are organised in a way that allows only the "best match" to become activated while other, potentially similar looking faces are inhibited. This idea is based on computer models of face recognition, but has rarely been tested with human viewers and its neural basis is completely unknown. We have now developed a series of novel experimental studies to fill this gap.Finally, while previous cognitive neuroscience research has mostly been constrained to examine face processing in groups, i.e. by collapsing data across a number of participants, it is crucial for any potential real-life application that familiarity can be detected in individual participants. In applied situations, it is not useful to know that a group of 10-20 participants on average shows a certain brain response, as evidence is typically required about an individual witness or suspect. This project will therefore develop a novel neural measure to detect familiarity in individuals. Moreover, to be of practical relevance, such a measure needs to work reliably even when participants are trying to conceal familiarity - for example to avoid implicating conspirators in criminal investigations. We will therefore test for robustness against attempted deceit.Overall, by taking idiosyncratic familiarity into account as well as by shifting the research focus from the "where/when" to "how", and from the group to the individual level, this project will generate innovative findings on how the human brain recognises familiar faces - a question of high theoretical importance. In addition, our results will contribute to solving a problem of substantial practical relevance.