人脸视错觉是人类进化的结果

看看这些图，在每张图中我(原文作者，下同)都能看到脸的样子，我猜你也一样。

之所以会这样是因为大多数人都对脸很痴迷。我们到处都能看到脸，而我们所看到的脸其实并不是脸。最有名的例子应该是月球上的男人了。最奇怪的例子则要数一个人花了2.8万美元买了一个不新鲜的烤奶酪三明治，因为这个三明治不小心被烤出了Virgin Mary的轮廓。

这种现象被称作人脸视错觉(也称作幻想性视错觉)，这并非什么新奇的名词。但没有人对为何我们的大脑会造成这种错觉知之甚详。本周我偶然看到了一个与这些问题相关的调查。这项名为《土司上的耶稣》(Seeing Jesus in Toast)的调查发表在《大脑皮层》杂志上。今年秋天它还获得了另类诺贝尔奖，被誉为“先令人发笑后引人深思”的研究。

该研究给20名志愿者展示了上百幅“噪点图像”(包含了黑、白和灰色斑点的正方形)然后告诉志愿者这上百幅图中有一半图像中包含了难以检测到的脸。(参与者都接受过训练，能够看到这类图像中明显包含着的脸，因此他们能用这种能力来搜寻噪点中的脸)。看过一副噪点图之后，志愿者会按下一个按钮来告诉研究人员她是否看到了图像中的脸。这些参与者们不知道的是，这些噪点图像中根本没有任何显而易见的脸。一周之后，这些参与者又回到实验室经历了相同的过程。然而这次，他们被告知一半的图像中含有难以检测到的字母。事实上，没有图像中有字母——这些图像和他们上周看过的一模一样。

所有这些实验在脑部扫描仪中进行，这样科学家就能够比较在发生人脸视错觉、字母视错觉和不发生幻想性视错觉的时候，大脑的哪部分较为活跃。结果表明发生人脸视错觉的时候，大脑中的纺锤状脸部区域(英文简称FFA)变活跃了，但发生字母视错觉或者不发生幻想性视错觉的时候，这一区域无变化。更重要的是，参与者大脑中的FFA越活跃，她在大脑中合成的图片就更像一张人脸。

先前有研究表明FFA这个区域是专门用来处理真实的脸(也因此得名)，它在想象脸的时候会变活跃表明FFA同我们对脸的抽象概念有关，它能对脸的基本模式(两只眼睛和一个鼻子)作出反应。为何我们的大脑会这么容易就想象出一张脸？这有一个强制的进化解释：这种人们会在模糊视觉信息中检测脸的倾向可能是人们高度适应社交生活中脸面重要性的结果。

[附：英文版原文]

ONLY HUMAN: December 17, 2014

Why Do We See the Man in the Moon?

by Virginia Hughes

Take a look at the slideshow above. The photos depict, in order: tower binoculars, a tank tread, tree bark, headphones, a tray table, a toilet, eggs, and more tree bark. Yet I perceived every one of them as a face, and I bet you did, too.

That’s because, as I wrote about a few weeks back, most people are obsessed with faces. We see faces everywhere, even in things that are most definitely not faces. The most famous example is probably the man in the moon. The weirdest has got to be the person who reportedly paid $28,000 for an old grilled cheese sandwich whose burn marks outline the face of the Virgin Mary.

This phenomenon, called face pareidolia, isn’t new (Leonardo da Vinci even wrote about it as an artistic tool). But nobody knows much about how or why our brains create this illusion. This week I came across a fascinating brain-imaging study that begins to investigate these questions. The paper, published in the journal Cortex, is titled “Seeing Jesus in Toast,” and this fall it won an Ig Nobel Prize, awarded “for achievements that first make people laugh then make them think.”

The study hinges on a clever method for inducing pareidolia inside of a brain scanner. The researchers showed 20 volunteers hundreds of “noise images” — squares comprised of black, white, and gray blobs — and told them that half of the images contained hard-to-detect faces. (The participants had been through a training period in which they saw clearly defined faces in such images, so they were used to the act of searching for a face within the noise.) After seeing a noise image, the volunteer would press a button indicating whether she saw a face in it or not. Unbeknownst to the participants, none of the noise images contained any overt faces.

The scientists reasoned that trials in which participants reported seeing a face were examples of pareidolia. To confirm this, the researchers took all of the images in which a participant saw a face and combined them into an average image. They then subtracted from that the average of all of the images in which the same participant did not see a face. The result of that subtraction, somewhat amazingly, was a crude face shape, suggesting that participants really were experiencing face pareidolia.

A week later, the same participants came back in the lab and went through a similar procedure. This time, though, they were told that half of the noise images they saw contained a hard-to-detect letter. In reality, none of them did — in fact, the images were exactly the same as those they saw the previous week.

All of these experiments happened inside of a brain scanner, allowing the scientists to compare which parts of the brain are activated during face pareidolia, letter pareidolia, and no pareidolia.

It turns out that a particular brain area — the right fusiform face area (FFA) — showed enhanced activation during face pareidolia but not letter pareidolia or no pareidolia. What’s more, the higher a volunteer’s activation in the right FFA, the more her subtracted composite image looked like a face, the study found.

This is an intriguing finding, the researchers say, because of what’s already known about the FFA. Previous studies had found that this area is specialized for processing true faces (hence the name). The fact that it’s also active for these imagined faces suggests that it’s involved in our more abstract conceptions of faces, as opposed to simply responding to the basic visual pattern of two eyes on top of a nose.

And why do our brains so easily create faces? There’s a compelling evolutionary explanation, the researchers write. “This tendency to detect faces in ambiguous visual information is perhaps highly adaptive given the supreme importance of faces in our social life.”

Regardless of what’s going on in my brain, there’s something delicious about looking at photos of face pareidolia, don’t you think? If you have your own examples, please share — the weirder the better!