根据一项研究，探测回忆活动引发的神经活动的大脑扫描可能让科学家――还可能让执法官员――确定一个人是否认出了一张特定的面孔。利用功能磁共振成像（fMRI），Jesse Rissman及其同事发现了与受试者识别面容有关的神经活动的模式，这些受试者研究了一个图像数据库中的数以百计的面容，然后看了一系列的照片，这些照片包括了一些来自这个数据库的面容，而另一些面容是新的。这组科学家在受试者表明他们是否识别出了每一个面容的时候进行了扫描。这组作者报告说，模式识别软件在大脑扫描中分离出了一种记忆的“神经特征”， 而且这种特征在各个受试者身上是一致的。然而，这组作者指出，fMRI扫描在知识检测方面有严重的限制。特别是这种技术无法可靠地区分正确地报告识别出了一幅图像和错误地声称识别出了此前没有见过的面容的受试者。尽管这些发现提示fMRI可以探测记忆，这组作者强调说，这种技术可能无法在一些情况下工作，而且“假阳性”可能带来严重的法律后果。
论文 #10-01028: "Detecting individual memories through the neural decoding of memory states and past experience," 作者 Jesse Rissman, Hank Greely和Anthony Wagner
媒体联系人：Jesse Rissman，斯坦福大学心理系
Department of Psychology, Stanford University, Stanford, CA
电话：650-724-9515
电子邮件：jesse.rissman@stanford.edu
Published online before print May 10, 2010, doi: 10.1073/pnas.1001028107
Detecting individual memories through the neural decoding of memory states and past experience
Jesse Rissmana,1, Henry T. Greelyb, and Anthony D. Wagnera,c,1
Abstract
A wealth of neuroscientific evidence indicates that our brains respond differently to previously encountered than to novel stimuli. There has been an upswell of interest in the prospect that functional MRI (fMRI), when coupled with multivariate data analysis techniques, might allow the presence or absence of individual memories to be detected from brain activity patterns. This could have profound implications for forensic investigations and legal proceedings, and thus the merits and limitations of such an approach are in critical need of empirical evaluation. We conducted two experiments to investigate whether neural signatures of recognition memory can be reliably decoded from fMRI data. In Exp. 1, participants were scanned while making explicit recognition judgments for studied and novel faces. Multivoxel pattern analysis (MVPA) revealed a robust ability to classify whether a given face was subjectively experienced as old or new, as well as whether recognition was accompanied by recollection, strong familiarity, or weak familiarity. Moreover, a participant's subjective mnemonic experiences could be reliably decoded even when the classifier was trained on the brain data from other individuals. In contrast, the ability to classify a face's objective old/new status, when holding subjective status constant, was severely limited. This important boundary condition was further evidenced in Exp. 2, which demonstrated that mnemonic decoding is poor when memory is indirectly (implicitly) probed. Thus, although subjective memory states can be decoded quite accurately under controlled experimental conditions, fMRI has uncertain utility for objectively detecting an individual's past experiences.
