Abstract: Criminal crime has become a universal social problem faced by all countries. With the concealment of criminal behavior and the diversification of criminal means, the traditional methods of investigation have become increasingly unable to meet the needs of the current criminal investigation. Therefore, an effective and reliable lie detection technique is urgently needed to meet the current severe situation of combating crime. This study adopted GKT (Guilty knowledge test) paradigm, with the suspect as the object, with real crime scene material, investigated the validity and reliability of the ERP index (P300\CNV) which associated with lie detection during real criminal environment. Twenty-six subjects were participate in the present study, which are all male and real criminal suspects. And the age range is 19-46 year old. The GKT paradigm involved three stimuli: target stimulus, probe stimulus and irreverent stimulus. Those stimuli all presents as pictures, and the emergence proportion of the target, probe and irrelevant stimuli were 1:1:4. Participants were required to respond to whether the stimulus they had ever seen before a key appears immediately. At the same time, EEG activity was recorded. Repeated measures of variance analysis were performed on the behavior data (accuracy and response times (RTs)) and the ERP index (P300 and CNV), separately. Behavior results showed that, the accuracy of target stimuli were significantly smaller than probe stimuli and irreverent stimuli while the accuracy of probe stimulus and irrelevant stimulus was not significant. In addition, the RTs of probe stimuli were significantly smaller than target stimuli and irreverent stimuli, whereas the RTs of target stimulus was greater than irrelevant stimulus, but not significant. Moreover, the ERP results showed that, compared with target stimuli and irreverent stimuli, the P300 (400-500 ms, localized in Fz、Cz、Pz、Oz、POz、FCz) amplitude of the probe stimuli was greater than the target stimulus and irrelevant stimulus, whereas the P300 amplitude of target stimulus was significantly greater than the irrelevant stimulus. The CNV (800～1700ms, localized in Fz、Cz、Pz、Oz、POz、FCz) amplitude of the irrelevant stimulus was significantly lower than the amplitude of the probe stimulus and target stimulus, whereas the difference between the probe stimulus and target stimulus was not significant. Furthermore, bootstrap method was used to calculate the detection rate with P300 peak value as the index, and the result found that the detection rate of P300 was 92.5%. Taken together, these findings showed three conclusions. First, ERP lie detection technology is applicable to lie detection in the field of criminal investigation and the GKT paradigm could indeed induced the P300 and CNV components. Second, compared with CNV, P300 is a reliable lie detection index, which can effectively detect the details of criminal, and has higher detection rate and better stability, further promote the application of ERP lie detection technology in criminal investigation. Finally, as a lie-detection index, the validity of the CNV component need to be tested in the future, and its generation may require certain conditions.

Key words: GKT delayed-response paradigm, P300, CNV, ERP, lie detection

摘要： 刑事犯罪已成为当今各国面临的普遍性社会问题，因此急需可靠有效的测谎技术以适应当前打击犯罪的严峻形势。本研究采用犯罪知识测试范式（Guilty knowledge test, GKT），以犯罪嫌疑人为对象，以真实犯罪现场为素材，探究真实环境下与测谎探测相关的脑电指标（P300、CNV）的有效性和可靠性。行为数据显示，目标刺激的反应正确率显著小于探测和无关刺激，探测刺激的反应时显著小于目标刺激和无关刺激。脑电数据显示，相较于目标刺激和无关刺激，探测刺激诱发的P300波幅最大且P300的检出率为92.5%，而在CNV指标上目标刺激诱发的波幅最大, 无关刺激次之，探测刺激最小。结果表明，与CNV相比，P300能够有效检测被试所隐藏的犯罪细节信息，且检出率较高，稳定性较好，是一种较为可靠的测谎指标。

关键词: GKT范式, P300, CNV, ERP, 测谎