Abstract: The concealed information test (CIT) is a lie detection method, which can determine if a criminal suspect knows the crime-related information and then infer whether the suspect is guilty. The CIT has high internal validity but it is easy to misjudge an innocent person who knows the crime-related information as a guilty one. Therefore, it is necessary to improve CIT's detection accuracy for the guilty and the knowledgeable innocent persons. The complex trial protocol (CTP) is a modified CIT which can effectively resist countermeasures, but it hasn’t been successfully applied to discriminate the guilty from the knowledgeable innocent.
The present study designed a novel CTP based on self-referential encoding, which measured the early posterior negativity (EPN) and P300. The probe or one of the irrelevant stimuli was randomly presented in the first phase of a trial, and the target (participant’s name) or one of the non-targets (others’ names) was randomly presented in the second phase of a trial. The participants needed to make self-related or self-unrelated responses in two phases when they saw the stimuli.
The results revealed that: (1) the area under the curve (AUC) for P300 to discriminate the guilty from the unknowledgeable innocent was 0.922, which was significantly higher than 0.5 (the chance level). The AUC for P300 to discriminate the guilty from the knowledgeable innocent was 0.605, which was not significantly different from 0.5, and the false positive rate of the knowledgeable innocent was 75%. The AUC for P300 to discriminate the knowledgeable innocent from the unknowledgeable innocent was 0.859, which was significantly higher than 0.5. (2) The AUC for EPN to discriminate the guilty from the unknowledgeable innocent was 0.770, and the AUC for EPN to discriminate the guilty from the knowledgeable innocent was 0.721. The two AUCs were significantly larger than 0.5. The false positive rate of the knowledgeable innocent was 12.5%. The AUC for EPN to discriminate the knowledgeable innocent from the unknowledgeable innocent was 0.516, which was not significantly different from 0.5.
In summary, the present study indicated that: (1) P300 can effectively discriminate the guilty from the unknowledgeable innocent, but can not effectively discriminate the guilty from the knowledgeable innocent. P300 is easy to misjudge the knowledgeable innocent as the guilty person. (2) The discrimination of EPN was weaker than that of P300 in discriminating the guilty from the unknowledgeable innocent. However, the discrimination of EPN was superior to that of P300 in discriminating the guilty from knowledgeable innocent. EPN is not easy to misjudge a knowledgeable innocent person as a guilty one. (3) When the CTP based on self-referential coding is applied to detect lies, P300 can be used to discriminate the guilty from the unknowledgeable innocent and EPN can be used to discriminate the guilty from the knowledgeable innocent.

Key words: complex trial protocol, self-referential encoding, knowledgeable innocent, EPN, P300