In Deng et al (2021), the self-referential complex trial protocol effectively distinguished guilty participants from knowledgeable innocents in a mock theft case; however, the test’s power to discriminate guilty from innocent was limited (AUC=0.72). The present study made some notable changes from Deng and colleagues (2021). First, the guilty participants completed a simulated murder task at a virtual dock, and the knowledgeable innocent participants witnessed the simulated murder. The probe (i.e., the victim’s face) or one of the irrelevant stimuli (i.e., an unknown bystander’s face) was randomly presented in the first phase, followed by the target (i.e., the participant’s birthday) or one of the non-targets (i.e., an unrelated date) in the second phase. All participants were required to respond with “related to me” or “unrelated to me” button responses when seeing the face and the dates. This study further explored whether the discriminative power of the modified self-referential CTP for distinguishing the guilty from the knowledgeable innocent in a simulated murder case under virtual reality conditions could improve upon an AUC of 0.72. The results demonstrated that: (1) the modified self-referential CTP could effectively discriminate the guilty from the knowledgeable innocent in murder case (AUC=0.91); (2) it also could effectively distinguish the guilty from the unknowledgeable innocent (AUC=0.93) and correctly classify the knowledgeable innocent as innocent. The modified self-referential CTP is an effective method for distinguishing the guilty from the knowledgeable innocent.