Detecting simulated amnesia with event-related brain potentials.

Abstract

Two experiments were done in which groups of normal college undergraduate subjects were instructed to simulate autobiographical amnesia related to head injury. The subjects were tested for autobiographical knowledge with pencil-and-paper recall and recognition tasks, and on another day, with recognition tasks in which event-related potentials were recorded in response to the same autobiographical items, this time presented on a display screen. In the first study, three Event-Related Potential (ERP) block types were utilized: 1) a block in which the subject's birthdate was an oddball with p = 0.11 and other dates (not relevant to the subject) had p = 0.89, 2) a block in which phone numbers were used as stimuli, and 3) a block in which (mothers' maiden) names were used as stimuli. Only birthdate blocks were used in the second study. A P300 ERP was seen in response to rare, personally relevant items. In both studies, the main effect of stimulus type (personally relevant versus not personally relevant) on P300 amplitude was significant (p < 0.0001) and there was no main effect of block type in the first study. Under a sophisticated malingering instruction set, about 15% of the items were correctly recalled and about 50% were correctly recognized by simulators in pencil-and-paper tests. Under a naive malingering instruction set, most simulating subjects failed to recall and recognize autobiographical items. In the first study, an arbitrarily but a priori derived discrimination criterion based on a subject's average P300 amplitude afforded 92% correct discrimination of simulating individual subjects for birthdates and phone numbers. The value was 77% for mothers' maiden names. The same criterion applied to the single birthdate block in the second study yielded 93% correct discrimination. The results suggest the P300 may be useful in detection of malingered amnesia.