Memory conjunction errors in normal and amnesic subjects

Amnesic and control subjects were presented compound words and performed either a deep or a shallow encoding task. They later received a surprise ''old''/''new'' recognition test or perceptual identification test that contained old, recombined, partially new, and completely new words. In recognition, controls were better able than amnesics to discriminate old from recombined stimuli; however, the groups were equally able to discriminate partially or completely new stimuli from recombined stimuli. There were no between-group differences in perceptual identification performance; both groups showed strong priming for old stimuli but not for recombined stimuli. The research demonstrates that memory illusions can result from miscombining parts of previously experienced stimuli, and that, relative to controls, amnesics suffer from a selective inability to intentionally remember how stimulus parts were interrelated. (C) 1996 Academic Press, Inc.