Studies of learning in patients with focal cerebral lesions have demonstrated that component memory processes are dissociable. Patients with global amnesia due to bilateral medial temporal-lobe or bilateral diencephalic lesions show normal acquisition of motor, perceptual, and cognitive skills despite severe deficits in learning declarative (i.e., factual) information. Patients with unilateral frontal-lobe pathology have deficits in remembering the temporospatial context of new learning episodes, but not the content of these episodes. Dissociations among component memory processes related to the etiology of amnesia have begun to emerge in studies of remote memory function: Bilateral medial temporal-lobe lesions produce a retrograde amnesia that is temporally limited; in contrast, Korsakoff s syndrome and dementia, involving bilateral diencephalic lesions and widespread cerebral pathology, respectively, produce remote memory deficits that are temporally extensive. Our study addressed the selectivity of deficit in remote memory processes with respect to the age of the memory, the nature of the declarative information (personal versus public, content versus date), and the type of memory test (recall versus recognition). Subjects included patients with Alzheimer’s disease (AD), Parkinson’s disease (PD), long-standing focal brain trauma, the patient H.M., who underwent bilateral medial temporal-lobe resection, and healthy control subjects. Remote memory for personal events was assessed using Crovitz’s test. Subjects were asked to relate a personally experienced event incorporating each of 10 nouns, from any period of their life, and to say when the events occurred. After 24 hours, subjects had to reproduce the episodes recalled the previous day. When subjects had difficulty responding, they were cued with examples of specific episodes on Day I , and components of Day 1 memories on Day 2. Responses were scored according to temporospatial specificity on a three-point scale. Remote memory for famous public events from five decades was evaluated using Squire’s four-choice verbal recognition test, and a pictorial task, the Famous Scenes Test. In the recall version of the latter, subjects related the events represented in the picture and estimated the date when it was taken. The recognition version had a three-choice verbal format; the two foils represented real-life events that had occurred about 5 and 15 years before or after the event depicted in the picture. Recognition of content and date for these events was assessed separately but simultaneously. Patients with AD were the most impaired in recalling autobiographical episodes even when cued, but did not differ from control subjects in the age of the episodes