Previous studies have produced inconsistent results concerning the two components of autobiographical memory—personal semantic memory and episodic memory. Results in subjects with mild cognitive impairment (MCI) and dementia of Alzheimer's type (DAT) have varied concerning the existence of a temporal gradient in retrograde amnesia. These results have important theoretical implications regarding multiple trace theory versus standard consolidation models of long-term memory (LTM). We investigated whether this variability arises from differences in the methods used in assessing autobiographical memory. We examined patterns of memory impairment in 20 healthy elderly controls, 20 MCI subjects, and 10 DAT subjects using the Autobiographical Memory Interview (AMI) of Kopelman and the Autobiographical Interview (AI) of Levine. Both the AMI and AI were modified to allow for the test scores to be derived from a single interview without fatiguing the subjects. On the AMI, DAT subjects were significantly impaired on both components of autobiographical memory—episodic memory and personal semantics—with episodic memory showing a significant though gentle temporal gradient sparing childhood memories. Using the AI test, subjects with DAT showed impaired recall of episodic details (but not personal semantics), again with a gentle temporal gradient. Differences between the two interview methods (fewer epochs in the AMI; fewer memories per epoch in the AI) were found to have a significant impact on the pattern of findings; fewer epochs in the AMI brought out the temporal gradient, and fewer memories per epoch (in the AI) diminished it. These data show the importance of technical details of the different tests in favouring one versus another LTM theory. The data are not purely compatible with either theory. © 2012 Wiley Periodicals, Inc.