Results of several prior studies, in which Alzheimer's disease (AD) patients missed the same concepts on multiple tasks, have been used to substantiate the theory that AD causes concept-specific loss of information from semantic memory However, sample sizes in these studies are modest, test-retest intervals small, and typically only a few tasks were used. In the present study 69 An subjects were annually administered 11 tasks, each using the same 13 concepts. Only a few instances were observed in which a concept was missed across all 11 tasks. When performances on the Oral Reading and Dictation tasks were removed from analysis, because of their questionable reliance on semantic memory, the number of missed concepts rose only modestly. A substantial rise in the number of missed concepts occurred, however, when performances on the four multiple- choice tasks were removed. Interpreting the larger number of missed concepts on the five remaining generative semantic tasks as evidence of item-specific loss is problematic, nonetheless, because the generative semantic tasks were among the hardest in the battery and the frequency with which an individual subject missed a concept across all tasks accorded with the subject's dementia severity level. Results also indicate that task difficulty, more than concept specificity, determine whether a concept is missed. Overall, results suggest that a concept will “disappear” when all of the tasks in which it is a stimulus become too difficult for the patient to perform. Study results call into question the appropriateness of using batteries of effortful, attention demanding tasks for ascertaining whether AD causes item-specific loss of conceptual knowledge.