Semantic memory refers to acquired generic knowledge, including knowledge of facts, concepts, and word meanings. Alzheimer's disease (AD) results in a progressive disruption in semantic memory that ultimately has a devastating effect on comprehension, memory for events, thinking, planning, and communicating. This proposal is directed at clarifying the nature of the disruption in semantic memory produced by AD. The project is guided by and intended to further develop a general cognitive architecture called MEM (Johnson, 1983, in press-a,b; Johnson & Hirst, in press). MEM is a theoretical framework for understanding both normal memory and disruption of function, for example, in delusions, confabulation, and amnesia. According to MEM, memory is organized at the most global functional level into two major systems, a perceptual memory system and a reflective memory system. In this framework, semantic memory is supported by both perceptual and reflective systems. The generalized cognitive deficits in AD are hypothesized to result from a disruption of the reflective system, especially a disruption of the ability of executive agendas to recruit other component reflective processes (cf. Nebes, 1989). Based on this characterization of AD, we expect that underlying structural relations among concepts (e.g., a dog, is an animal) and among attributes of concepts (e.g., cars have engines) may remain intact, but patients increasingly have a more difficult time using this information in cognitive tasks because reflective processes central to their use are disrupted. Experiments are proposed that assess the impact of AD on both the availability and use of semantic information of various types. Converging evidence will be obtained from several tasks, including verification of semantic relations (e.g., Is a sparrow a bird?), judgments of goodness or importance (e.g., Which is more central to the concept of a car, engine or hubcaps?), naming objects, and free association. These studies will also provide useful data for identifying the cognitive functions of brain areas damaged in AD (e.g., Van Hoesen & Damasio, 1987).