While it has been commonly assumed that a deficit in semantic memory underlies many of the clinical and cognitive features of early Alzheimer's disease (AD), there has been little agreement on what constitutes a proper theoretical description of this impaired system. Currently, most theories of this disorder fall into two categories: those that posit degraded semantic representations, and those that posit impaired retrieval operations with relatively intact semantic representations. It is argued that each position has both empirical and logical disadvantages that have prevented the development of a consensus on how to describe the pathology of semantic memory in AD. In this paper we present the details and supportive evidence for an alternative account of the semantic memory deficit of AD: the Gain/Decay hypothesis. The core claim of the Gain/Decay hypothesis is that a reduction in the time constant of spreading activation in AD produces dynamic changes in the availability of semantic representations that depend on the time frame in which this information must be accessed. The implications of this hypothesis for a range of experimental and clinical phenomena are discussed, as are possible biological correlates of the hypothesized alterations in the modulation of activation. (JINS, 1999, 5, 641–658.)