Cerebrovascular deposition of amyloid is a frequent observation in Alzheimer's disease patients. It can also be detected sporadically in normal aged individuals and is further found in familial diseases linked to specific gene mutations. The source and mechanism of this pathology are still unknown. It has been suggested that amyloidogenic proteins are derived from blood, the vessel wall itself, or from the central nervous system. In this article evidence is reviewed for and against each of these hypotheses, including new data obtained from transgenic mouse models. In APP23 transgenic mice that develop cerebral amyloid angiopathy (CAA) in addition to amyloid plaques, the transport and drainage of neuronally produced amyloid-beta (A beta) seem to be responsible for CAA rather than vascular A beta production or blood uptake. Although a number of mechanisms may contribute to CAA in humans, these results suggest that a neuronal source of A beta is sufficient to induce vascular amyloid deposition. The possibility to cross genetically defined mouse models of CAA with other mutant mice now has the potential to identify molecular mechanisms of CAA.