Several variants of the beta amyloid protein, differing only at their carboxy terminus (beta 1-39, beta 1-40, beta 1-42, and beta 1-43), have been identified as the major components of the cerebral amyloid deposits which are characteristic of Alzheimer's disease. Kinetic studies of aggregation by three naturally occurring beta protein variants (beta 1-39, beta 1-40, beta 1-42) and four model peptides (beta 26-39, beta 26-40, beta 26-42, beta 26-43) demonstrate that amyloid formation, like crystallization, is a nucleation-dependent phenomenon. This discovery has practical consequences for studies of the beta amyloid protein. The length of the C-terminus is a critical determinant of the rate of amyloid formation ("kinetic solubility") but has only a minor effect on the thermodynamic solubility. Amyloid formation by the kinetically soluble peptides (e.g., beta 1-39, beta 1-40, beta 26-39, beta 26-40) can be nucleated, or "seeded", by peptides which include the critical C-terminal residues (beta 1-42, beta 26-42, beta 26-43, beta 34-42). These results suggest that nucleation may be the rate-determining step of in vivo amyloidogenesis and that beta 1-42 and/or beta 1-43, rather than beta 1-40, may be the pathogenic protein(s) in AD.