Protein aggregation is a prominent feature of many neurodegenerative diseases, such as Alzheimer's, Huntington's, and Parkinson's diseases, as well as spongiform encephalopathies and systemic amyloidoses. These diseases are sometimes called protein misfolding diseases, but the latter term begs the question of what is the "folded" state of proteins for which normal structure and function are unknown. Amyloid consists of linear, unbranched protein or peptide fibrils of approximately 100 A diameter. These fibrils are composed of a wide variety of proteins that have no sequence homology, and no similarity in three-dimensional structures--and yet, as fibrils, they share a common secondary structure, the beta-sheet. Because of the prominence of amyloid deposits in many of these diseases, much effort has gone into elucidation of fibril structure. Recent advances in solid-state NMR spectroscopy and other biophysical techniques have led to the partial elucidation of fibril structure. Surprisingly at the time, for beta-amyloid, a set of 39-43-amino-acid peptides believed to play a pathogenic role in Alzheimer's disease, the beta-sheets are parallel with all amino acids of the sheets in-register. Since the time of those observations, however, it has become clear that there is no universal structure for amyloid fibrils. While many of the amyloid fibrils described thus far have a parallel beta-sheet structure, some have antiparallel beta-sheets, and other, more subtle structural differences among amyloids exist as well. Amyloids demonstrate conformational plasticity, the ability to adopt more than one stable tertiary fold. Conformational plasticity could account for "strain" differences in prions, and for the fact that a single polypeptide can form different fibril types with conformational differences at the atomic level. More recent data now indicate that the fibrils may not be the most potent or proximate mediators of cyto- and neurotoxicity. This damage is not confined to cell death, but also includes more subtle forms of damage, such as disruption of synaptic plasticity in the central nervous system. Rather than fibrils, prefibrillar aggregates, variously called "micelles," "protofibrils," or ADDLs (beta-amyloid-derived diffusible ligands in the case of beta-amyloid) may be the more proximate mediators of cell damage. These are soluble oligomers of aggregating peptides or proteins, but their structure is very challenging to study, because they are generally difficult to obtain in large enough quantities for high-resolution structural techniques, and they are temporally unstable, rapidly changing into more mature, and eventually fibrillar forms. Consequently, the mechanisms by which they disrupt cellular function are also not well understood. Nevertheless, three broad, overlapping, nonexclusive sets of mechanisms have been proposed as responsible for the cellular damage caused by soluble, oligomeric protein aggregates. These are: (1) disruption of cell membranes and their functions [e.g., by inserting into membranes and disrupting normal ion gradients]; (2) inactivation of normally folded, functional proteins [e.g., by sequestering or localizing transcription factors to the wrong cellular compartment]; and (3) "gumming up the works," by binding to and inactivating components of the quality-control system of cells, such as the proteasome or chaperone proteins.