A major problem in understanding memory is how it can be very long-lasting and stable from early childhood until death, despite massive interruptions in brain state as extreme as prolonged comas. Current prominent candidates for molecular substrates for long-term memory storage have focused on macromolecules such as calmodulin-dependent protein kinase II (CaMKII) coupled with the NMDA receptor and protein phosphatase 2A (2), protein kinase M zeta (PKMζ) (3), and cytoplasmic polyadenylation element binding protein (CPEB) (4), all of which are inside postsynaptic spines. To retain information despite metabolic turnover, all such candidates need to have some sort of bistable switch (e.g., state of phosphorylation or prion conformation) and a mechanism by which older copies of the molecule pass on their status to newer copies to preserve the information. A major problem is that individual intracellular molecules typically last at most a few days before being turned over. Therefore, the information would have to survive being copied tens of thousands of times in a long-lived human, despite metabolic interruptions. Such robust fidelity would be extremely difficult to engineer. Even dynamic computer memory with sophisticated refresh and error correction circuits cannot cope with even a momentary hiccup in its power supply. Instead, long-term information storage in both computers and human civilizations requires writing the information onto physically stable storage media (e.g., magnetic disks, clay tablets, or acid-free paper), which do not require frequent energy-dependent recopying. Aside from some nuclear pore constituents, all of the known really long-lived proteins are insoluble [extracellular matrix] components such as crystallin, elastin, collagen, and proteoglycans (5), which gain stability by extensive cross-linkage and remoteness from intracellular degradative machinery, such as proteasomes, lysosomes, and autophagy.