Experimental observations on the structure and physicochemical properties of TMV protein assemblies have led to a fundamental switch in the model of the self-assembly process: rather than being nucleated by the hypothetical two-layer disk, virus assembly appears to be initiated by interaction of the specific RNA sequence with a short helical aggregate of the coat protein arranged as in the virus. Formation of the 20s nucleating aggregate involves the binding of an average of half a proton per protein subunit. This proton-binding site can be identified with the carboxyl-carboxylate pair that is formed between top and bottom protein surfaces at a radius of 58 A in the virus helix. Because the 20s aggregate consists of about two helical turns, only one carboxyl-carboxylate pair will be formed between each top-bottom pair of protein subunits. Limitation of the length of the 20s helical aggregate at neutral pH can be accounted for by disorder of the inner loop of the protein chain, due to electrostatic repulsion among the carboxyl groups that form the anomalous proton-binding site at 25 A radius in the ordered virus structure. To grow beyond two to three turns, inner loops of the protein at the interior of the helix must be ordered in the close-packed arrangement. The electrostatic repulsion opposing this ordering can be overcome by binding of the viral RNA at neutral pH, by calcium binding, or by proton binding in slightly acid solution. Virus disassembly upon infection appears to result from the low intracellular calcium and proton concentration compared to the extracellular environment, which increases the electrostatic repulsion among the negatively charged groups involved in calcium and proton binding, thereby allowing cellular ribosomes to competitively bind the viral RNA. Disk aggregates of TMV protein, which form at high ionic strength in alkaline solution, do not appear to be involved in virus assembly. The stacked-disc aggregate, which was previously presumed to be built of a polar stack of the hypothetical polar two-layer aggregate, is, in fact, a bipolar structure. Because the bonding between turns of the disc structures is different from that of the virus helix, direct switching between these structures by the postulated dislocation does not occur. TMV assembly does appear to involve conservation of bonding specificity, as initially presumed, but only in helical packing arrangements of the protein subunits. Switching from disordered to ordered conformations of the protein, dependent on changes in the electrostatic interactions among the protein subunits, appears to be critical in controlling the assembly process.