The role of ATP hydrolysis in actin polymerization has been a puzzle, since it is known that polymer formation is possible without the ATPase activity and that the ATPase lags behind polymerization. We have used beryllium fluoride and G-ADP actin monomers to form F-ADP-BeF3- filaments that are a stable analog for either the ATP or the ADP-P(i) state. Electron microscopy and computed three-dimensional reconstruction have been used to compare this state to control actin, F-ADP, polymerized from G-ATP. We find, at a high degree of statistical significance, that subdomain-2 of the actin protomer in the ADP-BeF3- state is in a conformation very similar to that found in the atomic model for F-actin of Holmes and co-workers, but becomes disordered after the release of the phosphate. This breaks one of the longitudinal bonds in the filament, consistent with biochemical observations that phosphate release destabilizes F-actin. We have also found that lithium, which reduces the dissociation rate constant of actin filaments, induces a structural state indistinguishable from that of ADP-BeF3-. Further, in all states about ten C-terminal residues are displaced from the above mentioned model, but that the fit of the rest of the monomer is in excellent agreement, supporting the uniqueness of the solution they found and precluding a significantly different arrangement of the actin monomer in the filament.