Several models have been proposed for the quaternary structure of alpha-crystallin. Some suggest the subunits are arranged in concentric shells. Others propose that the subunits are in a micelle-like arrangement. However, none is able to satisfactorily account for all observations on the protein and the quaternary structure of alpha-crystallin remains to be established. In this review, factors contributing to the assembly and polymerization are examined in order to evaluate the different models. Consideration of the variations in particle size and molecular weight under different conditions leads to the conclusion that alpha-crystallin cannot be a micelle or a layered structure. Instead, it is suggested that the protein may be assembled from a 'monomeric' unit comprising eight subunits arranged in two tetramers with cyclic symmetry. The octameric unit is proposed to be disc-like particle with a diameter of 9.5 nm and a height of 3 nm. The larger particles, chains and sheet-like structures commonly observed are assembled from the octamers. Structural predictions indicate that the polypeptide may be folded into three independent domains which have different roles in the structural organization and functions of the protein. It is suggested that the tetramers are stabilized through interactions involving the second domain (residues 64-104) while assembly into the octamers and higher polymers requires hydrophobic interactions involving the N-terminal domain. Deletion of parts of this domain by site directed mutagenesis revealed that residues 46-63 play a critical role in the assembly. Current research aims to identify the specific amino acids involved.