About 15 % of patients homozygous with Z mutant a1-PI develop liver disease. The disorder is linked to accumulation of high molecular weight a1-PI polymers in hepatocytes. Understanding the mechanism of polymerization is crucial for the rational design of therapeutics and for assessing the safety of a1-PI products. We suggest a new model of polymerization, which is at odds with a previously proposed loop-sheet (i.e. head-to-tail) monomer polymerization model. In contrast, we propose that a1-PI dimer (perhaps head-to-head) is the polymerizing unit. Our hypothesis results from an observation that a1-PI polymers, formed in 1.4 M guanidine-HCl, after folding by dilution, polymerize further. Dimers, isolated from such solutions and from polymer solutions generated by heating, form polymers in buffer at ambient temperature. The rate of polymerization is temperature and concentration dependent. Until now, both denaturant- and heat-induced polymerization studies have been used to support the loop-sheet polymerization model. However, polymers formed from dimers produced by these two methods migrate differently by native PAGE. This suggests that the dimers are different and, thus, the polymerization mechanism(s) may be different. Further work is needed to determine if either of these mechanisms is physiologically relevant. However, the ability of both types of dimers alone to polymerize in vitro suggests that dimer may also be the relevant polymerizing unit in hepatocytes.