The amino acid sequence and disulfide bridges of bovine serum albumin reveal nine double loops formed by the bridges. The pattern and size of loops and connecting segments between loops indicate that a basic repeating unit (domain) consists of "large double loop-short connecting segment-small double loop-long connecting segment-large double loop-connecting segment between "domains" and repeats exactly three times. Thus, albumin arose by duplication of the primordial single domain gene, followed some time later by a half-gene duplication to give the ancestor of the present triple domain structure. These duplications are estimated to have occurred about 700 million years ago, based on the differences between domain (75 to 82%), the difference between bovine and human albumins (20%),and reported time estimates for globin and immunoglobulin evolution. Still farther back in time, the single domain arose by triplication of a gene of a primordial subdomain of about 77 amino acid residues, consisting of one large double loop with a segment on each end. The structure resulting from this triplication consisted of three large loops and three long connecting segments. Later, a gene deletion for part of a connecting segment and part of the middle large loop occurred to produce the large-small-large loop pattern. A proposed three dimensional structure for the "subdomain" shows spatial and sequence similarity with the G-H helical regions of myoglobin or hemoglobin, suggesting that it arose by duplication and separation of a gene for a C-terminal segment of a primitive globin. The CYS residues were introduced as an adaptive, convergent evolutionary event after separation from globin, but before duplication of the subdomain gene.