In relation to the primary sequence and three-dimensional structure of rabbit muscle glycogen phosphorylase, we have carried out a comparative sequence analysis of phosphorylases from human, rat, Dictyostelium, yeast, potato and Escherichia coli. Based on sequence similarity, a large region of the protein is shared by these enzymes extending from alpha-helix-1 to the last alpha-helix-33. Conserved residues are equally distributed between the N and C-terminal domains and occur primarily in buried residues. Phylogenetic analysis indicates that the two isozymes within either E. coli, potato or Dictyostelium are more closely related to each other than they are to other phosphorylases. Yeast phosphorylase is most closely related to the Dictyostelium isozymes. Mammalian muscle and brain isozymes are more closely related to each other than to the liver isozyme and the muscle isozyme is evolving at the slowest rate. All phosphorylases exhibit high conservation of active site and pyridoxal phosphate binding residues. Most phosphorylases also exhibit high conservation of sugar binding residues in the glycogen storage site. Phosphorylation and AMP binding site residues are poorly conserved in non-mammalian phosphorylases. In contrast, glucose-6-P binding residues are highly conserved in four of the seven non-mammalian enzymes. Analysis of interacting pairs of dimer contact residues indicates that they can be grouped into three relatively independent networks. One network contains phosphorylation and AMP binding residues and is poorly conserved in non-mammalian enzymes. A second network contains glucose-6-P binding residues and is highly conserved in enzymes containing a conserved glucose-6-P binding site. A third, conserved network contains residues within the tower helix and gate loop. A model for the evolution of allostery in phosphorylase is proposed, suggesting that glucose-6-P inhibition was an early control mechanism. The later creation of primarily distinct ligand binding sites for AMP/phosphorylation control may have allowed the establishment of a separate dimer contact network for propagating conformational changes leading to activation rather than inhibition of enzyme activity.