The design of recombinant globins as oxygen storage and delivery pharmaceuticals must address two key protein engineering problems, one dealing with the improvement of apoprotein stability and the other with the structural determinants of ligand binding rates. Thirteen naturally occurring apomyoglobins were observed to unfold following the same two transitions as sperm whale apomyoglobin but with variable stabilities. Even with highly similar proteins, it is difficult to identify the individual amino acids and the specific intramolecular noncovalent forces that confer protein stability. Site-directed mutagenesis of apomyoglobins suggests that single point mutations have complex, but generally small effects on the unfolding of apomyoglobin. However, two pig multiple mutants were constructed on the basis of substitution trends in stable mammalian myoglobins, and these genetically engineered proteins had markedly increased overall stabilities. This result suggests that analysis of naturally occurring variants may be the best way to select modifications that inhibit denaturation. Oxygen binding was measured by both conventional and ultrafast laser photolysis techniques for more than 70 myoglobin mutants at 25 different positions. Intramolecular geminate rebinding of oxygen to sperm whale myoglobin occurs on a nanosecond timescale at room temperature and shows two well separated kinetic components, indicating at least two internal sites from which the photodissociated ligands return to the iron. Xenon accelerates the fast reaction but decelerates and diminishes the slower reaction. The rates and proportions of the two components and xenon effects on them vary widely for different mutants and suggest photodissociated ligands occupy xenon site 4 in the distal pocket and xenon site 1 below the plane of the heme. Rebinding from these positions corresponds to the slower geminate phase for oxygen rebinding. Computed bimolecular rate constants for ligand entry suggest that only residues immediately adjacent to the bound ligand and Phe46 are important for ligand movement into and out of the protein. Ligand entry and exit do not appear to occur via transiently connected channels through the interior of the protein. Instead, the kinetic data presented here suggest that ligands enter and escape through a region circumscribed by the distal histidine, Phe46, and the heme propionates.