Site-directed mutagenesis has been used to probe the interactions that stabilize the equilibrium and burst phase kinetic intermediates formed by apomyoglobin. Nine bulky hydrophobic residues in the A, E, G and H helices were replaced by alanine, and the effects on protein stability and kinetic folding pathways were determined. Hydrogen exchange pulse-labeling experiments, with NMR detection, were performed for all mutants. All of the alanine substitutions resulted in changes in proton occupancy or an increased rate of hydrogen-deuterium exchange for amides in the immediate vicinity of the mutation. In addition, most mutations affected residues in distant parts of the amino acid sequence, providing insights into the topology of the burst phase intermediate and the interactions that stabilize its structure. Differences between the pH 4 equilibrium molten globule and the kinetic intermediate are evident: the E helix region plays no discernible role in the equilibrium intermediate, but contributes significantly to stabilization of the ensemble of compact intermediates formed during kinetic refolding. Mutations that interfere with docking of the E helix onto the preformed A/B/G/H helix core substantially decrease the folding rate, indicating that docking and folding of the E helix region occurs prior to formation of the apomyoglobin folding transition state. The results of the mutagenesis experiments are consistent with rapid formation of an ensemble of compact burst phase intermediates with an overall native-like topological arrangement of the A, B, E, G, and H helices. However, the experiments also point to disorder in docking of the E helix and to non-native contacts in the kinetic intermediate. In particular, there is evidence for translocation of the H helix by approximately one helical turn towards its N terminus to maximize hydrophobic interactions with helix G. Thus, the burst phase intermediate observed during kinetic refolding of apomyoglobin consists of an ensemble of compact, kinetically trapped states in which the helix docking appears to be topologically correct, but in which there are local non-native interactions that must be resolved before the protein can fold to the native structure.