The question of which stage of the triplet di-pi-methane rearrangement controls the regiochemistry and excited-state rate was resolved in the present investigation. In this study, using m-cyanodibenzobarrelene, the m-cyano substituent was located in a strategic location such that bridging between the vinyl and substituted benzo moiety would lead to stabilized diradical I but nonstabilized diradical II. Conversely, the substitution was such that with bridging between the vinyl and unsubstituted benzo ring, diradical II would be stabilized but diradical I would not. The study of the di-pi-methane rearrangement of this system thus weighed the relative importance of stabilization of diradical I versus stabilization of diradical II along the excited triplet hypersurface. Although in m-cyanodibenzobarrelene the two alternative pathways lead to the same m-cyanodibenzosemibullvalene, m-cyanodibenzobarrelene labeled with deuterium on the vinyl bridge affords different products via the two pathways. Experimentally, it was found that the di-pi-methane rearrangement of dideuterio-m-cyanodibenzobarrelene was completely regioselective. The reaction proceeded selectively to afford the regioisomer resulting from formation of the more delocalized diradical I. Stabilization of diradical II proved not to be competitive. In the same study, a theoretical approach was applied to the rearrangement of barrelene itself. Ab initio computations were carried out with geometry optimization of each point on the triplet surface. Additionally, energies were obtained for points on the ground-state surface with triplet geometries. Also, for strategic points, geometry-optimized ground-state energies were obtained. These computations provided insight into the observed photochemistry. The most striking feature of the theoretical results was the appearance of a real energy minimum for diradical I with the appearance of a triplet reaction transition state corresponding to formation of diradical I from triplet barrelene.