The early stages of the bacteriorhodopsin photocycle, including the J(625), K-590, and L(550) intermediates and the role of water molecules within the protein interior, are studied by means of molecular dynamics simulations. Our calculations examine two model’s for the excited state potential surface governing the observed all-trans --> 13-cis photoisomerization : one surface hindering a C-14-C-15 single-bond corotation and the other surface allowing such corotation. The investigations use as a starting structure a model of bacteriorhodopsin based on electron-microscopy studies and subsequent molecular dynamics refinement. The following questions are addressed : How does the binding site guide retinal’s photoisomerization? How does the photoisomerization depend on features of the excited state potential surface? Can one recognize a J(625) intermediate? How does water participate in the early part of the pump cycle? How is the initial photoreaction affected by a lowering of temperature? To model the quantum yield, i.e., the dependence of the dynamics on initial conditions, 50 separate isomerization trials are completed for each potential surface, at both 300 and 77 K, the trials being distinguished by different initial, random velocity distributions. From these trials emerge, besides all-transretinal, three different photoproducts as candidates for the K-590 intermediate : (1) 13-cis-retinal, with the Schiff base proton oriented toward Asp-96; (2) 13-cis-retinal, highly twisted about the C-6-C-7 bond, with the Schiff base proton oriented perpendicular to the membrane normal; (3) 13,14-dicis-retinal, with the Schiff base proton oriented toward the extracellular side. Two candidates for the K-590 intermediate, case 2 and case 3 above, were subjected to simulated annealing to determine corresponding L(550) structures. We suggest that photoproduct 2 above most likely represents the true K-590 intermediate. Water molecules near the Schiff base binding site are found to play a crucial role in stabilizing the K-590 state and in establishing a pathway for proton transfer to Asp-85.