The absorbance changes that accompany the light-driven proton-pumping cycle of bacteriorhodopsin measured over a broad range of times, wavelengths, temperatures, and pH values have been globally fitted to the kinetic model K in equilibrium with L in equilibrium with X in equilibrium with M in equilibrium with N in equilibrium with O----bR. A remarkably good fit to the data was obtained by optimizing the rate constants at 20 degrees C and the corresponding activation energies at each pH value, together with the extinction coefficients for each intermediate, which were assumed to be independent of both pH and temperature. Back-reactions are included for all but the last step of the cycle and are found to be essential for fitting the data. The rates of these reactions are large, and the analogous irreversible model produced significantly worse fits to the data. Small systematic differences between the fit and the experimental data associated with the X, M, and O intermediates, together with the inability of the model to produce spectra for the X and M intermediates consistent with their assignment as molecular species, indicate that this model must be an incomplete description of the photocycle. We suggest that these problems arise from the presence of additional occupied states that are difficult to distinguish on the basis of their visible absorption spectra alone.