The linear dichroism spectrum of rhodopsin in sonicated bovine disk membranes was measured 30, 60, 170, and 600 ns after room temperature photolysis with a linearly polarized, 7-ns laser pulse (lambda = 355 or 477 nm). A global exponential fitting procedure based on singular value decomposition was used to fit the linear dichroism data to two exponential processes which differed spectrally from one another and whose lifetimes were 42 +/- 7 ns and 225 +/- 40 ns. These results are interpreted in terms of a sequential model where bathorhodopsin (BATHO, lambda max = 543 nm) decays toward equilibrium with a blue shifted intermediate (BSI, lambda max = 478 nm). BSI then decays to lumirhodopsin (LUMI, lambda max = 492 nm). It has been suggested that two bathorhodopsins decay in parallel to their products. However, a Monte Carlo simulation of partial photolysis of solid-state visual pigment samples shows that one mechanism which creates populations of BATHO having different photolysis rates at 77 K may not be responsible for the two decay rates reported here at room temperature. The angle between the cis band and 498-nm band transition dipoles of rhodopsin is determined to be 38 degrees. The angles between both these transition dipoles and those of the long-wave-length bands of BATHO, BSI, and LUMI are also determined. It is shown that when BATHO is formed its transition dipole moves away from the original cis band transition dipole direction. The transition dipole then moves roughly twice as much towards the original cis band direction when BSI appears. Production of LUMI is associated with return of the transition dipole almost to the original orientation relative to the cis band, but with some displacement normal to the plane which contains the previous motions. The correlation between the lambda max of an intermediate and its transition dipole direction is discussed.