Ab initio optical rotation data from linear-response coupled-cluster and density-functional methods are compared to both gas-phase and liquid-phase polarimetry data for the small, conformationally flexible molecule epichlorohydrin. Three energy minima exist along the C-C-C-Cl dihedral angle, each with strong, antagonistic specific rotations ranging from ca. -450 to +500 deg/[dm (g/mL)] at 355 nm. Density-functional theory(specifically the B3LYP functional) consistently overestimates the optical rotations of each conformer relative to coupled-cluster theory (in agreement with our earlier observations for conformationally rigid species), and we attribute this to density-functional theory's underestimation of the lowest-lying excitation energies of epichlorohydrin. Length- and velocity-gauge formulations of the coupled-cluster response function lead to slightly different specific rotations (ca. 7% at short wavelengths). We have determined well-converged Gibbs free energy differences among the conformers using complete-basis-set extrapolations of coupled-cluster energies including triple excitations to obtain Boltzmann-averaged specific rotations for comparison to the gas-phase results. The length-gauge coupled-cluster data agree remarkably well with experiment, with the velocity-gauge coupled-cluster and density-functional data bracketing the experimental results from below and above, respectively. Liquid-phase conformer populations reported earlier by Polavarapu and co-workers from combined infrared absorption and theoretical analyses differ markedly from the gas-phase populations, particularly for polar solvents. Nevertheless, Boltzmann-averaged specific rotations from both coupled-cluster and density-functional calculations agree well with the corresponding experimental intrinsic rotations, although the theoretical specific rotations for the individual conformers do not take solvent effects into account. PCM-based estimates of conformer populations lead to poor agreement with experiment.