Of all the atmospheric constituents, aerosol content is shown to be responsible for the greatest variations in direct luminous efficacy. Some clarity is brought to the comparison between Linke's and Angstrom's turbidity coefficients, respectively [T.sub.L] and [Beta]. Grenier's recent formulation of the optical thickness of a water and aerosol free atmosphere is presented here in a simplified expression. Based on these results and Dogniaux's illuminance turbidity factor, [T.sub.i1], two direct luminous efficacy models are derived, one of which is tuned to our experimental data. The input parameters are optical air mass, [Beta], and water vapour content in the tuned version. These models perform significantly better than any of twelve other models found in the literature when compared to 1 yr's measurements from each of two sites in the U.S. and Switzerland. In both sites, [Beta] was derived from horizontal visibility estimated in a nearby airport.