To address a recent issue of whether contemporary clear-sky shortwave radiative transfer models do or do not portray reality, we have collocated satellite measurements of reflected shortwave radiation at the top of the atmosphere, made as part of the Earth Radiation Budget Experiment, with measurements of surface insolation made at 24 stations located throughout southern Canada. From this, a clear-sky data set has been constructed by using both the satellite and the surface measurements to identify clear days, and the data set extends over a 4-year period (1985--1988). Two quite different types of shortwave radiative transfer models were employed, one a model-derived algorithm that converts the top-of the-atmosphere measurements to surface insolation, and the other a stand-alone column model which computes the surface insolation independently of the satellite measurements. Both models incorporate prescribed aerosols, and they were compared to the clear-sky data by using satellite-measured atmospheric water vapor as model input. The models are in excellent agreement with the clear-sky measurements when averaging is performed over all stations and over all seasons. The relative bias error in surface insolation is roughly 1% for both models, despite the fact that one model is constrained to the top-of-the-atmosphere measurements and the other is not. Thus the models' incorporation of tropospheric aerosols seems quite realistic when compared to the averaged data. For daily means and individual stations, however, the models' relative bias errors in surface insolation are observed to range from -6% to 10%, and it is demonstrated that this is caused by temporal and geographic variability in tropospheric aerosol loading. Because the models incorporate prescribed aerosols, they produce a relative bias error of roughly -6% for aerosol-free conditions. The opposite limit of 10% corresponds to aerosol loading in excess of that prescribed in the models. There is no evidence that the models incorrectly portray any important physical processes, and we find no support for a suggestion that there is an unknown and substantial source of shortwave absorption by water vapor. Our comparisons do not, however, rule out recently suggested minor absorbers of shortwave radiation. Also, since a very stringent clear-sky identification procedure was used with the goal of removing forest fire smoke as well as clouds, our study does not rule out possible atmospheric absorbers resulting from urban pollution. ¿ 1998 American Geophysical Union