Direct measurements of solar radiation at sea were carried out in the Western Mediterranean Sea during several cruises in the years 1989--1994. The data collected were used to perform a test on the empirical formula proposed by Reed <1977>. The agreement between measured and empirical data is noteworthy, taking into account the coarseness of the formula. Nevertheless, Reed's estimates show a bias of about 5 W m-2. In order to investigate the causes of this bias, the data were analyzed according to the time of year and the cloud cover index. The results reveal that the computed insolation is overestimated under clear sky conditions and underestimated for cloudy sky. The disagreement of clear sky predictions can be ascribed to a regional misevaluation of both aerosols and water vapor attenuation. The inadequacy of the cloud cover index is indicated to explain the discrepancy of the estimates under cloudy sky. In any case, the mean difference between observations and formula is very slight, and this supports the use of this type of formula to compute long time average insolation over the sea. ¿ American Geophysical Union 1996