The accuracy to which the atmospheric component of the longwave flux at the ocean surface can be estimated by using simple empirical formulae, with information from standard ship meteorological reports as input, has been tested by comparison with radiometric measurements made on cruises in the midlatitude North Atlantic and the Southern Ocean. Over a wide range of atmospheric conditions the best agreement between the estimated and measured downwelling longwave flux is obtained by using the formula of Clark et al. <1974>, which has a mean bias error with respect to our combined data set of -0.7 W/m2. Two other formulae were considered: the Bunker <1976> formula, which is found to overestimate the atmospheric longwave flux at values less than about 350 W/m2, resulting in an overall bias of 9.4 W/m2, and the Bignami et al. <1995> formula, which underestimates the flux over the entire range considered, giving a bias of -26.5 W/m2. The difference in performance of the Clark and Bunker formulae is ascribed to the different parameterizations assumed for the effect of clouds and atmospheric humidity on the longwave flux. It is suggested that the Clark formula be used in future climatological studies at midlatitudes. ¿ 1997 American Geophysical Union