Irradiance measurements on a horizontal surface often deviate from theory where the irradiance is supposed to be proportional to the cosine of the angle of incidence. This discrepancy is known as the cosine error. In this paper, three different methods for cosine error correction are investigated. The simplest method is based on the assumption of an isotropic sky radiance distribution, regardless of sky conditions, and the irradiance is treated as a single component. In the second method the irradiance is divided into one direct solar and one diffuse sky component, where the latter is assumed to have an isotropic distribution. Finally, a third method also divides the irradiance into two components but under the assumption of an anisotropic sky radiance distribution. Irradiances under general sky conditions are found by interpolation between clear and overcast cases on the basis of sunshine duration or cloud cover. The three methods are applied to data from a Robertson-Berger sunburning meter located in Norrk¿ping, Sweden. Both methods, where the irradiance is divided into two components, produce acceptable and similar results, while the isotropic one-component method does not. ¿ 2000 American Geophysical Union