Due to limitations in the diffusers used for global irradiance measurements, the method usually applied to correct global irradiance measurements is investigated. Based on sky radiance measurements, the assumption of a homogeneous sky radiance distribution generally used to calculate the diffuse part of the global cosine error is accurate to within ¿1.5% in the UVB for varying atmospheric and geographic conditions. At longer wavelengths, the assumption of a homogeneous sky radiance distribution should not be used for global cosine corrections since it underestimates the true diffuse cosine error by up to 10% at 500 nm. The accuracy of diffuse irradiance measurements using shading disks is investigated and its dependence on different aerosol amounts is shown. The cosine correction of global irradiance measurements is shown to be insensitive to errors in the determination of the fraction of direct to global irradiance. The uncertainty in the cosine correction of global irradiance measurements is therefore less than ¿2% in the UVB. ¿ American Geophysical Union 1996