Three different types of cirrus cloud field, reconstructed in three dimensions directly from midlatitude observations by a cirrus stochastic model, are used to study the effects of three-dimensional radiative transfer in both the long-wave and short-wave spectral regions. Calculations of three-dimensional radiative transfer (3D), the independent column approximation (ICA) and the plane-parallel approximation are compared to quantify the effects on heating rates, radiative fluxes and related properties. Locally the heating rate difference between 3D and ICA reaches more than 10 K day(-1) in both the long-wave and short-wave, depending upon the distributions of ice water content, which indicates that horizontal radiation transport plays an important role in structures of heating rate. The domain-averaged heating profiles of 3D agree within a few tenths of a K day(-1) with ICA but show a systematic low bias. The domain-averaged heating rates in cloud layers are increased in 3D by up to 7% in the long-wave and more than 20% in the short-wave. The root-mean-square differences at individual points are up to ten times larger than the corresponding domain-averaged differences, representing the cancellation of opposing 3D effects. The ICA biases in long-wave net flux and emissivity have their maximum values (similar to 2-3%) near cloud top for the thinnest cloud with lowest fractional coverage. In general, ICA tends to reduce the reflected upwelling short-wave flux at the top of clouds; the layer-averaged albedo at cloud top agrees with 3D within 1%, although the corresponding RMS difference may differ by up to 30% from 3D at high solar zenith angles. Similar results are found for whole-sky (cloudy plus clear) short-wave reflectances and transmittances for which ICA agrees with 3D within 5%. For domain-averaged short-wave absorptance, however, ICA errors can reach 20%. The corresponding RMS differences may differ by up to 50% in reflectance and transmittance but exceed 200% in absorptance. The effects of solar zenith angle are also discussed. Copyright (c) 2007 Royal Meteorological Society.