The sensitivity of the soil dust aerosol cycle to radiative forcing by the soil dust aerosol particles themselves is studied. Four experiments with the NASA GISS atmospheric general circulation model, which includes a soil dust aerosol model, are compared, all using a prescribed climatological sea surface temperature as a lower boundary condition. In one experiment, dust is included as a dynamic tracer with no radiative effect, whereas dust interacts with radiation in the other simulations. The single-scattering albedo of dust particles is prescribed to be globally uniform in the experiments with radiatively active dust, although this albedo is varied from experiment to experiment. On a global scale the radiative forcing by dust generally causes a reduction in the atmospheric dust load, corresponding to a decreased dust source flux. The dust source flux and its changes are analyzed in more detail for the main source regions. This analysis shows that the reduction varies both with the season and with the single-scattering albedo of the dust particles. The experiments show that dust radiative forcing can lead to significant changes both in the soil dust cycle and in the climate state. To estimate dust concentration and radiative forcing by dust more accurately, dust size distributions and dust single-scattering albedo in the model should be a function of the source region, because dust concentration and the climate response to dust radiative forcing are sensitive to dust radiative parameters. ¿ 2001 American Geophysical Union