An extensive set of line-by-line plus doubling-adding reference computations for both clear and overcast skies has been utilized to develop, calibrate, and verify the accuracy of a new multiple-band solar parameterization, suitable for use in atmospheric general circulation models. In developing this parameterization the emphasis is placed on reproducing accurately the reference absorbed flux in clear and overcast atmospheres. In addition, a significantly improved representation of the reference stratospheric heating profile, in comparison with that derived from older, broadband solar parameterizations, has been attained primarily because of an improved parameterization of CO2 heating. The exponential-sum-fit technique is used to develop the parameterization of water vapor transmission in the main absorbing bands. An absorptivity approach is used to represent the heating contributions by CO2 and O2, and a spectral averaging of the continuum-like properties is used to represent the O3 heating. There are a total of two pseudomonochromatic intervals needed to do the radiative transfer problem in the vertically inhomogeneous atmosphere is 72. The delta-Eddington method is used to solve for the reflection and transmission of the homogeneous layers, while the adding method is used to combine the layers. The single-scattering properties of the homogeneous layers can account for all types of scattering and absorbing constituents (molecules, drops, ice particles, and aerosols), given their respective single-scattering properties and mass concentrations. With respect to the reference computational results the clear-sky heating rates are generally accurate to within 10%, and the atmospheric absorbed flux is accurate to within 2%. An analysis is made of the factors contributing to the error in the parameterized cloud absorption in the near infrared. Derivation of the representative drop coalbedo for a band using the mean reflection for an infinitely thick cloud (thick-averaging technique) generally results in a better agreement with the reference cloud absorbed flux than that derived using the mean drop coalbedo (thin-averaging technique), except for high, optically thin water clouds. Further, partitioning the 2500<&ngr;<8200 cm-1 spectral region into several more bands than two (the minimum required) results in an improved representation of the cloud absorbed flux, with a modest increase in the shortwave radiation computational time. The cloud absorbed flux is accurate to within 10%, and the cloudy layer heating rates are accurate to within 15%, for water clouds, while larger errors can occur for ice clouds. The atmospheric absorbed, downward surface, and upward top-of-the-atmosphere fluxes are generally accurate to within 10%. ¿ 1999 American Geophysical Union