A modified version of the nonlinear iterative Chahine algorithm is presented and applied to the inversion of spectral extinction data for particle sizing. Simulated data were generated in a λ range of 0.2-2 µm,and particle-size distributions were recovered with radii in the range of 0.14-1.4 µm. Our results show that distributions and sample concentrations can be recovered to a high degree of accuracy when the indices of refraction of the sample and of the solvent are known. The inversion method needs no a priori assumptions and no constraints on the particle distributions. Compared with the algorithm originally proposed by Chahine, our method is much more stable with respect to random noise, permits a better quality of the retrieved distributions, and improves the overall reliability of the fitting. The accuracy and resolution of the method as functions of noise were investigated and showed that the retrieved distributions are quite reliable up to noise levels of several rms percent in the data. The sensitivity to errors in the real and imaginary parts of the refraction index of the particles was also examined.