The inverse static light-scattering problem for spheres is treated numerically by employing genetic algorithms. Intensity functions are generated with variable amounts of noise and are used as input data in order to deduce both the refractive index and the radius of the particle. Both real and complex refractive indices are treated. Comparisons are made with a gradient-based minimization approach that requires fewer function evaluations but which often terminates in a local minimum. The genetic algorithm approach gives very good results for the parameters even in cases where the noise level in the data is relatively high.