Polarization and radiance profiles of parhelia were measured as a function of scattering angle. The wavelength dependence of the width of the parhelion polarization peak agrees with Fraunhofer diffraction theory, indicating that the broadening of the halos is caused primarily by diffraction. Hence our hypothesis that a spread of interfacial angles is the dominating cause of halo broadening, has proved untenable. The conflicting results between the width of the halo polarization peaks and the observed size distribution of the replicated crystals originates from a strong size-dependent collection efficiency in the sampling. This implies that shapes of sampled crystals need not necessarily be representative for the shapes of the halo-making crystals in a swarm.