The density matrix approach to polarized radiative transfer is reviewed, with particular emphasis on the physical assumptions that are at the basis of the recent developments achieved by means of this formalism. In particular it is shown that two of the basic hypotheses (the hypothesis of neglecting correlation effects between thermal velocity and density matrix, and the hypothesis of neglecting atomic polarization in the atomic ground level) are highly questionable for the description of resonance polarization -and its modifications due to the presence of a magnetic field- in spectral lines formed in the solar atmosphere.