[1] The influences of the directional variations and the polarization of the marine reflectance on the retrieval of the inherent optical properties (IOP) (i.e., absorption, scattering, and backscattering coefficients) of water constituents in coastal waters are examined. First, an inversion algorithm based on artificial neural network (NN) methodology is developed using a synthetic data set. The simulations were carried out using a radiative transfer model that accounts for the polarization state of light in the ocean. The simulated data included various directional effects of the particles. The data set is also constrained by observations collected in optically representative coastal waters. In particular, the relationships that exist between the IOP were taken into account, thus making the data set realistic. The results showed that the total IOP were correctly retrieved while the performance of the algorithm to derive the IOP of each water component significantly degrades. However, the inclusion of the directional variations and the polarization of the reflectance in the algorithm improved the accuracy of retrieval of the scattering properties by 15% - 60% and 65% - 75%, respectively. The phytoplankton and noncovarying particles (i.e., nonalgal particles) scattering and backscattering coefficients were derived with an accuracy of 25% and 15% respectively. These results demonstrate the potential of using the polarized signal to separate the total IOP into contribution of biogenic and highly refractive particles in coastal waters. Therefore the development of in situ instrumentation able to measure the polarization properties of the particles is recommended.