Target scattering decompositions have become the standard tools for scattering classification and for target geophysical parameter extraction from polarimetric synthetic aperture radar (SAR) data. Incoherent decompositions are based on the characteristic decomposition of the target coherency matrix and the use of its eigenvalues and eigenvector parameters for a unique characterization of target scattering. The presence of speckle, however, may introduce biases to the incoherent parameters derived from the characteristic decomposition of the coherency matrix sample. In this study, it is shown that both coherency eigenvalues and eigenvector parameters estimates may be significantly biased, with a bias that decreases with an increase in the size of the coherency sample processing window. The statistics of cross-channel coherence we have previously derived for circular complex Gaussian processes can be used to determine the minimum processing window size required for unbiased estimation of the incoherent deco...