Several methods which have been proposed to aid the identification of polarity in Farns-worth-Munsell (FM) 100-Hue test plots are described. The assumptions which underlie one type of method are discussed and problems identified. Two methods are taken as representative of the two types of analysis and their performance compared on plots from congenital and acquired color defectives. The key difference between the two methods and the types of method which they represent lies in the imposition of symmetry on the plot. From the performance of 200 diabetic subjects, the problems created by the presence of monopolar plots, plots with unequal magnitudes of errors in the two poles of bipolar plots, and plots in which the two poles are not diametrically opposite are highlighted. It is concluded that the simpler type of analysis is sufficient and retains more information.