In F2 populations, gametic and zygotic selection may affect the analysis of linkage in different ways. Therefore, specific likelihood equations have to be developed for each case, including dominant and codominant markers. The asymptotic bias of the "classical" estimates are derived for each case, in order to compare them with the standard errors of the suggested estimates. We discuss the utility and the efficiency of a previous model developed for dominant markers. We show that dominant markers provide very poor information in the case of segregation distortion and, therefore, should be used with circumspection. On the other hand, the estimation of recombination fractions between codominant markers is less affected by selection than is that for dominant markers. We also discuss the analysis of linkage between dominant and codominant markers.