In a previous paper, we presented a quantitative genetic model of the evolu
tion of genomic imprinting in mammals - differential gene expression of an
embryonic growth factor gene depending on the parental origin. We have show
n that the verbal argument of the genetic conflict hypothesis can be justif
ied by a formal genetic model. The model predicts the evolution of an expre
ssed paternal allele and a silent maternal allele for a growth-enhancing ge
ne increasing maternal resource acquisition, while the reverse pattern will
evolve for an inhibitor gene. This may, however, be prevented if there are
recessive deleterious mutations on coding regions in the population. In th
is paper, we examine potential problems for the genetic conflict hypothesis
and discuss how the theory can account for them. First, we show that the r
everse pattern of genomic imprinting of a growth factor gene can evolve if
the risk of abortion in early gestation is enhanced by the overproduction o
f the growth enhancer. Such a pattern is observed in Mash2. Second, paterna
l disomies (double dose of paternal origin and none from maternal origin) w
ith respect to a part of a chromosome sometimes result in a decrease (rathe
r than an increase) in embryo size. Here we show that this can be explained
if the imprinted genes regulate the allocation between placenta and embryo
proper by modifying the developmental fate of cells. Third, an alternative
non-conflict hypothesis is studied in the same modelling framework, which
states that genomic imprinting has evolved because it reduces the risk of s
pontaneous development of parthenogenetic embryos that cause serious risk t
o the mother's life (ovarian time-bomb hypothesis). Finally, we point out t
hat genes on the X chromosome receive selection different from autosomes, r
esulting in the evolution of genomic imprinting in the opposite direction.