Genomic Imprinting is a very interesting phenomenon that has been observed starting from both the classical genetics and the molecular genetics points of view. 

In the first setting, characters were discovered (primarily in Mammalians and in Plants) that do not follow the classical mendelian mode of inheritance. Specifically, some characters are expressed in the progeny only if inherited from one of the parents.

For example, the first character observed in the mouse in 1974 was a deletion at the Tme locus on Chr. 17 that showed in utero lethality only when transmitted by the mother and not if transmitted by the father. In other words, this means that the locus from the mother was indeed essential and could not be compensated by a normal allele from the father, which apparently can not be used in the embryo. 

Just as a joke, we can see this (invented) example. The allele "long cilia" can be carried by either Mickey Mouse or Minnie. However, the phenotype is seen only if it comes from Mickey Mouse.

but 

From the Molecular Genetics point of view, on the other side, this problem has to do with the following question: as a norm, is gene expression bi-allelic or mono-allelic ?   Of the two copies of each gene that we have, is only one expressed or both of them ?

The answer is: most genes are expressed bi-allelically.