This is a fairly simple question if you think about the principals behind it. Let me explain.
First however, let's assign the alleles some symbols.
T - is the dominant allele (Upper case is always dominant by convention)
t - is the recessive allele (lower case is always recessive by convention)
P1: TT x tt (Parental generation, one parent is homozygous dominant for the trait in question, the other is homozygous recessive.)
You can draw up a punnet square if you wish to find out the probabilities of which alleles the kids will have but due to one parent being homozygous recessive, all children will have one dominant allele and one recessive allele. Since they have one dominant allele, they all show the dominant phenotype. We'll pretend dominant T produces black hair and recessive t red hair. So, in this case, all children have one dominant T allele, so, all have black hair.
F1: Tt, Tt, Tt, Tt
Now, we'll select one of the heterozygous children - Tt.
This person is in the 2nd generation, one of their parents had black hair (TT), the other parent had red hair (tt).
They are heterozygous for this trait. They have black hair but they also carry one allele for red hair, as shown above (Tt).
This child later goes onto marry another person(from another family) who is either heterozygous or homozygous recessive for this trait.
Case 1: Heterozygous partner
P:Tt x Tt
F1:TT/Tt/Tt/tt
25% chance of the child being homozygous recessive for the trait (tt) and thus displaying the recessive trait after 'skipping' a generation.
Case 2:Homozygous partner
P:Tt x tt
F1:Tt/Tt/tt/tt
50% chance of child being homozygous recessive for the trait (tt) and displaying the recessive trait after 'skipping' a generation.