There are three possible states for the canoes to be in. Both on shore A, one on each, and both on shore B. Let the probabilities of each be p(AA), p(AB) and (BB), respectively.
When many villagers have crossed the river, there is some probability for the canoes to be in each of those three states. When many plus one villagers have come and gone, the probabilities are the same.
if p(AA), p(AB) and p(BB) are the situation after n people, the situation after n+1 people are:
p(AA) = ½p(AA) + ½p(AB)
p(AB) = ½p(AA) + ½p(BB)
p(BB) = ½p(BB) + ½p(AB)
For example, for both canoes to be on shore A after the n+1st villager they needed to be either on shore A already or on separate shores, and then the n+1st villager crossed from B to A. Solving that given p(AA) = p(AB) = p(BB) = 1/3. So, a person need to holler with a chance of one in three, not one in four.
With three canoes, the states are AAA, AAB, ABB and BBB. Making a similar system of equations gives p(AAA) = p(AAB) = p(ABB) = p(BBB) so the likelihood of needing to holler is one in four.
This is an example of the Markov chain.