Beside the fact that maximum parsimony is computationally cheap, what other good arguments are there for it? Is there any model behind this principle? Why would one expect this principle to provide right phylogeny in any situation at all ?