Unit content
Maximum parsimony for comparing phylogenetic tree hypotheses
When several phylogenetic trees could explain the same character data, maximum parsimony prefers the tree requiring the fewest inferred evolutionary changes.
This is not the claim that evolution always follows the simplest path. It is a rule for comparing hypotheses given a character model.
Worked example
Suppose an ancestral state is 0 and three species have:
| Species | Character X | Character Y |
|---|---|---|
| A | 0 | 0 |
| B | 1 | 1 |
| C | 1 | 1 |
Consider two candidate trees:
- $(A,(B,C))$
- $(B,(A,C))$
On tree 1, each character can be explained by one change from 0 to 1 on the branch ancestral to B and C: two changes total.
On tree 2, explaining B and C both having state 1 while A has 0 generally requires separate changes or a change plus reversal for each character. The total is larger.
Maximum parsimony therefore favors tree 1 for these data.
Parsimony can be misled by convergent evolution, reversals and very different rates of change among lineages. Modern phylogenetics therefore often uses explicit statistical models instead. Nevertheless, parsimony teaches the central inference problem clearly: a tree is evaluated by how economically it explains the distribution of observed character states.