Fitch's Algorithm

What Parsimony Leaves Out

Minimum-mutation scoring is useful, but it is not guaranteed evolutionary history.

highlighted = computed this step

Minimum mutations are not guaranteed history

Parsimony finds a minimum-mutation explanation on the given tree. Homoplasy and convergent changes can hide repeated history, and long-branch attraction can favor a misleading topology.

minimum mutations≠certain history\text{minimum mutations} \ne \text{certain history}
Parsimony model boundaryThe table separates what parsimony optimizes from what biology may require.Parsimony givesIt leaves openminimum mutation counthistorical truthscore for a given topologywhich topology is trueone equally good labelingbranch lengths and probabilities

The score is not a full evolutionary model

Fitch scoring assumes the topology is already given. It gives no branch lengths or probabilities, and ties can leave multiple equally parsimonious ancestral labelings.

parsimony score≠probabilistic model\text{parsimony score} \ne \text{probabilistic model}
Parsimony model boundaryThe table separates what parsimony optimizes from what biology may require.Parsimony givesIt leaves openminimum mutation counthistorical truthscore for a given topologywhich topology is trueone equally good labelingbranch lengths and probabilities