While we're on the subject of tree-building from sequence data, has anyone tackled the estimation of error in branch-length calculations? References appreciated. Rick rbm8p at virginia.edu