Terry, You sound awfully dogmatic about your position below? What leads to such surety? By constructing a tree I hope we are meaning the same thing --- estimating the location in time of the tree nodes and perhaps their connectivities. If you reread my post, it suggests just the opposite of your final sentence/question. The purpose of bringing in the (unseen) haplotypes at the tree nodes may be a required ingredient of a more accurate method of tree reconstruction; not that reconstructions can not be made otherwise --- so I don't necessarily disagree with the very first part of your first sentence. I'll try to reconstruct a simple case tomorrow which can be exhaustively analyzed. From: Terry Sent: Sunday, March 25, 2012 10:42 PM Ken, One can construct a tree just fine without going through the unstable approach of attempting to figure out what some haplotype might, or might not be, for any particular ancestor. You may choose to infer the value of an ancestor's haplotype somehow by a separate method, like taking some type of "average" of the input haplotypes or something, but I can't see that helping in any new method. The quality of modern tree reconstruction methods is not in doubt, so why lower that quality by forcing into the computation a potentially incorrect inference for the value of all past haplotypes. Terry