Reconstructing Ancestral States

[Elanora Smith]

I am using the Trace Character/Reconstruct Ancestral States function with the Likelihood Ancestral States mode. I have 8 "characters" aka geographic regions and I have a matrix of each species with a "1" or a "0" corresponding to if that species is present or absent in that region. For one of my characters/regions, the same exact probability of presence was generated at every single node. With the Mk1 model it gives a 50% prob and with the Asymm2 model it gives a 61% prob. Does this just mean that the data/model do not support the likelihood of any ancestor being in this region? 

[Wayne Maddison]

Hi -- Without seeing your tree and the character distributions, it's hard to know whether this result is unexpected. However, it certainly is possible that a reconstruction under likelihood could be fully ambiguous. This would be especially true if you let the parameter(s) of the model free to vary, because the highest likelihood may be at very high rates of change (especially if there's a very short branch somewhere in the tree), which can lead to full ambiguity.