This is very straightforward to do using parsimony.  You use a downpass/uppass approach and do it by hand (if two terminals have the same state, then their most recent common ancestral node probably had that state, too), or you can probably also use the MacClade package of Mesquite to accomplish the same end.

Hi,

You don't always need branch lengths to reconstruct ancestral character states; it depends on how your topology was built and the type of analyses you want to perform on the tree.

I would recommend you read this : http://sites.biology.duke.edu/cunningham/pdfs/science%202.pdf

Good luck !

Thank you very much, Vincent.

I will read the paper you recommended.

Thank you very much, Andrew.

Yes, I have tried mesquite, now what I want to do is making a compare between different methods.

You mean morphological characters? If you know the topology of the groups thus you know the set of "group" characters and their rows of evolution. Within the groups you may differ the common species characters which have similar rows of evolution in neighbor groups (parallelisms), and the characters which may have the unique rows for one species group only. I think you ask about such unique character series, is it so?    

hello, Alexander. Actually, the character is genome size (or 2C).

Do I understand you correctly that you have, or could make, a phylogeny at least for some of the species, and that you have a continuous trait (2C)? If so, one possibility is to set up a Bayesian phylogenetic analysis with sequences where available and missing data for the rest, alongside the trait matrix to be optimised on the phylogeny. You could also constrain taxon groups as monophyletic where you have this information. Such an analysis is well possible to set up in e.g. BEAST and would (in my opinion) outperform a parsimony analysis by far. 

thank you, Thomas. I will consider your suggestion and wish to have a better inference.

Actually, the main problems in this project is the molecular phylogeny relationship  :

1, the sequence data is not available for most of the species I studied;

2, BUT, my species could clustered into several groups, and I have  sequence data of other species in those groups; 

so, available sequence doesn't match at species level, but group level.

I wish you could correctly understand, thank you again. 

yes