I tried to find it with no sucess

Hi,

this reference might be relevant, and/or some citations in it.

The size of the character state space affects the occurrence and detection of homoplasy: Modelling the probability of incompatibility for unordered phylogenetic characters

By:Cuthill, JH (Cuthill, Jennifer Hoyal) . JOURNAL OF THEORETICAL BIOLOGY Volume:366 Pages:24-32 DOI:10.1016/j.jtbi.2014.10.033 Published:FEB 7 2015

Good hunt, Lauri

You get that problem all the time with morphological data: Loose definitions of character states result in a wide range of morphologies considered as similar which would be regarded as incompatible if more strict criteria for similarity had been applied. As a consequence the measured amount of homoplasy (in a cladistic analysis) is higher in case of loose definitions and few incompatibilities than for strict definitions and many incompatibilites. But that appears to be rather trivial ...

Michael,

all very true. But the practical problem may be - and in my case often is: I can guess that thing A and thing B are not homologous, but these things must be anyway defined in a way that can be defended. I.e. I have, in the absence of e.g. ontological data coded a similar structure at the same place with the same code. If I do not do so, I am guilty of a) 'knowing' things before analyses, and b) being unable to defend why these identical-looking things are coded differently. This is inherent in probably all morphological matrix building, perhaps  apart the most simplistic ones. What comes to 'too loose definitions of character states' I see them potentially equally subjective. Too strict definitions certainly cause a lot homoplasy as well. What is left is to be subjective, but in as educated manner as one can. Shapes or position of structures -even if truly homologous, may vary - to what extent to split the variation? Excessive amount of multistate characters may blur the phylogenetic signal (homology of the structure itself), as well. Well of course, there are methods for application of continuing characters, but in practice it may prove quite impractical to sacrify huge amounts of time to obtain statistics enough to use many such characters - if even there are material enough available. Hmm, I may drifted a bit out from the original question, sorry.

I don't know the reference, but it seems to me that perhaps what Farris was talking about is the idea that a character can have many homoplastic autapomorphies (these are not incompatible with the MP tree).  To be incompatible, a character needs to support a hypothesis of grouping that is different than the MPT.  Autapomorphies do not change the tree length on any tree, and support no group, so they are not incompatible, even if they are homoplastic. 

But what do I know?  Martin, you should ask Steve Farris.  His email address is listed at the end of any of his recent publications in Cladistics.

Hi Jan,

Thanks for giving me a possible out, but I think I was just confused.  What I said above is not correct.  Can you answer this for me:

A character is either compatible or not compatible with the most parsimonious tree, right?  Are there degrees of compatibility for an individual character?

Thanks for the explanation.  The issue is clearer to me now.  I have not thought much about cliques in a long time.