From ErrorEstimates.bf with the command

CovarianceMatrix (cmx, theLnLik);

I'm getting the error:

Error:LUDecompose doesn't work on singular matrices (row 33)
Call stack
1 : Calculate the Covariance Matrix storing into, cmx, the following likelihood function:theLnLik ;

I get the same error with the MPI, MP and SP versions of the svn downloaded around March 3, 2009.

I reduced the number of taxa from around 80 to around 25 and I didn't get the error.  I'm guessing there is a branch that has an omega at a high or low extreme with zero variance and covariance?

Dear Danny,

Yeah, sounds like the numerical Hessian is singular.
Try using the profile likelihood approach.

Cheers,
Sergei