geoR's Sigma-sq and Tau-sq

Hi:
By using

likfitobject$info.minimisation.function

after fitting data using likfit one can see
that the hessian has 2 dimensions less than the
number of parameters reported in

summary(likfitobject)

I already know that beta is computed from
the other parameters. The remaining 
'missing' dimension must be because
the optimization routine actually estimates
the ratio between sigma-sq and tau-sq rather
than each one separately. So how does geoR 
recover the values of sigma-sq and tau-sq reported 
in the summary? I have tried multiplying the estimated 
ratio and the total variance in the original data (transformed
with the mle of lambda) but that does not give
me the mles of sigma-sq and tau-sq reported in
summary.
Thanks for clarifications
Ruben