R help - Jul 2003 - glmmPQL with crossed random effects

Is it possible to fit a crossed random effects model
using glmmPQL and, e.g.
random=pdBlocked(list(pdIdent(~a-1),pdIdent(~b-1))) as
the random part of the model?  I seem to be getting
errors on R using glmmPQL in this way, although if I
try something very similar using 'lme' instead of
'glmmPQL', it works fine.  Does the data have to be of
a groupedData form for it to work (as it has to be for
lme) because glmmPQL doesn't seem to work with data of
the form groupedData?