Posterior distribution over a countable setMof continuous data-sampling distributions
piles up at
-projection of the true distribution 
on 
, provided that
the -projection is unique. If there are several -projections of on , then the
posterior probability splits among them equally.