A central result in the rational learning literature is that if the true me
asure is absolutely continuous with respect to the beliefs then, given enou
gh data, the updated beliefs merge with the true distribution. In this pape
r, we show that, under absolute continuity, weak merging occurs fast (at th
e rate 1/root t) with density one. Moreover, if weak merging occurs fast en
ough (at the rate 1/root t) then absolute continuity holds. These rates are
sharp. We also show that, under some conditions, if weak merging occurs at
the rate 1/root t then absolute continuity holds.