In an earlier paper [Shah Journal of Mathematical Economics, 1995, 24(
5), 461-495], we studied Bayesian learning in an intertemporal, stocha
stic setting. The results stated in the present paper are significant
generalizations and simplifications of the earlier results. We study a
sequence of games in which the stochastic kernel that links the state
variable is unknown. Successive stage games are played by successive
generations of identical players. Each generation makes a rational for
ecast of the state, given their belief about the true kernel and the e
quilibrium implemented by the players. Beliefs are updated by Bayes' r
ule after observing the actual state. Our present results concern the
dynamic behavior of the forecasts process. We show there is partial le
arning in the sense of being able to (asymptotically) predict the futu
re as well as an omniscient modeler who knows the true transition stru
cture and understands the stochastic evolution of the state and player
s' actions. Moreover, we characterize the limiting behavior of the for
ecasts process in terms of the set of full information rational expect
ations forecasts.