The main objects here are nonlinear dynamic systems, stemming from sto
chastic models of economic growth. The issue is whether and how agents
can identify their best behavior in the long run, knowing neither the
system dynamics nor the exact law of the random variables. To come to
grips with this problem we model a process of adaptive adjustments, a
nd provide reasonable conditions under which the resulting tatonnement
converges to optimal behavior. (C) 1998 Elsevier Science B.V. All rig
hts reserved.