Economics lacks a good theory of the pricing and output decisions of a
monopolist which does not know its demand-we inevitably assume that t
he monopolist knows much more about demand conditions than is reasonab
le. I present a model in which demand information is generated endogen
ously. When information is endogenous the monopolist has an incentive
to experiment with price and quantity. I derive the direction of exper
imentation, characterize an important value function arising from dyna
mic programming problems with learning, and relate the results to Blac
kwell's comparison of experiments.