We introduce Bayesian learning into a stochastic growth model and study the effect of experimentation on the optimal level of the investment decision and on the amount of information gathering. When more investment produces more information, experimentation can push towards a lower level of investment, thus reducing information acquisition. Symmetrically, when more investment reduces information, experimentation can increase investment and again decrease information gathering. These results run contrary to intuitions generated by the standard literature on experimentation. (C) 1998 Elsevier Science B.V. All rights reserved.