Abstract: This paper reports an experiment conducted to examine the effects of changing the form of strategic interactions (Bertrand or Cournot), and changing underlying information conditions on the usefulness of the 'ALM,' a merger simulation device used by the U.S. Department of Justice. The experiment consists of twenty 60-period quadropolies. Aftr period 40 a merger is induced. Behaviorally, Cournot markets are much more variable than Bertrand markets. Also, providing information about others' actions and earnings dampens merger-related comparative statics effects. Consequently, the ALM predicts post-merger performance poorly. Nevertheless, results do not undermine a previous finding that the ALM screens out nonproblematic consolidations.