In an experimental investigation of Milgrom and Roberts' (1982) model,
play consistently converges to a unique equilibrium, providing eviden
ce of sophisticated strategic behavior that the theory predicts. Play
starts with monopolists at their myopic maxima, followed by an attempt
to pool, and then lif ilo pooling equilibrium exists) separation, sug
gesting myopia rather than a forward-looking process. When both pure-s
trategy pooling and separating equilibria exist, equilibrium selection
is a function of the past history of play. An adaptive learning model
characterizes the major features of the data and provides predictions
of intermediate-term behavior that is otherwise lacking in the theory
.