This paper reports the results of an experiment designed to test the usefulness of alternative solution concepts to explain players' behavior in noncooperative games with preplay communication. In the experiment subjects communicate by plain conversation prior to playing a simple game. In this setting, we find that the presumption of individualistic and independent behavior underlying the concept of Nash equilibrium is inappropriate. Instead, we observe behavior to be coordinated and correlated. Statistical tests reject Nash equilibrium as an explanation of observed play. The coalition proof correlated equilibrium of the game, however, explains the data when the possibility of errors by players is introduced.