Cheap talk with simultaneous versus sequential messages

Recent experimental studies find excessive truth-telling and excessive trust in one sender/one receiver cheap talk games with an essentially unique and babbling equilibrium. We extend this setup by adding a second sender into the play and study the behavior of the players both theoretically and experimentally. We examine games where senders are assumed to communicate with the receiver either simultaneously or sequentially as well as a game where the receiver chooses one of these two communication methods. The theoretical predictions for truth-telling, non-conflicting messages observed and trust frequencies are the same for both the simultaneous and sequential plays; however, we observe systematic differences between the treatments of these plays. While the truth-telling frequencies stay above the theoretical prediction of the one half during all the experiments, the nature of truth-telling seems to differ between sequential and simultaneous plays. Under simultaneous communication, the messages of senders are non-conflictive more than half of the time, while the non-conflicting messages are significantly more likely to be correct than not. The frequency of non-conflicting messages is lower under sequential plays due to the tendency of the second sender to revert the message of the first sender. We observe that subjects who prefer to get non-conflicting messages prefer simultaneous mode of communication more often. When acting as senders, these subjects also adjust their truth-telling frequencies so as to generate conflictive messages.