Indicative versus Imperative Meaning in Cheap-Talk Games: An Experiment

In a sender-receiver cheap talk game, where a sender, privately informed about a type, sends a message to a receiver, who then chooses an action, there is no concern for a meaning of the messages. Messages are used as an abstract device to pair the sender's types and the receiver's actions, and identical pairings might be obtained by swapping all the messages. However, anecdotal evidences and some experiments (e.g. Blume, 2022) suggest that messages might have an extraneous meaning (i.e. refer to something beyond their ``meaning in use''). The question we ask is whether such an extraneous meaning, if any, might be related to some strategic aspect of the played game. We design an experimental protocol that allows us to test whether the players of a cheap talk game endow the messages used with either an indicative or an imperative meaning, where indicative (resp. imperative) refer to whether the observed message is unequivocally associated with a given type (resp. a given action). We compare two treatments in order to determine whether one meaning emerges rather than the other, according to the existence of a conflict between the players (as to which action to associate with which type). On the other hand, we compare two other treatments, in which players can only coordinate on an imposed meaning, in order to test, in a second step, whether one meaning lasts more once a conflict is introduced. The experiment was carried out in February 2024 with 386 participants. The data collected are currently being analyzed. First results suggest that the considered meanings do are related to coordination and conflict.