If you are at a county fair with a friend, and you get separated somehow, how do you reunite with your friend? Most likely, you go someplace, like the main gate, or a big attraction, hoping that your friend will also think to go there. In short you are attempting to coordinate with your friend, but you can not communicate with him. You are trying to predict where he will go, and (you hope) he is trying to make a similar prediction about you. Often (though not as often as we might like) we solve these sorts of problems fairly easily.

However, it's not entirely clear how we do it. When you attempt to coordinate with another person, you must decide what to do based on what you expect that person to do. At the same time, you know that she is also attempting to predict what you will do. This seems to wind up in a vicious circle of you trying to predict what I'm trying to predict that your trying to predict that I....

Thomas Schelling suggested that we solve this sort of problem by concentrating on a focal outcome: a choice that is salient, unique or prominent in some way (Schelling, 1960).

He asked people to play what he called tacit coordination games, in which they had to coordinate with a parnter with whom they could not communicate. For example, subjects were told to name a letter of the alphabet, or a number, in the hopes of matching the answer given by their partner. Or they were told to imagine they had to meet a stranger in New York City without being given a specific location, and told to name the place where they would attempt to rendezvous with their unknown partner. The most common answers for the first two games were "a" and "1". For the New York problem, the most common answer was the clock at Grand Central Station. These answers are, in some sense, "obvious", but its not clear why they are obvious.

We are investigating this question by observing people while they play tacit coordination games. The difference between our games and Schelling's is that ours have a more definite mathematical structure. Subjects are presented with threee dots on a line, and asked to pick one of the dots. The stimuli are designed so that either the "central" or the "most isolated" dot are salient to varying degrees. We want to see if subjects can coordinate their choices in this game, and if so, how they manage to do so.