We run an experiment about a coordination game dealing with a strategic approach to bank run. We vary the size of the banks but keep fixed the tightness of the coordination problem faced by depositors. We explore the determinants of both the decisions and the experimentation rates of the participants. The probability to withdraw is higher in large banks, even though in the final rounds of all treatments the probability reaches very high levels. On the other hand, the experimentation rates are mainly affected by the distance between the share of expected players who withdraw and the parameter that measures the coordination tightness. Experimentation vanishes when this distance gets too large.