There is a substantial literature examining coordination in public goods games. We conducted an experiment to explore how varying patterns of thresholds affect the willingness of subjects to contribute to a public good. We had subjects play a multiperiod game where each subject was allocated an initial point endowment, told a threshold for the group and had to choose how much to contribute to the common pot. Each period is identical, except for the possibility of having a different threshold, which is always stated before the players make their contributions. We found that while contributions are similar for the increasing and decreasing threshold group types when thresholds were low, a sizeable gap opens up around the average threshold size. We found that for nearly every threshold, it is more profitable to be in an increasing than in a decreasing threshold group type. Early cooperation seems to facilitate the achievement of harder-to-reach thresholds, which require considerable contributions from all members of the group. These findings are also very robust in the regression specifications. Our findings shed light on the role of past cooperative success and threshold patterns on subsequent willingness to cooperate.