This study investigates a potential mechanism to promote coordination. With theoretical guidance using a belief-based learning model, we conduct a multi-period, binary-choice, and weakest-link laboratory coordination experiment to study the effect of gradualism - increasing the required levels (stakes) of contributions slowly over time rather than requiring a high level of contribution immediately - on group coordination performance. We randomly assign subjects to three treatments: starting and continuing at a high stake, starting at a low stake but jumping to a high stake after a few periods, and starting at a low stake while gradually increasing the stakes over time (the Gradualism treatment). We find that relative to the other two treatments, groups coordinate most successfully at high stakes in the Gradualism treatment. We also find evidence that supports the belief-based learning model. These findings point to a simple mechanism for promoting successful voluntary coordination.