Group tasks are often organized by a list: group members state their willingness to contribute by entering their names on a publicly visible, empty list. Alternatively, one could organize the group task by starting with a full list: every group member is already entered on the list and non-cooperators have to cross out their names. Indeed, strong behavioral differences are observed when comparing (otherwise identical) environments with empty and full lists in a laboratory experiment with repeated interaction. Cooperation in the empty list is high in early periods, but is decreasing. In the full list, cooperation starts low, but is actually increasing, surpassing cooperation in the empty list treatment in later periods. Two factors, diffusion of responsibility and unraveling of cooperation seem to drive the results.