Decisions about future behaviors are clearly shaped by thecontentof past experiences, but whether theorderof past experiences matters remains controversial. By analyzing the largest field experiment about prosocial behavior to date, a natural field experiment involving 14,383 volunteer crisis counselors over five years, we examine how the content and order of past experiences causally influence decisions about future behaviors — whether individuals continue volunteering or quit. Volunteers were repeatedly and randomly assigned to perform 1,976,649 prosocial behaviors that were either harder (suicide conversations) or easier (non-suicide conversations). We found that thecontentof past experiences mattered: Harder (versus easier) behaviors encouraged quitting. However, theorderof past experiences mattered far beyond their content alone: Harder behaviors caused disproportionately more quitting if they came in long “streaks” or at the “end.” These “streak”/“end” effects reveal important practical insights for leaders and policymakers seeking to boost prosocial behavior. For instance, a reordering intervention — assigning behaviors so as to avoid creating hard “streaks” — would reduce volunteer quitting rates by at least 22%, boosting prosocial behavior and likely saving lives.