This paper reports the results of a field experiment testing the effectiveness of different quality get-out-the-vote (GOTV) non-partisan phone calls. During the week preceding the November 2004 election, we randomly assigned registered voters in North Carolina and Missouri to one of three live phone calls with varying length and content. The scripts are (1) standard GOTV, (2) interactive GOTV, and (3) interactive GOTV with a request for mobilizing neighbors. We find that people assigned to the interactive GOTV treatment are more likely to turn out, while the effect of the “get your neighbors to vote” script is relatively as weak as that of the standard script. The findings suggest that interactive phone calls generally tend to increase voter turnout, but in order for a phone call to be effective, the message needs to be focused. The borderline statistical significance of the script that encourages neighbors’ participation invites replication of this experiment.