In collaboration with local election officials, we conducted a randomized field experiment in which postage-paid envelopes were provided to a random sample of 10,000 permanent vote-by-mail (VBM) voters in San Mateo County, California, in advance of the November 2, 2010, general election. We find that the treatment generated statistically significant but unexpected effects: postage-paid envelopes increased the probability that voters cast their ballots in person and decreased the probability that they cast their ballots by mail. These offsetting effects meant that the intervention produced no net change in voter turnout. We find that this pattern of countervailing effects is strongest among voters who frequently voted by mail in the past, those potentially most susceptible to disruptions in routine. Post-election interviews support the idea that the postage-paid envelopes created confusion for some voters. The results suggest that reforms designed to increase turnout by decreasing voting costs may have the unintended effect of disrupting routines.